[{"data":1,"prerenderedAt":3237},["ShallowReactive",2],{"sitemap-blog":3,"sitemap-news":2151,"sitemap-cases":3069},[4,10,15,21,27,33,39,44,49,54,59,64,69,74,79,84,90,95,100,105,110,115,120,125,130,135,140,145,150,155,160,165,170,175,180,185,190,195,200,205,210,215,220,225,230,235,240,245,250,255,260,265,270,275,280,285,290,295,300,305,310,315,320,325,330,335,340,345,350,355,360,365,370,375,380,385,390,395,400,405,410,415,420,425,430,435,440,445,450,455,460,465,470,475,480,485,490,495,500,505,510,515,520,525,530,535,540,545,550,555,560,565,570,575,580,585,590,595,600,605,610,615,620,625,630,635,640,645,650,655,660,665,670,675,680,685,690,695,700,705,710,715,720,725,730,735,740,745,750,755,760,765,770,775,780,785,790,795,800,805,810,815,820,825,830,835,840,845,850,855,860,865,870,875,880,885,890,895,900,905,910,915,920,925,930,935,940,945,950,955,960,965,970,975,980,985,990,995,1000,1005,1010,1015,1020,1025,1030,1035,1040,1045,1050,1055,1060,1065,1070,1075,1080,1085,1090,1095,1100,1105,1110,1115,1120,1125,1130,1135,1140,1145,1150,1155,1160,1165,1170,1175,1180,1185,1190,1195,1200,1205,1210,1215,1220,1225,1230,1235,1240,1245,1250,1255,1260,1265,1270,1275,1280,1285,1290,1295,1300,1305,1310,1315,1320,1325,1330,1335,1340,1345,1350,1355,1360,1365,1370,1375,1380,1385,1390,1395,1400,1405,1410,1415,1420,1425,1430,1435,1440,1445,1450,1455,1460,1465,1470,1475,1480,1485,1490,1495,1500,1505,1510,1515,1520,1525,1530,1535,1540,1545,1550,1555,1560,1565,1570,1575,1580,1585,1590,1595,1600,1605,1610,1615,1620,1625,1630,1635,1640,1645,1650,1655,1660,1665,1670,1675,1680,1685,1690,1695,1700,1705,1710,1715,1720,1725,1730,1735,1740,1745,1750,1755,1760,1765,1770,1775,1780,1785,1790,1795,1800,1805,1810,1815,1820,1825,1830,1835,1840,1845,1850,1855,1860,1865,1870,1875,1880,1885,1890,1895,1900,1905,1910,1915,1920,1925,1930,1935,1940,1945,1950,1955,1960,1965,1970,1975,1980,1985,1990,1995,2000,2005,2010,2015,2020,2025,2030,2035,2040,2045,2050,2055,2060,2065,2070,2075,2080,2085,2090,2095,2100,2105,2110,2115,2120,2125,2130,2135,2140,2145,2148],{"title":5,"description":6,"date":7,"category":8,"path":9},"被 AI 引用之后，用户点进来落在哪一页","引用带来的点击通常落在被引用页面，而那一页往往没有转化入口。这篇写承接断在哪、怎么在不打扰阅读的前提下加下一步，以及怎么测量。","2026-11-07","SEO","\u002Fblog\u002Fcitation-click-landing",{"title":11,"description":12,"date":13,"category":8,"path":14},"词库半年不维护就废了，一半的词已经没人搜","2026 年 4 月我们重看 2025 年的词表，870 个词里 210 个已写、130 个零曝光、60 个词义变了。词库不是建一次就完事，我们固定每月小更新、每季度重算。","2026-11-06","\u002Fblog\u002Fkeyword-library-maintenance",{"title":16,"description":17,"date":18,"category":19,"path":20},"内容的半衰期：哪些文章三年后还在带量","内容有半衰期，有的三个月就没流量，有的三年还在涨。我们跟踪了 120 篇内容的流量曲线，分出三类：快消型、稳增型、长尾型。半衰期怎么判断、怎么用它做排期都在这篇，数据来自我们对 120 篇内容的两年跟踪。","2026-11-05","内容运营","\u002Fblog\u002Fcontent-half-life",{"title":22,"description":23,"date":24,"category":25,"path":26},"国内三大运营商 4G 实测：同一页能差几秒","三个城市、三家运营商、同一台中端机，测同一批页面的首屏。差异最大的一组差了 1.7 秒。这篇写测试方法、差异来源，以及官网该按哪条线验收。","2026-11-04","工具评测","\u002Fblog\u002Fchina-mobile-network-field-test",{"title":28,"description":29,"date":30,"category":31,"path":32},"从 HTTP 迁到 HTTPS：技术 SEO 的 12 个检查项和两个坑","HTTPS 迁移看着简单，做错的代价不小：证书链、混合内容、301 配置、HSTS，每一项都可能让收录掉一截。我们迁移过 15 个站，把检查项整理成 12 条，按迁移前、迁移中、迁移后分好，照着过一遍能少踩坑。最后写了两个我们踩过的坑。","2026-11-03","技术SEO","\u002Fblog\u002Fhttps-migration-checklist",{"title":34,"description":35,"date":36,"category":37,"path":38},"跟老板汇报 SEO，他要听的不是排名是生意","一页、三个数、两个季度。这篇写面向老板的汇报结构：自然搜索带来多少线索、单条线索成本多少、下季度赌什么，以及哪些话不要讲。","2026-11-02","行业观察","\u002Fblog\u002Freport-seo-to-executives",{"title":40,"description":41,"date":42,"category":31,"path":43},"算法更新监测表：我们每周记录的 9 个指标","算法更新能不能应对，取决于有没有一套固定的监测表。我们每周记录 9 个指标，两年没断过。表的结构、每个指标的用法、怎么读数据都在这篇，可以直接照着搭，半小时能填完一周，照着做一周就能上手。","2026-11-01","\u002Fblog\u002Falgorithm-monitoring-dashboard",{"title":45,"description":46,"date":47,"category":37,"path":48},"对比页的效果怎么算：它很少是最后一跳","对比页的直接转化率只有 1.1%，看起来不如产品页，但把 90 天路径拉出来，28% 的成单在签约前访问过它。只看最后点击会严重低估这类页面的作用。","2026-10-31","\u002Fblog\u002Fcomparison-page-conversion-measure",{"title":50,"description":51,"date":52,"category":37,"path":53},"接站先看五个信任页面，缺一个都算风险","信任页缺失不影响收录，但影响转化和品牌词结果页。这篇写五个页面各查什么、缺失怎么补、以及为什么它和收录问题并列在诊断清单第一页。","2026-10-30","\u002Fblog\u002Ftrust-page-audit-checklist",{"title":55,"description":56,"date":57,"category":37,"path":58},"GEO 要不要单独列预算，先算清这三笔账","我们建议别设 GEO 专项预算，而是在内容和技术里各切一块。这篇写三笔账怎么算、常见的两种浪费，以及一个可参考的分配比例，附一套和老板沟通的拆法。","2026-10-29","\u002Fblog\u002Fgeo-budget-three-accounts",{"title":60,"description":61,"date":62,"category":19,"path":63},"内容 KPI 分四层，只看最后一层会做错决定","只看注册数，会把该做的事全砍掉。我们把内容考核分四层：产出、可见、互动、商业。前两层看新项目，后两层看成熟项目，权重我们定成 2:3:2:3。","2026-10-28","\u002Fblog\u002Fcontent-kpi-layers",{"title":65,"description":66,"date":67,"category":19,"path":68},"批量内容的质量底线：模板化到什么程度算过头","批量生产内容不是原罪，模板化过头才是。我们给批量内容定了四条底线：每篇有独家信息、结构可以有模板、案例不能重复、数据必须核实。一个批量 200 篇翻车的案例和我们的检查表都在这篇，四条底线要守住。","2026-10-27","\u002Fblog\u002Fcontent-batch-quality",{"title":70,"description":71,"date":72,"category":31,"path":73},"静态化加边缘缓存，把首屏压进一秒内","一个文档站改成静态生成加边缘缓存，首字节从 900 毫秒降到 120 毫秒。这篇写哪些页面能静态化、缓存怎么失效、内容更新怎么触发。","2026-10-26","\u002Fblog\u002Fedge-cache-static-generation",{"title":75,"description":76,"date":77,"category":31,"path":78},"内链锚文本：同一个词链 40 次反而没用","内链锚文本不是关键词堆得越多越好。我们做过对比：同一个目标页用 40 个相同锚文本链接，排名不如用 8 个自然锚文本的页面。三个原则、一次实验数据和一个常见误区都在这篇，看完能立刻对照着改。","2026-10-25","\u002Fblog\u002Fanchor-text-strategy",{"title":80,"description":81,"date":82,"category":37,"path":83},"客户说没效果，先和他对齐这三个词","大部分没效果是预期错位，少数是真的没做好。这篇写效果、周期、归因三个词怎么对齐，对齐之后如果还是不行，该怎么调整合同和交付。","2026-10-24","\u002Fblog\u002Fclient-says-no-results",{"title":85,"description":86,"date":87,"category":88,"path":89},"AI 概览上线后，哪些查询的点击真的没了","Google AI 概览上线后，信息型查询的点击确实在掉，但软件类产品词没受什么影响。从 2025 年 12 月起我们监测 40 个词、9 个月，按四类查询拆开看，掉多少、为什么、能做什么都写了。产品词反而稳，品牌词搜索量还涨了，值得细看。","2026-10-23","GEO","\u002Fblog\u002Fai-overviews-click-loss",{"title":91,"description":92,"date":93,"category":19,"path":94},"海外榜单页：自己写还是求人写，账不一样","榜单页是海外买家选型的第一站，但自己做榜单和进别人的榜单，成本和风险完全不同。我们两种都试过，进别人的榜单三个月带来每月 400 多次引荐流量。","2026-10-22","\u002Fblog\u002Fbest-of-list-pages",{"title":96,"description":97,"date":98,"category":8,"path":99},"别人提到你却没给链接，这件事值得追","无链接品牌提及是最好谈的外链机会。这篇写怎么用工具找到它们、第一封邮件怎么写、成功率大概多少、哪些提及不值得追，附我们半年的实际数据。","2026-10-21","\u002Fblog\u002Funlinked-brand-mentions",{"title":101,"description":102,"date":103,"category":37,"path":104},"GEO 多久能见效，我们观察是三个月起步","十二个项目里，最快的六周出现引用，多数在三到四个月。这篇写见效的三个阶段、四个变量、该看哪些早期信号，以及什么时候该判断失败。","2026-10-20","\u002Fblog\u002Fgeo-timeline-expectations",{"title":106,"description":107,"date":108,"category":8,"path":109},"零搜索量的长尾词值得写吗，我们只写三类","工具显示月搜 0 的词，我们站有 15% 的访问来自它们。但只写三类：报错原文、组合场景、新版本新工具。其余零量词，写了也是自娱自乐。","2026-10-19","\u002Fblog\u002Fzero-volume-longtail",{"title":111,"description":112,"date":113,"category":19,"path":114},"同一个功能，三家软件公司怎么写不一样","软件公司的内容会同质化，因为大家写的是“功能清单”。差异不在功能，在场景、代价和证据。我们给三个客户写同一个功能，三篇内容角度完全不同。三个差异化方向、示例和别做的事都在这篇，适合内容同质化的团队。","2026-10-18","\u002Fblog\u002Fcontent-differentiation",{"title":116,"description":117,"date":118,"category":31,"path":119},"老站迁到 HTTP\u002F2，先把这四个坑填了","协议升级本身不难，难的是升级之后旧的优化手法全部失效。这篇写域名分片、资源合并、服务端推送和 CDN 兼容四个坑，以及迁移后的验证方法。","2026-10-17","\u002Fblog\u002Fhttp2-migration-pitfalls",{"title":121,"description":122,"date":123,"category":31,"path":124},"HTTP 头也能做 SEO：X-Robots-Tag 与 Link 头的三个用法和坑","页面里的 meta 标签不是唯一入口，HTTP 头也能控制索引、传 canonical、声明多语言。X-Robots-Tag 和 Link 头用得对，能解决一批页面里改不了的问题；用错会互相冲突。用法、冲突场景和查法都在这篇，最后附了一个判断标准和两个踩过的坑。","2026-10-16","\u002Fblog\u002Fhttp-header-seo",{"title":126,"description":127,"date":128,"category":8,"path":129},"算法更新那周，我们给客户发一封什么信","波动第一天别发信，第三到第五天再发。这篇写怎么判断是不是算法更新、信里写哪五段、哪三句话绝对不能出现，以及之后的四周做什么。","2026-10-15","\u002Fblog\u002Falgorithm-update-client-letter",{"title":131,"description":132,"date":133,"category":8,"path":134},"关键词密度这个指标，2026 年可以彻底扔了","关键词密度是 2010 年代的指标。我们做过测试：同一篇文章密度从 0.8% 提到 2.3%，排名没有任何变化，反而影响可读性。现在该看什么、工具为什么还显示它，这篇讲清楚。我们拿 10 篇文章做了对照，结论很干净，覆盖 80% 以上就够了。","2026-10-14","\u002Fblog\u002Fkeyword-density-obsolete",{"title":136,"description":137,"date":138,"category":8,"path":139},"对比页的七个模块，顺序错了转化会掉一半","我们把一个客服系统的对比页按七个模块重排，内容几乎没改，申请试用从每月 46 涨到 80。对比页的顺序比内容更影响转化，结论和价格都要往前放。","2026-10-13","\u002Fblog\u002Fvs-comparison-page-structure",{"title":141,"description":142,"date":143,"category":37,"path":144},"融资和产品发布，媒体稿怎么写才有搜索价值","新闻稿发完就沉底，是因为只写了事件没写问题。这篇写媒体稿的三个可检索入口、标题怎么写、发出去之后还要做什么，以及发布前两周该准备什么。","2026-10-12","\u002Fblog\u002Ffunding-press-release-seo",{"title":146,"description":147,"date":148,"category":31,"path":149},"AI 爬虫来得勤不勤，能看出内容有没有人用","两个站对比，一个 GPTBot 日均一百二十次，另一个八次。这篇写频次怎么统计、高和低各说明什么、怎么和引用数据交叉比对，附一个四周内频次变化的例子。","2026-10-11","\u002Fblog\u002Fai-crawler-frequency-signal",{"title":151,"description":152,"date":153,"category":19,"path":154},"内容生产流水线：从选题到发布的 7 个环节，每个环节卡在哪","内容团队效率低，通常不是写得慢，是环节之间互相等。我们把生产拆成 7 个环节：选题、关键词、大纲、初稿、审稿、发布、复盘，每个环节定负责人和产出物。流程、卡点和一个跑顺的案例都在这篇，流程表可以直接抄。","2026-10-10","\u002Fblog\u002Fcontent-production-pipeline",{"title":156,"description":157,"date":158,"category":19,"path":159},"一个 300 篇的博客站，审完只留了 78 篇","2026 年 6 月我们审了一个五年的老博客站：留 78 篇、并 96 篇、删 84 篇、改 42 篇。客户最担心删了掉流量，实际两个月后流量比审之前还高 6%。","2026-10-09","\u002Fblog\u002Fblog-audit-300-posts",{"title":161,"description":162,"date":163,"category":31,"path":164},"弱网测试只开 3G 降速，会漏掉一半问题","我们把同一页放在三种弱网配置下测，首屏差了 2.3 秒。DevTools 的降速只模拟网络，不模拟丢包和真实手机的处境，这篇写我们的替代测法。","2026-10-08","\u002Fblog\u002Fweak-network-testing",{"title":166,"description":167,"date":168,"category":31,"path":169},"移动端技术 SEO：桌面和移动内容不一致的 4 个后果","移动优先索引意味着搜索引擎主要看移动版。桌面有内容、移动端没有，排名会按移动版算。2026 年 3 月到 6 月我们检查过 20 个软件站，4 个常见问题几乎每家都有。检查清单和修法都在这篇，照着查一遍一小时能看完。","2026-10-07","\u002Fblog\u002Fmobile-technical-seo",{"title":171,"description":172,"date":173,"category":8,"path":174},"排名掉了先别慌，先分清这四种情况","整站掉、单关键词掉、单页掉、显示掉但点击没掉，四种情况的处理完全不同。这篇写每种怎么确认、观察几天再动手，以及当天千万别做什么。","2026-10-06","\u002Fblog\u002Franking-drop-first-week",{"title":176,"description":177,"date":178,"category":8,"path":179},"国内备案和 Google 信任度，是两套不同的体系","备案是百度的入场券，Google 不看备案，看的是作者、来源、产品信息。软件公司出海时把国内那套资质搬过去，基本没用。两套信任体系怎么建，这篇写清楚。两边都做的公司，内容可以复用，信任建设必须分开做。","2026-10-05","\u002Fblog\u002Ficp-filing-google-trust",{"title":181,"description":182,"date":183,"category":8,"path":184},"替代页怎么写：搜竞品名字的人，其实是你的客户","“XX 替代”是 B2B 软件转化最高的一类词。我们给一个项目管理工具做了 12 个替代页，一年带来 1900 多次访问和 60 多条线索。","2026-10-04","\u002Fblog\u002Falternatives-page-seo",{"title":186,"description":187,"date":188,"category":37,"path":189},"领英公司页没人管，等于少了一个品牌入口","客户和候选人都会先搜领英。这篇写公司页要填的六项、员工页怎么联动、内容发什么，以及它为什么会影响品牌词的搜索结果页，附一个容易漏的设置。","2026-10-03","\u002Fblog\u002Flinkedin-company-page",{"title":191,"description":192,"date":193,"category":37,"path":194},"维基数据和知识面板，B2B 软件要不要管","我们给三个客户做过 Wikidata 条目，只有一个通过。这篇写知识面板的信息从哪来、什么公司适合做、条目为什么会被删，以及替代做法。","2026-10-02","\u002Fblog\u002Fwikidata-knowledge-panel",{"title":196,"description":197,"date":198,"category":8,"path":199},"别按竞品选题清单写内容，你复制不了他的推力","竞品发什么你发什么，是最省事也最没用的选题方法。2026 年 5 月一个客户跟风写了 12 篇同题，8 篇没进前 30；同期 6 篇场景文，4 篇进前 20。","2026-10-01","\u002Fblog\u002Fdont-copy-competitor-topics",{"title":201,"description":202,"date":203,"category":19,"path":204},"内容效果归因：注册到底算哪篇文章的功劳","一篇文章带来 10 个注册，另一篇带来 3 个，不代表第一篇更好。内容归因难在路径长、触点杂。我们用三层归因：直接来源、辅助触点、品牌搜索。算法、例子和两个误区都在这篇，两张表的做法也写了，每月出一次。","2026-09-30","\u002Fblog\u002Fcontent-attribution",{"title":206,"description":207,"date":208,"category":31,"path":209},"上线之后变慢，半小时内怎么定位到原因","发布后性能掉下去，先回滚还是先查。这篇写确认、对比、回滚的半小时流程，以及我们遇到最多的四种回归原因，和发版前的一次对比检查。","2026-09-29","\u002Fblog\u002Fdeploy-speed-regression",{"title":211,"description":212,"date":213,"category":31,"path":214},"站内搜索页被收录 8 万条：怎么清、怎么防","站内搜索结果页被搜索引擎大量收录，是软件站最常见的索引事故之一。我们处理过一个 8 万条站内搜索页被收录的站：清理用了 8 周，索引量降了 76%，核心页抓取翻倍。清理顺序和预防的三道闸都在这篇，照着做能少踩坑。","2026-09-28","\u002Fblog\u002Fsite-search-pages-indexed",{"title":216,"description":217,"date":218,"category":37,"path":219},"低价 SEO 套餐的成本账，便宜在哪省掉了什么","三千一个月的套餐，拆开看只有八小时人力和一批外链包。这篇算清低价套餐的成本结构、客户为什么买，以及六个月后重做的代价，附我们的三条建议。","2026-09-27","\u002Fblog\u002Fcheap-seo-package-cost",{"title":221,"description":222,"date":223,"category":19,"path":224},"一套内容改五个版本，是最蠢的多引擎策略","每个引擎单独写一版内容，是我们在预算充足的客户身上见过最浪费的做法。5 个版本的成本是 1 个版本的 3 倍，效果差异不到 10%。什么情况才该分开写，这篇讲清楚，也写了我们自己的一个教训。五版的预算集中到一版，效果更好。","2026-09-26","\u002Fblog\u002Fcontent-versioning-trap",{"title":226,"description":227,"date":228,"category":31,"path":229},"平台型产品的用户页面，别让搜索引擎全抓走","低代码平台有几十万个用户应用预览页，全被索引会拖垮整站质量。我们用三档规则处理，收录量降了 87%，有效页面排名反而升了。","2026-09-25","\u002Fblog\u002Fplatform-ugc-pages-indexing",{"title":231,"description":232,"date":233,"category":37,"path":234},"软件公司要不要做地图标注，看这一个条件","地图标注对纯线上 SaaS 基本没用，对有办公室、会接待客户的公司有用。这篇写判断条件、三家平台怎么填、常见的重复条目问题，以及怎么做效果判断。","2026-09-24","\u002Fblog\u002Fmap-listing-for-b2b",{"title":236,"description":237,"date":238,"category":31,"path":239},"AI 爬虫抓不到的四类内容，第一类最常见","我们抽查十二个站，被 AI 引用的内容九成在 HTML 正文里。这篇写登录墙、交互内容、PDF 和图片文字四类看不见的内容，以及怎么改回来。","2026-09-23","\u002Fblog\u002Fai-invisible-content-types",{"title":241,"description":242,"date":243,"category":19,"path":244},"内容页到注册，中间有三个位置能补","内容带来流量，接不住就白做。我们给内容页固定三个承接位置：读完后的一段话、正文里的功能链接、侧边栏的相关内容。改完试用率从 0.9% 到 2.1%。","2026-09-22","\u002Fblog\u002Fcontent-to-signup-path",{"title":246,"description":247,"date":248,"category":19,"path":249},"内容共创：和行业媒体、KOL 合作，写什么、怎么分工","内容共创不是找人转发。我们做过 12 次共创：联合报告、专家约稿、媒体专栏，效果差三倍，差别在选题和分工。三种形式怎么选、分工怎么定、效果怎么衡量都在这篇。12 次合作里我们只留下了 5 个长期伙伴，这 5 个才是重点。","2026-09-21","\u002Fblog\u002Fcontent-partnership",{"title":251,"description":252,"date":253,"category":31,"path":254},"给页面定性能预算，超过就拦住不让上线","没有预算的站，JS 只会一年比一年多。这篇写预算定哪四个数、怎么在 CI 里落地、超了怎么办，以及我们给软件官网的默认阈值，一个下午就能配上。","2026-09-20","\u002Fblog\u002Fperformance-budget",{"title":256,"description":257,"date":258,"category":31,"path":259},"nofollow 还有用吗：2026 年链接属性的正确用法","这三个链接属性被讨论了很多年，2026 年的实际作用已经很有限。我们的建议是：只在该用的地方用，别指望它们解决权重问题，也别把它们当惩罚工具。软件站最常误用的两处和正确用法都在这篇，用错地方反而伤抓取。","2026-09-19","\u002Fblog\u002Fnofollow-attributes-2026",{"title":261,"description":262,"date":263,"category":31,"path":264},"索引膨胀：收录 8 万页只有 3000 页有流量","收录多不等于好。我们给一个 8 万页索引的站做诊断，只有 3.8% 的页面有过展示。这篇讲索引膨胀怎么形成、怎么判断哪些该清，以及清理后流量为什么反而涨。","2026-09-18","\u002Fblog\u002Findex-bloat-cleanup",{"title":266,"description":267,"date":268,"category":37,"path":269},"承诺关键词排名的合同，我们一单都不签","过去两年推掉四个要求承诺排名的单子。这篇写为什么排不出承诺、承诺排名的公司靠什么做到，以及我们拿什么替代，客户为什么最后接受了。","2026-09-17","\u002Fblog\u002Fseo-no-ranking-guarantee",{"title":271,"description":272,"date":273,"category":37,"path":274},"夸克搜索的 AI 答案，正在抢走哪类查询","夸克的 AI 搜索答案正在抢走操作步骤类和概念解释类查询的点击，软件公司的文档站受影响最明显。我们监测了 20 个词、4 个月的数据，哪些查询还安全、软件公司该做什么，都写了，文末有一份可以直接用的应对清单。","2026-09-16","\u002Fblog\u002Fquark-ai-search-queries",{"title":276,"description":277,"date":278,"category":19,"path":279},"低代码平台的学习内容，是获客不是客服","教程视频放在帮助中心里，只有已付费的人看得到，等于把最好的获客素材锁起来了。我们把它搬到公开站，按角色分成三条学习路径，半年带来 2.8 万自然访问。","2026-09-15","\u002Fblog\u002Flow-code-academy-content",{"title":281,"description":282,"date":283,"category":37,"path":284},"官网信息和工商信息对不上，会出什么问题","一个站的公司名有三种写法，地址还是两年前的。这篇写哪些信息必须一致、不一致的四个后果、一次性统一的做法，附五页抽查清单。","2026-09-14","\u002Fblog\u002Fbusiness-info-consistency",{"title":286,"description":287,"date":288,"category":31,"path":289},"AI 认错你的公司，多半是三个来源不一致","我们抽查八个公开来源，一家公司的信息有三处对不上。这篇写 AI 怎么认实体、哪三项必须对齐、不一致怎么修，以及怎么验证修好了。","2026-09-13","\u002Fblog\u002Fentity-info-consistency-ai",{"title":291,"description":292,"date":293,"category":19,"path":294},"多语言内容：翻译、改写、本地化，三件事别混着做","把中文稿翻译成英文就发，是出海内容最常见的错误。翻译解决语言，改写解决搜索习惯，本地化解决用户场景，三层成本差三倍。什么时候用哪层、我们的流程和一个失败案例都在这篇，照着走能少花冤枉钱。","2026-09-12","\u002Fblog\u002Fmultilingual-content-workflow",{"title":296,"description":297,"date":298,"category":19,"path":299},"一个月二十篇和五篇，哪个更有用取决于一件事","我们统计过 4 个客户站一年的数据，篇数和自然流量几乎没有相关性，有效篇数才有。有效篇数指带独家信息、能被引用、能带来动作的那几篇。","2026-09-11","\u002Fblog\u002Fcontent-velocity-quality",{"title":301,"description":302,"date":303,"category":31,"path":304},"合成监控和真实用户监控，到底要不要都做","合成监控能发现回归，真实用户数据能反映体验。这篇写两者的偏差在哪、核心页面怎么选、采样率多少够用，以及自建 RUM 的最小方案。","2026-09-10","\u002Fblog\u002Frum-vs-synthetic-monitoring",{"title":306,"description":307,"date":308,"category":31,"path":309},"改域名后旧域名要挂多久，9 次迁移的恢复数据","旧域名 301 三个月就撤，是很多团队的做法，但我们在 9 次迁移里看到恢复期普遍在 8 周到 6 个月。这篇讲旧域名保留时长、成本，以及什么时候可以撤。","2026-09-09","\u002Fblog\u002Fold-domain-301-retention",{"title":311,"description":312,"date":313,"category":31,"path":314},"假蜘蛛比真蜘蛛多：我们怎么验证百度谷歌爬虫","日志里自称 Googlebot 和 Baiduspider 的请求，很多是假的：采集器、扫描器、竞品爬虫都会伪装。我们用 DNS 反查、IP 段核对和行为特征三步验证，一个站一天能筛出 3 万多条假蜘蛛请求。这套方法我们每周跑一次，基本能挡住假蜘蛛。","2026-09-08","\u002Fblog\u002Ffake-crawler-detection",{"title":316,"description":317,"date":318,"category":37,"path":319},"验收标准写不清，项目最后一定扯皮","排名、流量、转化写进验收，等于给项目埋雷。这篇写三类可验收的写法、哪些指标不能进合同，以及按月小验收的节奏怎么排，谈合同时照着抄就能用。","2026-09-07","\u002Fblog\u002Fseo-acceptance-criteria",{"title":321,"description":322,"date":323,"category":37,"path":324},"搜狗、360、神马三家的收录差异，我们对比了 5 个站","三家国内引擎的收录差异比想象中大：360 收录最快但卡备案，搜狗最慢但流量质量稍好，神马只收移动端且量最小。5 个站、3 个月的对比数据和处理建议都在这篇，还有一个我们没预料到的发现。读完能决定要不要投这三家。","2026-09-06","\u002Fblog\u002Fsogou-360-shenma-index-compare",{"title":326,"description":327,"date":328,"category":8,"path":329},"无代码应用模板页，别只放一张预览图","用户搜“库存管理 模板”是想拿过来就能用，一张预览图回答不了这个问题。我们把模板页加上字段结构、权限设置和改造步骤，安装转化从 4% 到 11%。","2026-09-05","\u002Fblog\u002Fno-code-app-template-seo",{"title":331,"description":332,"date":333,"category":37,"path":334},"搜索结果里有负面，先分清三种情况再动手","负面信息处理错方式，往往越处理越糟。这篇写事实错误、真实投诉、垃圾负面三类怎么区分，各自该怎么处理，以及为什么不建议买删帖。","2026-09-04","\u002Fblog\u002Fnegative-serp-handling",{"title":336,"description":337,"date":338,"category":8,"path":339},"内容保鲜：AI 更愿意引用最近更新的页面吗","我们更新了二十篇旧文，六篇在四周内被重新引用。但保鲜不等于改日期，这篇写哪类内容值得更新、怎么改才有效、以及无效更新的两种做法。","2026-09-03","\u002Fblog\u002Fcontent-freshness-ai-citation",{"title":341,"description":342,"date":343,"category":19,"path":344},"更新日志和“最后更新”日期，用对了能救回一批老文","老文更新后没有更新标记，等于白改。我们在标题下加一行“某年某月更新”，并维护一份更新日志，六周后这批页面的点击率涨了 0.6 个百分点。","2026-09-02","\u002Fblog\u002Fupdate-log-last-updated",{"title":346,"description":347,"date":348,"category":19,"path":349},"内容维护日历：哪些文章每年必更，哪些三年不用动","内容发完就不管，两年后一半的文章在拖后腿。我们按内容类型排了维护日历：产品文档每季度、教程每年、观点文两年、数据文跟着数据走。哪些必更、哪些别碰、日历怎么排都在这篇，照着排能省一半维护时间。","2026-09-01","\u002Fblog\u002Fcontent-refresh-calendar",{"title":351,"description":352,"date":353,"category":25,"path":354},"测速工具那么多，我们最后只留三个","实验室工具、合成监控、真实用户数据，三类各有用处。这篇写 Lighthouse、PageSpeed Insights、WebPageTest 和国内测速平台各管什么，哪些可以放弃。","2026-08-31","\u002Fblog\u002Fcwv-monitoring-tools",{"title":356,"description":357,"date":358,"category":31,"path":359},"多语言站用子目录、子域名还是独立域名：软件公司怎么选","三种结构没有绝对优劣，看你的市场、团队和 SEO 目标。2025 年 9 月到 2026 年 6 月，我们给 12 个软件公司做过选择：做周边市场选子目录，做重点市场选子域名，品牌独立运营才用独立域名。判断依据和迁移的坑都在这篇。选错再改的成本很高。","2026-08-30","\u002Fblog\u002Fmultilingual-site-structure",{"title":361,"description":362,"date":363,"category":31,"path":364},"扁平结构和层级结构，软件站该选哪个","扁平结构不是所有页面都放首页，层级结构也不等于深目录。这篇用两个真实站点对比，给出按内容类型分层的做法：什么该 2 层，什么该 4 层，以及判断标准。","2026-08-29","\u002Fblog\u002Fflat-vs-hierarchical-structure",{"title":366,"description":367,"date":368,"category":37,"path":369},"SEO 报价按词还是按人天，我们为什么选人天","按词报价客户会挑词，按排名承诺我们不敢接。这篇写人天报价的结构、每月工时表怎么给客户看，以及为什么报价方式决定了合作方式。","2026-08-28","\u002Fblog\u002Fseo-pricing-per-day",{"title":371,"description":372,"date":373,"category":8,"path":374},"被 Google 忽略的长尾词，在 Bing 上反而有量","有些长尾词在 Google 上竞争太激烈或搜索量太小，在 Bing 上竞争低、转化好。我们给 3 个客户做了 Bing 长尾，单个注册成本是 Google 的三分之一，选词标准、具体做法和一个我们放弃的词，都在这篇。Bing 长尾不是主战场，但性价比高。","2026-08-27","\u002Fblog\u002Fbing-long-tail-keywords",{"title":376,"description":377,"date":378,"category":19,"path":379},"低代码和无代码，搜索词根本是两拨人","同一个产品，客户搜“低代码”和搜“无代码”的心态完全不同。我们按两套词分开做内容，一个客户半年线索涨了 80%，成本没变。","2026-08-26","\u002Fblog\u002Flow-code-vs-no-code-keywords",{"title":381,"description":382,"date":383,"category":37,"path":384},"仿冒官网和山寨下载站，怎么发现怎么处理","一个客户搜品牌加下载，前五条里三条是山寨站。这篇写三种仿冒类型、取证要留什么材料、五个处理渠道，以及官网自己怎么防，附每月十分钟的排查法。","2026-08-25","\u002Fblog\u002Fcounterfeit-site-protection",{"title":386,"description":387,"date":388,"category":19,"path":389},"表格比段落更容易被 AI 引用，我们对比了两种写法","同一组数据，表格版四十五天被引用九次，段落版三次。这篇写什么该做成表、表怎么写才被摘得出来、图片表格为什么没用，以及怎么验证。","2026-08-24","\u002Fblog\u002Ftables-ai-citation",{"title":391,"description":392,"date":393,"category":19,"path":394},"长尾词聚合：一个页面覆盖 20 个词，怎么写才不是缝合怪","长尾词一个个写文章成本太高，聚合到一个页面是常见做法。但聚合不等于把词堆进去。我们判断“该不该聚合”看三个条件，写法上分四层。一个页面覆盖 20 个词、排名进前 10 的案例也在这篇，照着结构写就行。","2026-08-23","\u002Fblog\u002Flongtail-aggregation",{"title":396,"description":397,"date":398,"category":19,"path":399},"数据文的结构：先给反常识结论，再交代口径","数据文不是把数字堆出来。我们固定四段结构：反常识结论、数据来源和口径、两三个细分发现、适用边界。顺序错了，读者读三段就关掉。","2026-08-22","\u002Fblog\u002Fdata-post-structure",{"title":401,"description":402,"date":403,"category":31,"path":404},"移动端比桌面慢三秒，差距不只在网络","同一页面桌面 1.8 秒、移动 5.1 秒。很多人以为差在网络，其实 CPU 降速和布局复杂度占大头。这篇写四个差距来源和移动端专属的测试方法。","2026-08-21","\u002Fblog\u002Fmobile-desktop-cwv-gap",{"title":406,"description":407,"date":408,"category":31,"path":409},"CDN 缓存配错，改版后蜘蛛三天没看到新内容","一次改版上线三天，搜索引擎抓到的还是旧页面，原因是 CDN 缓存把 HTML 缓存了 7 天。缓存和抓取的关系、三个配置原则、怎么验证都在这篇，附我们踩过的两次坑。这层不出问题则已，出问题就是全站级别的，别等出事再查。","2026-08-20","\u002Fblog\u002Fcdn-cache-crawl",{"title":411,"description":412,"date":413,"category":31,"path":414},"带参数的页面 canonical 怎么写，三种场景","参数页的 canonical 是最容易写错的一类。筛选、排序、追踪三种参数，写法完全不同。这篇给每种的正确配置、一个 2026 年 6 月的案例，以及写错之后怎么排查。","2026-08-19","\u002Fblog\u002Fcanonical-on-parameter-pages",{"title":416,"description":417,"date":418,"category":37,"path":419},"项目延期不可怕，可怕的是最后一天才说","延期三周，提前十天告知，客户反而更信任我们。这篇写发现延期当天要做的四件事、跟客户说的四句话模板，以及三种不该出现的解释。","2026-08-18","\u002Fblog\u002Fproject-delay-communication",{"title":421,"description":422,"date":423,"category":31,"path":424},"百度和 Google 对 JS 渲染的支持，差得比想象中大","Google 能渲染大部分 JS，百度只渲染一部分，而且不稳定。今年 3 月我们测了 4 种前端架构在两边抓取里的表现，客户端渲染的站在百度上基本等于裸奔。判断方法、改法都写了，看完能判断自己的站有没有问题。","2026-08-17","\u002Fblog\u002Fjs-rendering-google-baidu",{"title":426,"description":427,"date":428,"category":8,"path":429},"迁移指南是云服务最接近成交的内容","搜“从 AWS 迁移到 XX”的人已经在做决策，内容只要回答得够具体，线索质量就很高。我们写了 8 篇迁移指南，一年带来 470 条企业线索，成单 11 个。","2026-08-16","\u002Fblog\u002Fcloud-migration-guide-seo",{"title":431,"description":432,"date":433,"category":19,"path":434},"白皮书要不要留资才能下载，我们试过两种","留资能拿到线索，但会挡掉九成读者和引用。这篇写两种模式的取舍、留资表单怎么精简、不留资时怎么衡量白皮书的价值，附我们的折中方案。","2026-08-15","\u002Fblog\u002Fgated-whitepaper-download",{"title":436,"description":437,"date":438,"category":8,"path":439},"给 AI 准备一张产品事实页，比写十篇软文有用","一页列清产品名、定位、功能边界、价格、集成和限制，上线六周后被三个引擎引用。这篇写事实页放什么、怎么写、放哪里、怎么维护，附一份字段清单。","2026-08-14","\u002Fblog\u002Fai-fact-sheet-page",{"title":441,"description":442,"date":443,"category":19,"path":444},"竞品内容监控：每周 30 分钟，我们看这四件事","竞品内容监控不是每天刷对方博客。我们每周花 30 分钟看四件事：新发内容、排名变化、外链来源、SERP 占位。看什么、怎么记、哪些不用跟，都在这篇，附我们的监控表结构，省得自己设计，照着填就行，一周 30 分钟够用。","2026-08-13","\u002Fblog\u002Fcompetitor-content-monitoring",{"title":446,"description":447,"date":448,"category":8,"path":449},"作者页别只放头像，放三样能核对的东西","作者页的作用是让人验证“这个人真做过这件事”。我们固定放三样：做过什么项目、写过哪些文章、在哪能查到他。三样都能点开核对，比任何头衔有用。","2026-08-12","\u002Fblog\u002Fauthor-page-credentials",{"title":451,"description":452,"date":453,"category":31,"path":454},"首页背景视频值不值得那两秒，我们算过账","一个 12 兆的背景视频让移动端首屏到了 6 秒。这篇算清视频的字节、解码和内存成本，写什么情况值得放、怎么延迟加载、以及改成点击播放后的效果。","2026-08-11","\u002Fblog\u002Fhomepage-video-weight",{"title":456,"description":457,"date":458,"category":31,"path":459},"站点地图的更新日期乱写，等于告诉蜘蛛别来","每次生成 sitemap 都把 lastmod 写成当前时间，是 CMS 的常见默认行为。搜索引擎发现这个字段不可信就会忽略它。这篇讲正确用法、增量拆分，以及一个 2026 年 5 月的对比测试。","2026-08-10","\u002Fblog\u002Fsitemap-lastmod-correct-usage",{"title":461,"description":462,"date":463,"category":31,"path":464},"服务器慢 1 秒，抓取频次掉多少：我们测了 4 个站","服务器响应时间和抓取频次的关系，比大多数人想的直接。我们给 4 个站做了前后对比：TTFB 从 1.8 秒降到 0.4 秒后，Googlebot 抓取量平均涨 42%，百度蜘蛛涨 31%。测试方法和三个隐蔽的拖慢原因都在这篇，看完能判断自己的站要不要优化。","2026-08-09","\u002Fblog\u002Fserver-response-crawl",{"title":466,"description":467,"date":468,"category":37,"path":469},"跨部门卡住的 SEO 项目，问题多在开发排期","一个 title 模板改动等了三周，一次 robots 误改让全站被屏蔽六小时。这篇写 SEO 需求怎么进开发排期、工单怎么写、找谁推进才有效。","2026-08-08","\u002Fblog\u002Fdev-queue-blocking-seo",{"title":471,"description":472,"date":473,"category":19,"path":474},"Helpful Content 并入核心后，内容站还怎么写","Helpful Content 在 2024 年 3 月并入核心更新后，“为搜索而写”这套彻底失效。内容站要活下来只有两条路：有产品支撑，或者有独家信息。我们选了第三条：不做纯内容站。这个判断是怎么来的、反例长什么样，都写在这篇里，代价不小。","2026-08-07","\u002Fblog\u002Fhelpful-content-core-integration",{"title":476,"description":477,"date":478,"category":31,"path":479},"云服务状态页，也是一块能排名的内容","故障通告页看起来和 SEO 无关，但“XX 云 故障”这类词有真实搜索量，搜不到官方口径就会传谣言。我们把状态页改成可索引结构，故障期间自然访问涨了 5 倍。","2026-08-06","\u002Fblog\u002Fcloud-status-page-seo",{"title":481,"description":482,"date":483,"category":19,"path":484},"客户案例里的数字，哪些能写哪些不能写","案例页最常见的错误是把客户的业绩当自己的功劳。这篇写三类数字的写法、客户授权的流程、没有数据时怎么写案例，以及我们的案例生产流程。","2026-08-05","\u002Fblog\u002Fcase-study-numbers-authorization",{"title":486,"description":487,"date":488,"category":19,"path":489},"GEO 内容的审稿流程，我们改成了三道关","以前一篇稿子写完就发，现在要过事实卡、可引用段落、上线后引用验证三道。返工多了一倍，但同样的话题，被 AI 引用的比例明显不同，流程可直接照搬。","2026-08-04","\u002Fblog\u002Fgeo-fact-check-workflow",{"title":491,"description":492,"date":493,"category":8,"path":494},"排名涨了流量没涨，先查曝光和点击率两件事","排名是从曝光加权算出来的，曝光涨排名会好看，流量不一定涨。2026 年 2 月我们把报告里的排名换成曝光和点击率，客户问题少了一半。","2026-08-03","\u002Fblog\u002Ftraffic-vs-ranking",{"title":496,"description":497,"date":498,"category":19,"path":499},"内容预算怎么分：写手、工具、外包、试验的比例","内容预算不是“有多少钱写多少稿”。我们按 5:2:2:1 分：写手和编辑一半、工具和数据两成、外部合作两成、试验一成。每块怎么花、什么时候该调、一个失败的分配案例都在这篇，预算有限的团队可以照着调，比例不是死的，按阶段调。","2026-08-02","\u002Fblog\u002Fcontent-budget-split",{"title":501,"description":502,"date":503,"category":31,"path":504},"一张图三个尺寸：srcset 配对了能省七成流量","一个站图片总重 3.8 兆，配上 srcset 和图片 CDN 之后降到 1.1 兆。这篇写 sizes 怎么写、格式怎么选，以及高分辨率屏和懒加载的坑。","2026-08-01","\u002Fblog\u002Fresponsive-images-srcset",{"title":506,"description":507,"date":508,"category":31,"path":509},"索引膨胀：站内 60% 页面不该被收录，怎么控制","索引膨胀的意思是搜索引擎收录了大量没有搜索价值的页面，占着抓取预算还拉低整站评价。我们给一个 8 万页的站做清理，压到 1.2 万页，核心页抓取频次翻倍。判断标准和三种控制手段都在这篇，照着做能少走弯路。","2026-07-31","\u002Fblog\u002Findex-bloat-control",{"title":511,"description":512,"date":513,"category":31,"path":514},"CDN 和防火墙把蜘蛛拦了，日志里看不出来的坑","蜘蛛被 WAF 拦掉时，日志里可能只剩空白或者 403，甚至完全没有记录。我们遇到 5 次这类事故，最久的一次被拦了 6 周。这篇讲怎么发现、怎么配白名单、怎么进变更流程。","2026-07-30","\u002Fblog\u002Fcdn-waf-blocking-crawlers",{"title":516,"description":517,"date":518,"category":37,"path":519},"SEO 项目的交付物清单，缺一项就会扯皮","交付物不是只有报告。这篇列出我们每个项目固定交付的九项内容，写清每项的格式、频率和验收方式，避免月底为做没做争论，附九项清单和验收写法。","2026-07-29","\u002Fblog\u002Fseo-deliverables-list",{"title":521,"description":522,"date":523,"category":31,"path":524},"百度抓取诊断报 403 和 503，我们怎么查源站","抓取诊断报 403 或 503，九成不是百度的问题，是源站的防火墙、CDN 或 UA 拦截把蜘蛛挡了。我们处理过 20 多次这类问题，我们处理过 20 多次这类问题，验证方法、三种常见原因和修复动作都在这篇，也写了一个反复出现的坑。","2026-07-28","\u002Fblog\u002Fbaidu-crawl-diagnosis-errors",{"title":526,"description":527,"date":528,"category":19,"path":529},"云厂商的教程页，是最大的自然流量入口","“怎么在 XX 上部署 YY”这类教程，搜索量比产品词大得多。我们按“任务 + 技术栈”建了 60 个教程页，一年带来 11 万自然访问。","2026-07-27","\u002Fblog\u002Fcloud-tutorial-content-seo",{"title":531,"description":532,"date":533,"category":37,"path":534},"开源项目的 issue 响应速度，是品牌的信任信号","开发者会翻你的 issue 列表判断项目活不活。这篇写响应速度怎么影响选型、模板和标签怎么设、issue 里的问题怎么变成内容。","2026-07-26","\u002Fblog\u002Fgithub-issue-response-trust",{"title":536,"description":537,"date":538,"category":37,"path":539},"AI 把你的产品说错了，纠正比投诉有用","我们抽查四个引擎的六十个问题，十二个答案里有事实错误。这篇写四类常见错误、错从哪来、四条纠正路径，以及为什么投诉平台通常没用。","2026-07-25","\u002Fblog\u002Fai-product-facts-wrong",{"title":541,"description":542,"date":543,"category":19,"path":544},"新产品上线前 8 周的内容排期：预热、教育、发布、跟进","软件产品发布当周才开始写内容，已经晚了。我们把内容排期提前 8 周：预热期讲问题、教育期讲方案、发布期讲产品、跟进期讲案例。四个阶段各写什么、什么时间发，都在这篇，附一个发布首月的数据，都是我们自己跟的案例。","2026-07-24","\u002Fblog\u002Fproduct-launch-content-plan",{"title":546,"description":547,"date":548,"category":19,"path":549},"多语言先做哪个市场，我们用四个分数排顺序","不是所有市场都值得做多语言。我们按询盘来源、搜索需求、竞品覆盖、支持能力四项打 10 分制，先做分数最高的一个市场，做完再考虑第二个。","2026-07-23","\u002Fblog\u002Fmultilingual-market-priority",{"title":551,"description":552,"date":553,"category":31,"path":554},"preload 用错反而更慢，先看这三个判断","一个站 preload 了九个资源，只有三个真用上，首屏反而慢了两百毫秒。这篇写 preload、preconnect、dns-prefetch 各管什么，什么时候不该用。","2026-07-22","\u002Fblog\u002Fpreload-preconnect-misuse",{"title":556,"description":557,"date":558,"category":31,"path":559},"SoftwareApplication 结构化数据：9 个站加完后的效果和坑","软件公司该加的 JSON-LD，不是 Article 和 FAQPage，而是 SoftwareApplication。2025 年 12 月到 2026 年 5 月我们给 9 个站加完，6 个出现富结果，3 个没有任何变化。加什么字段、三种常见错误、什么情况别加，都在这篇。加完能看到效果的前提是页面信息本身完整，这点很多人忽略。","2026-07-21","\u002Fblog\u002Fsoftware-app-schema",{"title":561,"description":562,"date":563,"category":31,"path":564},"Organization 结构化数据：六个字段写对才有用","Organization 标记容易被当成模板活，填错却会拖累全站。logo、sameAs、url、name、address、contactPoint 这六个字段我们逐个过，附一个 2026 年 4 月的排查记录。","2026-07-20","\u002Fblog\u002Forganization-schema-fields",{"title":566,"description":567,"date":568,"category":37,"path":569},"SEO 项目排期，前三个月别排满内容","排期表上第一个月就写满 20 篇，执行到第三周一定卡住。这篇写我们按阶段排期的做法：技术、内容、外链各占多少，以及验收节点怎么设，项目延期多半是排期结构的问题。","2026-07-19","\u002Fblog\u002Fseo-project-scheduling",{"title":571,"description":572,"date":573,"category":37,"path":574},"百度新站沙盒期到底多久：我们跟了 6 个站","百度没有官方承认的沙盒期，但我们跟的 6 个新站里，5 个在前 3 个月有低展现期，平均 11 周后才稳定出词。哪些事能缩短、哪些纯属白做，这篇按 6 个站的记录写清楚，附一个我们没料到的发现。6 个站的记录都在正文里。","2026-07-18","\u002Fblog\u002Fnew-site-sandbox-baidu",{"title":576,"description":577,"date":578,"category":31,"path":579},"云服务的区域页，一个 region 一页才吃得到流量","用户会搜“XX 云 新加坡 机房”“XX 云 法兰克福 延迟”。我们把 9 个区域拆成独立页，半年带来 1.4 万自然访问，企业咨询多了三成。","2026-07-17","\u002Fblog\u002Fcloud-region-pages-seo",{"title":581,"description":582,"date":583,"category":19,"path":584},"技术社区发文章，别把官网内容复制一遍","同一篇文章搬到掘金、CSDN、开源中国，搜索权重会被分散。这篇写哪些内容适合分发、怎么改开头和结尾、原文链接怎么放才有用。","2026-07-16","\u002Fblog\u002Fdev-community-reposting",{"title":586,"description":587,"date":588,"category":19,"path":589},"内容发布后的 90 天：我们固定做的 4 件事","一篇文章发出去只是开始。我们给每篇重点内容排了 90 天运营动作：第 1 周查收录、第 2 到 4 周补内链和分发、第 5 到 8 周看数据、第 9 到 12 周决定加料还是放弃。每步的具体动作都在这篇，照着排能少走弯路，重点是第 1 周和第 12 周。","2026-07-15","\u002Fblog\u002Fcontent-lifecycle-90-days",{"title":591,"description":592,"date":593,"category":19,"path":594},"内容审计表格该有九个字段，少一个结论就会错","审计表不是把文章列出来。我们固定九个字段：URL、曝光、点击、排名、外链、转化、主题、意图、建议动作。少一个字段，判断就会偏，最后做出来的动作也是错的。","2026-07-14","\u002Fblog\u002Fcontent-audit-sheet-fields",{"title":596,"description":597,"date":598,"category":31,"path":599},"把关键 CSS 内联进 HTML，首屏能省半秒","CSS 阻塞渲染，把首屏样式内联进 HTML 能提前首屏时间。这篇写关键 CSS 怎么抽、四 KB 上限怎么守、剩余样式怎么异步加载，以及三个常见坑。","2026-07-13","\u002Fblog\u002Fcritical-css-inline",{"title":601,"description":602,"date":603,"category":31,"path":604},"1200 张产品图只被收录 40 张，问题在三个地方","产品图不收录，不是图片不够好看。我们查了 6 个软件站，问题集中在懒加载写法、图片 URL 参数和页面结构三处。1200 张图修完收录到 780 张，图片搜索带来的访问涨了 3 倍。三个修法都不难，难在有人愿意去查。文末有修完的数据。","2026-07-12","\u002Fblog\u002Fimage-seo-technical",{"title":606,"description":607,"date":608,"category":31,"path":609},"抓取频次上不去，先看服务器那 200 毫秒","很多人以为抓取少是\"权重不够\"，其实服务器响应时间的影响更直接。我们把一个客户的 TTFB 从 1.1 秒压到 180 毫秒，Googlebot 抓取量一个月涨了 2.7 倍。这篇讲测量、优化和验证。","2026-07-11","\u002Fblog\u002Fcrawl-frequency-server-response",{"title":611,"description":612,"date":613,"category":19,"path":614},"外包写手的稿子，我们通常改掉三成","外包稿的常见问题不是文笔，是行业细节和判断力。这篇写我们的约稿单怎么填、改稿改哪三处、什么样的稿子直接退，以及怎么和写手长期合作。","2026-07-10","\u002Fblog\u002Ffreelance-writer-editing",{"title":616,"description":617,"date":618,"category":37,"path":619},"2026 年上半年三次算法波动，我们记录的共同点","2026 年上半年我们记录到三次明显波动：1 月、3 月、5 月。共同点是内容站先掉、产品页更稳、恢复周期 4 到 6 周。每次波动的时间、幅度、掉了哪些页面和我们的应对都记下来了。三次都不在官方更新列表里，但影响够大。","2026-07-09","\u002Fblog\u002Falgorithm-volatility-first-half-2026",{"title":621,"description":622,"date":623,"category":19,"path":624},"插件权限写得清不清楚，直接影响装不装","用户装插件前会看权限说明，写得含糊就流失。我们把一个插件的权限页从 200 字扩到 900 字，商店页安装转化从 3.2% 到 5.6%。","2026-07-08","\u002Fblog\u002Fextension-permissions-trust",{"title":626,"description":627,"date":628,"category":19,"path":629},"软件公司在知乎回答问题，别写成产品介绍","我们跟踪过四十个知乎回答的引流效果，写得像产品介绍的基本零转化。这篇写选题怎么挑、回答怎么组织、哪些做法会被反感，附一个我们自己的开头写法。","2026-07-07","\u002Fblog\u002Fzhihu-answer-strategy",{"title":631,"description":632,"date":633,"category":19,"path":634},"掘金发了 12 篇技术文，带来的点击和询盘是多少","2025 年 9 月到 2026 年 6 月，我们帮一个客户在掘金发了 12 篇文章。总阅读 6.8 万，官网引荐 1200 次，询盘 4 个。这篇复盘哪几篇有效、为什么，以及做掘金内容的三个前提。","2026-07-06","\u002Fblog\u002Fjuejin-content-revenue-review",{"title":636,"description":637,"date":638,"category":19,"path":639},"专题页怎么组织：一个入口带二十篇，别做成第二支柱页","支柱页按知识组织，专题页按任务组织。混着做过一次，专题页变成第二支柱页，两个都不排。这篇写专题页该有什么、跟支柱页的区别，以及什么情况下该做。","2026-07-05","\u002Fblog\u002Ftopic-hub-page-structure",{"title":641,"description":642,"date":643,"category":19,"path":644},"同一个话题别人写过了，第二篇还有机会吗","关键词被大站占满，不代表不能写。我们判断“第二篇有没有机会”看三个信号：SERP 里有没有旧内容、有没有没被覆盖的子问题、我们有没有独家素材。三种可以写、三种别写的情况都在这篇，看完能判断一个词值不值得做第二篇。","2026-07-04","\u002Fblog\u002Fsecond-article-chance",{"title":646,"description":647,"date":648,"category":31,"path":649},"第三方脚本治理，先砍掉不产生价值的那几个","一个站 23 个第三方脚本，9 个从没带来过转化。这篇写怎么清点、分三类处理、加载时机怎么调，以及上线前的准入规则，附一份季度审计清单。","2026-07-03","\u002Fblog\u002Fthird-party-script-audit",{"title":651,"description":652,"date":653,"category":31,"path":654},"500 页的站，每个页面都能 3 次点击到达吗","我们抽查过 30 个软件站，平均有 34% 的页面从首页点 4 次以上才能到。内链深度不是玄学，它直接决定抓取频次和排名。抽查方法、三种补链方式和一次改版后的数据都在这篇。先抽查再加链接，两周能看到变化。","2026-07-02","\u002Fblog\u002Finternal-link-depth",{"title":656,"description":657,"date":658,"category":31,"path":659},"必应站长工具的收录报告，被低估的第二个数据源","做外贸和工具类软件的站，必应的流量常被忽略。必应站长工具的索引浏览、URL 提交、关键词研究三个模块我们都会用，这篇讲它比 GSC 多给了什么、少给了什么。","2026-07-01","\u002Fblog\u002Fbing-webmaster-tools-index",{"title":661,"description":662,"date":663,"category":19,"path":664},"内容排期怎么排，一周三篇的节奏能撑多久","一周三篇听着不多，按我们的实际产能只能撑六周。这篇写排期要考虑的四项隐性成本、我们怎么排季度表，以及为什么别按篇数考核写手，附我们的季度排法。","2026-06-30","\u002Fblog\u002Fcontent-calendar-cadence",{"title":666,"description":667,"date":668,"category":8,"path":669},"本地包对软件公司基本没用，别在这上面花预算","本地包（Local Pack）对软件公司几乎没有价值，只有做本地服务、本地部署的例外。我们监测 14 个软件站，本地包带来的访问占比中位数 0.3%，投入产出比极低，只有三种业务值得做。我们的建议是：把这笔预算挪到产品内容上。","2026-06-29","\u002Fblog\u002Flocal-pack-software-companies",{"title":671,"description":672,"date":673,"category":31,"path":674},"桌面客户端下载页，被忽略的搜索入口","用户会搜“XX 下载 Windows 版”“XX 安装失败”。我们给一个桌面工具重做下载区，下载页自然访问涨 3.1 倍，安装失败相关咨询降了一半。","2026-06-28","\u002Fblog\u002Fdesktop-app-download-page-seo",{"title":676,"description":677,"date":678,"category":37,"path":679},"客户评价放哪里才有用：官网、评测站还是社群","官网评价可控但没人信，第三方评测站可信但不可控。这篇写三种载体各怎么用、官网评价怎么写才有效、结构化数据的合规边界，附一个四句话模板。","2026-06-27","\u002Fblog\u002Ftestimonials-placement",{"title":681,"description":682,"date":683,"category":8,"path":684},"开发者工具公司：把 GitHub 组织页做成第二官网","我们的客户里做开发者工具的，GitHub 组织页带来的注册有时超过官网博客。这篇讲 README 的四块结构、Issue 区怎么当内容库、以及怎么让 AI 搜索引用你的 README。","2026-06-26","\u002Fblog\u002Fgithub-org-second-homepage",{"title":686,"description":687,"date":688,"category":19,"path":689},"选题的 5 个否决条件：我们为什么不追热点","选题会最大的浪费不是选错题，是花了时间讨论一个根本不该做的题。我们定了 5 个否决条件：没有搜索需求、没有转化路径、写不出独家、竞品已经写透、维护成本过高。每条都有判断方法，选题会上照着过一遍就行。","2026-06-25","\u002Fblog\u002Ftopic-rejection-criteria",{"title":691,"description":692,"date":693,"category":8,"path":694},"不买关键词工具能不能挖词，能，但慢三倍","预算为零也能挖词。四个免费来源：搜索后台、站内搜索、客服工单、竞品目录。我们给一个零预算客户做了三周，挖出 210 个词，进排期的 31 个。","2026-06-24","\u002Fblog\u002Fkeyword-research-without-tools",{"title":696,"description":697,"date":698,"category":31,"path":699},"JS 包体拆完还是慢，问题在加载顺序","一个站 JS 从 2.1 兆拆到 600K，首屏还是 2.8 秒。拆包解决下载，解决不了执行和依赖顺序。这篇写拆错的三种情况、关键 chunk 怎么排、真正的大头在哪。","2026-06-23","\u002Fblog\u002Fjs-bundle-load-order",{"title":701,"description":702,"date":703,"category":31,"path":704},"筛选参数把抓取预算吃光了：电商站的 5 个处理规则","一个 B2B 电商站的筛选组合生成了 6 万多个 URL，抓取预算的 62% 花在参数页上，核心产品页一周才被抓一次。我们用 5 条规则把参数页从 6 万压到 800，收录率和询盘都回来了。照着做能把参数页数量压掉 90%，还能把收录率拉回来。","2026-06-22","\u002Fblog\u002Ffaceted-navigation-crawl",{"title":706,"description":707,"date":708,"category":31,"path":709},"换内容管理系统不掉排名，URL 和模板先对齐","从 WordPress 换到自研 CMS、从自研换到 Headless，我们做过 4 次。掉排名基本不是 CMS 本身的问题，而是迁移时没对齐 URL、模板和渲染。附一份迁移前后对照检查表。","2026-06-21","\u002Fblog\u002Fcms-migration-url-template",{"title":711,"description":712,"date":713,"category":19,"path":714},"选题会开成聊天会，问题出在没带数据","三十分钟出八个能落地的选题，前提是会前准备三份数据。这篇写选题会只回答哪三个问题、哪些讨论纯属浪费、会后选题卡怎么填，开完就能用。","2026-06-20","\u002Fblog\u002Ftopic-selection-meeting",{"title":716,"description":717,"date":718,"category":8,"path":719},"页面体验分对排名的影响，没工具厂商说的那么大","页面体验被工具厂商说成了排名关键。我们优化了 11 个站的核心网页指标，流量提升中位数 8%，没有一个靠这个从 20 名进前 10。什么时候值得做、优先级怎么排，这篇讲清楚，也写了一个被误诊的案例，结论可能和你想的不一样。","2026-06-19","\u002Fblog\u002Fpage-experience-signal",{"title":721,"description":722,"date":723,"category":8,"path":724},"插件官网该做什么页面，商店页替代不了","插件商店页只有一屏，讲不清权限、定价和场景，用户装完还有一堆疑问。我们给 5 个插件站补了场景、权限、定价、帮助四类页面，官网自然访问平均涨 2 倍。","2026-06-18","\u002Fblog\u002Fbrowser-extension-site-seo",{"title":726,"description":727,"date":728,"category":19,"path":729},"公众号文章能帮官网做搜索吗，答案和你想的不一样","公众号内容在微信里搜得到，但基本不参与百度排名，外链也是屏蔽的。这篇讲公众号对搜索真正有用的三种方式、哪些活不值得花时间，附我们建议的内容分发顺序。","2026-06-17","\u002Fblog\u002Fwechat-article-seo-value",{"title":731,"description":732,"date":733,"category":37,"path":734},"招聘页是很多软件公司漏掉的信任信号","招聘页常年是官网流量前五的页面，但大部分公司只挂了几个职位。这篇写招聘页为什么客户也会看、该写什么、JobPosting 标记怎么处理。","2026-06-16","\u002Fblog\u002Fhiring-page-trust-signal",{"title":736,"description":737,"date":738,"category":19,"path":739},"教程文缺三样东西：前置条件、验证方法、失败信号","教程写不好，八成不是文笔问题。我们重写过一篇 SDK 接入文档，补齐前置条件、验证命令和报错解释之后，相关工单从每月 34 条降到 18 条。","2026-06-15","\u002Fblog\u002Ftutorial-post-structure",{"title":741,"description":742,"date":743,"category":19,"path":744},"一篇长文拆成 6 个渠道：我们的内容复用流程和偷懒清单","写完一篇 4000 字长文只发一个渠道，是最常见的浪费。我们把它拆成 6 个渠道：公众号、Newsletter、知乎、视频脚本、社交媒体卡片、销售物料。拆什么、改什么、哪些不能复用都在这篇，拆分清单可以直接抄，我们每周都在用。","2026-06-14","\u002Fblog\u002Fcontent-repurposing",{"title":746,"description":747,"date":748,"category":31,"path":749},"中文字体加载慢，问题不在字体文件大小","一个中文字体包动辄八兆，但真正拖慢首屏的是加载时机和字形切换。这篇写子集化怎么做、font-display 三种取值怎么选、什么时候干脆别用自定义字体。","2026-06-13","\u002Fblog\u002Ffont-loading-chinese",{"title":751,"description":752,"date":753,"category":31,"path":754},"百度对 JS 渲染支持到什么程度，我们测了 6 个页面","百度能执行 JS，但能力有限。我们设计 6 个测试页面，分别测 React 渲染、接口异步数据、动态内链、懒加载、Web Component、JSON-LD 注入，记录百度的实际抓取结果。","2026-06-12","\u002Fblog\u002Fbaidu-js-rendering-test",{"title":756,"description":757,"date":758,"category":31,"path":759},"百度索引量掉了 40%，我们从哪查起","百度索引量一周掉 40%，先别改内容。我们按“先看抓取、再看站点、最后看内容”的顺序查，6 次里有 4 次问题出在技术层。排查步骤、每一步看什么数据、修复顺序都在这篇。6 次里有 4 次是技术问题，这个比例值得记住。","2026-06-11","\u002Fblog\u002Fbaidu-index-volume-drop",{"title":761,"description":762,"date":763,"category":19,"path":764},"我们的审稿清单：一篇文章要过十二关","事实、结构、语言、立场四层共十二项检查，一篇稿子平均改两轮半。清单写出来之后，新人第一稿的返工率明显下降，审稿也不再靠感觉。","2026-06-10","\u002Fblog\u002Feditorial-review-checklist",{"title":766,"description":767,"date":768,"category":31,"path":769},"hreflang 配错的代价，比不做多语言站还大","hreflang 配错的代价比不做多语言还大：Google 会把页面当成重复内容，中文站排名被英文站压。我们处理过 4 个案例，最常见的三种错法、排查方法和一份能直接抄的配置模板，都在这篇，最后附了一个我们自己做的恢复案例，照着查能省一天。","2026-06-09","\u002Fblog\u002Fhreflang-mistakes",{"title":771,"description":772,"date":773,"category":8,"path":774},"Chrome 应用商店的列表页，怎么写才有人装","商店站内搜索是插件主要流量来源，但排序更看留存和评分。我们改了 4 个插件的标题、截图和描述，两个月安装量从 9000 到 2.3 万。","2026-06-08","\u002Fblog\u002Fchrome-web-store-listing-seo",{"title":776,"description":777,"date":778,"category":19,"path":779},"知乎机构号怎么做出搜索流量，别把它当广告位","我们替 4 个客户运营过知乎机构号，两年下来最有效的动作不是发文章，是回答问题。这篇讲选题怎么找、回答怎么写、机构号和个人号怎么配合，附三个实际数据。","2026-06-07","\u002Fblog\u002Fzhihu-org-account-seo",{"title":781,"description":782,"date":783,"category":37,"path":784},"安全与合规页，是软件公司最被低估的信任页","我们看三十个软件站，只有六个有独立的安全页。采购在决策前会搜这些问题，而大部分官网答不上。这篇写六类信任页各写什么、合规表述的边界。","2026-06-06","\u002Fblog\u002Ftrust-pages-security-compliance",{"title":786,"description":787,"date":788,"category":31,"path":789},"多语言内容上线前的九项检查，第四项最常漏","hreflang 配错、语言切换自动跳转、内容只是翻译没重写，是多语言站最常见的三个坑。这份九项检查清单，我们每个多语言项目上线前都过一遍。","2026-06-05","\u002Fblog\u002Fmultilingual-site-checklist",{"title":791,"description":792,"date":793,"category":19,"path":794},"搜索意图会漂移：三年前的文章，可能已经答错了问题","同一个关键词，三年前的搜索结果和现在可能完全不是一回事。我们复查了 40 个老词，有 11 个的意图发生了变化，最典型的是从“了解”变成“购买”。怎么发现、怎么改、什么时候该重写都在这篇，附我们每季度的复查方法。","2026-06-04","\u002Fblog\u002Fkeyword-intent-drift",{"title":796,"description":797,"date":798,"category":31,"path":799},"海外 CDN 在国内慢在哪，选型先看三个指标","一个站换回国内 CDN，首字节从 1.1 秒降到 240 毫秒。这篇写跨境链路的问题、国内选型的三个指标，以及国内海外双线怎么搭。","2026-06-03","\u002Fblog\u002Fdomestic-cdn-selection",{"title":801,"description":802,"date":803,"category":31,"path":804},"rel=next 没用了之后，分页页面怎么做技术 SEO","Google 早就不把 rel=next\u002Fprev 当收录信号，很多站的分页配置还停在五年前。2026 年 4 月我们给 8 个内容站重做了分页逻辑：可抓取、有上限、尾部不 noindex。三个目标、具体做法和一个反常识结论都在这篇。这套做法我们用了两年，8 个站都跑通了。","2026-06-02","\u002Fblog\u002Fpagination-seo-rules",{"title":806,"description":807,"date":808,"category":31,"path":809},"多语言站 hreflang 落地：从生成到校验的流程","hreflang 难的不是语法，是上千个页面的一致性。我们把 hreflang 做成生成器加校验脚本，接进发布流程，新增语言版本不会再漏。这篇给具体做法和三条校验规则。","2026-06-01","\u002Fblog\u002Fhreflang-pipeline-validation",{"title":811,"description":812,"date":813,"category":8,"path":814},"诊断最容易漏的一块，是转化路径这一环","40 个有自然流量的页面，只有 3 个有明确的转化入口。技术、收录、排名都查完了，流量进来了却没人留资，问题往往在这里，我们怎么查、先修哪几页都写在这。","2026-05-31","\u002Fblog\u002Fdiagnosis-conversion-path",{"title":816,"description":817,"date":818,"category":31,"path":819},"六个搜索引擎的收录优先级，我们这么排","六个引擎不能平均用力。我们的收录优先级是 Google、Bing、百度、360、搜狗、神马头条。理由不是流量大小，是抓取效率、收录确定性、流量价值的综合排序，附具体操作顺序，还有一个我们遇到的例外情况，照着排能省两周排查时间。","2026-05-30","\u002Fblog\u002Fsearch-engine-index-priority",{"title":821,"description":822,"date":823,"category":37,"path":824},"B2B 软件做原创数据内容，比写 100 篇博客有用","我们让一个客户停掉每周两篇行业资讯，把调研做成一年一份的年度报告，一年拿到 47 条自然外链。原创数据是软件公司少有的、别人抄不走的内容资产。","2026-05-29","\u002Fblog\u002Fb2b-original-data-content",{"title":826,"description":827,"date":828,"category":8,"path":829},"地图结果和自然排名：两种入口各接什么需求","搜\"深圳 ERP 实施\"，前三条是地图，后面才是网页。这篇讲地图入口和自然入口分别接什么需求、优先级怎么排、以及同时做的时候三个必须对齐的衔接点。","2026-05-28","\u002Fblog\u002Flocal-pack-vs-organic",{"title":831,"description":832,"date":833,"category":8,"path":834},"品牌名搜出来第一页，有几条内容是你控制的","品牌搜索结果页是客户做决策前的最后一站。这篇写该占住的六个位置、怎么盘点、第三方条目怎么挤压，以及实体标记要补什么，附六个必查的品牌查询词。","2026-05-27","\u002Fblog\u002Fbrand-serp-first-page",{"title":836,"description":837,"date":838,"category":19,"path":839},"博客、文档、落地页、案例：四类内容各管什么，别互相抢","软件公司做内容最容易犯的错，是让四类内容干同一件事。博客负责获客、文档负责留存、落地页负责转化、案例负责信任，分工错了会互相抢词。我们的分工表、一个抢词案例和三个判断边界都在这篇。","2026-05-26","\u002Fblog\u002Fcontent-types-division",{"title":841,"description":842,"date":843,"category":19,"path":844},"一天写四篇：我们的批量写作流程和它的代价","一天四篇不是四篇都在写。我们上午写两篇结构固定的短稿，下午写两篇要查资料的，晚上不回看。试了一个月产出 15 篇，但也暴露了代价。","2026-05-25","\u002Fblog\u002Fbatch-writing-workflow",{"title":846,"description":847,"date":848,"category":31,"path":849},"浏览器、CDN、服务端三层缓存，只配一层等于没配","静态资源该缓存一年，HTML 只该缓存五分钟。我们见过把两层配反的站，用户每次访问都在重新下载。这篇写三层各配什么、怎么配错。","2026-05-24","\u002Fblog\u002Fcache-three-layers",{"title":851,"description":852,"date":853,"category":31,"path":854},"尾斜杠和大小写：同一页面两个 URL 的隐形重复","\u002Fdocs\u002Fapi 和 \u002Fdocs\u002Fapi\u002F 是不是同一页，\u002FAbout 和 \u002Fabout 呢？这两类重复比参数更隐蔽。这篇给服务器层的统一规则、五类变体的处理顺序，以及一个 2026 年 1 月的修复记录。","2026-05-23","\u002Fblog\u002Ftrailing-slash-case-duplicate",{"title":856,"description":857,"date":858,"category":31,"path":859},"软 404 比死链更伤：我们用日志抓出 3000 条并清完","软 404 是页面已经没了、服务器还返回 200，搜索引擎会一直抓、一直收、一直不给排名。我们用日志加状态码分布抓出 3000 条，清理后抓取频次涨了 55%。查法、分类和修复顺序都在这篇，清理前后的数据对比在文末，值得每季度跑一次。","2026-05-22","\u002Fblog\u002Fsoft-404-detection",{"title":861,"description":862,"date":863,"category":8,"path":864},"新站和老站的诊断顺序不一样，别用同一张清单","新站先证明能被收录，老站先清历史包袱。这篇写两种站各自的前两个月怎么排：先查什么、跳过什么、什么时候能判断有没有救，附两条判断标准。","2026-05-21","\u002Fblog\u002Fnew-site-vs-old-site-diagnosis",{"title":866,"description":867,"date":868,"category":37,"path":869},"头条搜索的量在短视频那边，图文站别抱太大希望","字节的搜索量大部分发生在抖音和头条 App 内部，纯图文网页拿不到多少。我们拿 5 个站从 2025 年 9 月测到 2026 年 4 月：头条搜索带来的访问平均不到总流量 1%，只有一个品类例外。看完能省下至少一周的试错时间和预算。","2026-05-20","\u002Fblog\u002Ftoutiao-search-traffic-reality",{"title":871,"description":872,"date":873,"category":31,"path":874},"集成页面是 B2B 软件被低估的搜索资产","用户会搜“XX 和 Salesforce 怎么对接”。我们把一个客户从 3 个集成页扩到 46 个，一年带来 2.1 万自然访问，还成交了两个集成需求的大单。","2026-05-19","\u002Fblog\u002Fintegration-pages-seo",{"title":876,"description":877,"date":878,"category":8,"path":879},"商户的地址电话写不一致，本地搜索流量会白白丢掉","我们给一个客户做本地信息核对，在 23 个平台上查出 17 处地址或电话冲突。这篇讲一致性检查的方法、最常见的三种冲突，以及改完之后本地搜索结果和电话量的实际变化。","2026-05-18","\u002Fblog\u002Fnap-consistency-audit",{"title":881,"description":882,"date":883,"category":37,"path":884},"友情链接之外，软件公司还能怎么和别人换资源","换链接的边际收益在降，换内容和渠道更稳。这篇写三种资源互换、谈判时先给对方什么理由、以及合作台账怎么记，附一个我们谈成过的合作形态。","2026-05-17","\u002Fblog\u002Fcontent-partnership-over-links",{"title":886,"description":887,"date":888,"category":8,"path":889},"排名集体下滑时，我们按这六步清单处理老文","2026 年 4 月一个客户 47 个页面三周内排名下滑，技术没动、外链没少。我们按六步清单处理，六周后 31 个回到原位，12 个确认救不回来。","2026-05-16","\u002Fblog\u002Frefresh-losing-rankings",{"title":891,"description":892,"date":893,"category":8,"path":894},"关键词蚕食：两个页面抢同一个词，怎么判断、怎么处理","站内两个页面抢同一个词，表现是排名来回跳、流量没涨。我们查过 18 个站，平均每个站有 6 组蚕食。判断方法、三种处理方式和一次处理后的数据都在这篇，也附了一个常被误判的情况，照着查能少走弯路。","2026-05-15","\u002Fblog\u002Fkeyword-cannibalization",{"title":896,"description":897,"date":898,"category":31,"path":899},"TTFB 超过 800 毫秒，先别急着换服务器","TTFB 由五段组成，换服务器只解决其中一段。这篇写怎么用 curl 把五段时间拆开、后端慢的三种典型情况，以及什么情况下换机器才有用。","2026-05-14","\u002Fblog\u002Fttfb-diagnosis-steps",{"title":901,"description":902,"date":903,"category":31,"path":904},"内链深度超过四层，产品页基本抓不到","我们从 6 个站的日志里统计出点击深度和抓取频次的关系，4 层是分水岭。这篇讲怎么查深度、怎么用内链把重要页面拉浅，附一个文档站把深度从 5.2 压到 2.8 的改造数据。","2026-05-13","\u002Fblog\u002Finternal-link-depth-four",{"title":906,"description":907,"date":908,"category":31,"path":909},"站内 70% 相似的页面，该合并还是保留，我们有个判断标准","站内相似页面处理错了，比不处理更伤。我们用“搜索意图是否相同”做判断：意图相同就合并，意图不同就保留并做差异化。合并的三种做法、两个不该合的例子和一次翻车都在篇里，最后有个判断标准可以直接用。","2026-05-12","\u002Fblog\u002Fduplicate-content-merge",{"title":911,"description":912,"date":913,"category":37,"path":914},"诊断报告别写 80 页，客户只看得下 5 页","我们写过 80 页的诊断报告，客户只翻了前 5 页。这篇写怎么把报告压到 5 页：一页结论、一页问题、一页优先级、一页动作、一页数据附录。","2026-05-11","\u002Fblog\u002Fdiagnosis-report-length",{"title":916,"description":917,"date":918,"category":31,"path":919},"神马搜索只做移动端，这套适配不做等于白做","神马搜索只做移动端，PC 站做得再好也没用。它的三个硬门槛是移动端可访问、落地页规范、备案，缺一个收录就卡住。我们帮 6 个客户做过神马适配，踩过的坑和具体做法都在这篇，最后有一段是适配后的效果数据。","2026-05-10","\u002Fblog\u002Fshenma-mobile-adaptation",{"title":921,"description":922,"date":923,"category":19,"path":924},"采购流程里的那些词，是 B2B 软件最值钱的长尾","RFP 模板、供应商评估表、招标要求，这些词每月搜索量不大，但搜的人已经在走采购流程。我们靠 12 个页面拿到 40 多条企业线索。","2026-05-09","\u002Fblog\u002Frfp-procurement-keywords",{"title":926,"description":927,"date":928,"category":8,"path":929},"百度地图商户入驻，我们踩过的八个坑和改法","从执照名称到门牌号核对，百度地图商户入驻的坑比想象的多。这篇列我们实际遇到的八个问题：名称不一致、地址被合并、电话占位、类目选错等，每条都给改法。","2026-05-08","\u002Fblog\u002Fbaidu-map-merchant-pitfalls",{"title":931,"description":932,"date":933,"category":8,"path":934},"外链一个月加多少条算正常，我们有个经验值","一个站一个月涨了 800 条外链，三个月后掉了一半。这篇写自然增速的量级、突增的三种来源、怎么判断异常，以及我们自己的建链节奏，遇到突增先查来源。","2026-05-07","\u002Fblog\u002Fbacklink-velocity-normal",{"title":936,"description":937,"date":938,"category":19,"path":939},"从 SERP 前 10 反推大纲：我们 30 分钟搭一篇文章骨架的方法","写大纲最省力的方法不是拍脑袋，是把目标词前 10 名的结构全部拆开，看哪些子问题被反复覆盖、哪些没人讲。我们 30 分钟能搭出一篇的骨架，步骤、表格模板和一个反例都在这篇，照着走一遍就能上手。","2026-05-06","\u002Fblog\u002Fserp-outline-reverse",{"title":941,"description":942,"date":943,"category":8,"path":944},"打竞品写得最差的二十个页面，比追爆款有效","竞品写得好的页面你追不上，写得差的页面才有机会。我们按字数、数据、结构、更新时间给 400 多篇竞品文章打分，挑出最差 20 篇，客户做了 8 篇，6 篇进前 30。","2026-05-05","\u002Fblog\u002Fhit-competitor-weak-pages",{"title":946,"description":947,"date":948,"category":31,"path":949},"移动端 CLS 的三个特有来源，桌面测不出来","一个站桌面 CLS 0.02、移动端 0.21，查下来是地址栏、输入法和折叠菜单三个移动端独有的场景。这篇写成因、改法，以及真机怎么测。","2026-05-04","\u002Fblog\u002Fcls-mobile-viewport-shift",{"title":951,"description":952,"date":953,"category":31,"path":954},"大站 sitemap 怎么分片：5 万页站点的三条规则和一次失败","一个 5 万页的 SaaS 文档站，sitemap 之前只有 40% 的页面被收录。我们按内容类型拆成 18 个分片、给 lastmod 立了规矩，2025 年 11 月开始改，8 周后收录率到 78%。分片规则、失败教训和验证方法都在这篇。如果你也在管大站，这篇能省一次返工，少走三个月弯路。","2026-05-03","\u002Fblog\u002Fxml-sitemap-segmentation",{"title":956,"description":957,"date":958,"category":31,"path":959},"分页页的 canonical 要指向自己，别指向第一页","分页页指向第一页、给深层分页加 noindex，这两个做法都在伤害收录。我们按三种分页场景给配置表，附一个 2026 年 3 月的模板站案例：收录从 300 涨到 1100。","2026-05-02","\u002Fblog\u002Fpagination-canonical-rules",{"title":961,"description":962,"date":963,"category":8,"path":964},"接站前 48 小时，我们只做这五件事","接手一个站的头两天，别急着出方案。我们固定做五件事：备份数据、确认权限、跑一次爬虫、读日志、和客户对齐目标，之后才谈改什么。","2026-05-01","\u002Fblog\u002Ffirst-48-hours-site-takeover",{"title":966,"description":967,"date":968,"category":37,"path":969},"Bing 收录慢不是你的问题，是它的抓取节奏","Bing 收录慢是常态，不是你的站有问题。我们监测的 14 个站里，Bing 首批收录平均 9 天，Google 3 天，中文站更慢。这篇讲清它的抓取节奏、哪些动作能加快、哪些纯属浪费预算，别再把时间花在反复提交上，无效操作也列在文末。","2026-04-30","\u002Fblog\u002Fbing-indexing-slow",{"title":971,"description":972,"date":973,"category":8,"path":974},"客户案例页写得像新闻稿，就别指望有搜索流量","案例页是 B2B 软件搜索意图最强的页面之一，但多数写成“某客户上线后效果显著”。我们改了 9 个案例页的结构，半年自然访问涨 2.4 倍。","2026-04-29","\u002Fblog\u002Fb2b-case-study-seo",{"title":976,"description":977,"date":978,"category":8,"path":979},"谷歌商户资料对软件公司有没有用，我们试了八个月","我们给 3 个客户开过谷歌商户资料，八个月的数据看下来：对本地搜索有用，对 B2B 直接询盘几乎没用，但它是品牌词搜索里的一个占位。这篇讲开通流程、适用场景和三个别做的动作。","2026-04-28","\u002Fblog\u002Fgbp-for-software-company",{"title":981,"description":982,"date":983,"category":8,"path":984},"外链的锚文本别都用一个词，分布比数量重要","一个站 320 条外链，七成锚文本是同一个词，排名没起来还多了风险。这篇写五类锚文本怎么配、过度优化怎么查、以及和对方沟通时怎么说。","2026-04-27","\u002Fblog\u002Fbacklink-anchor-text-distribution",{"title":986,"description":987,"date":988,"category":19,"path":989},"簇页写太薄，损失最大的不是它自己是支柱页","一个簇里 12 篇有 7 篇不到 600 字，支柱页一直卡在 40 名。删掉薄页、补了 3 篇有数据的，三个月后支柱页到 12 名。薄页拖累的是整组。","2026-04-26","\u002Fblog\u002Fthin-cluster-pages-risk",{"title":991,"description":992,"date":993,"category":31,"path":994},"站点地图里塞五万个 URL，反而害了收录","sitemap 不是越多越好。我们给一个 5 万条 URL 的站做减法，砍到 1.2 万条后有效页面的展示量涨了 37%。这篇说清该放什么、不该放什么，以及三个判断标准。","2026-04-25","\u002Fblog\u002Fsitemap-what-to-include",{"title":996,"description":997,"date":998,"category":25,"path":999},"自建看板还是买现成工具，我们算了笔账","一个 SEO 看板自建要 3 人天加每月维护，买现成工具一年几千块。这篇按团队规模分三档给结论，附我们踩过的维护成本明细，按团队规模给结论。","2026-04-24","\u002Fblog\u002Fdashboard-build-vs-buy",{"title":1001,"description":1002,"date":1003,"category":31,"path":1004},"robots.txt 在百度和 Google 下的三个不同行为","robots.txt 不是一份通用配置。百度和 Google 对它的解读至少有三处不一样：Crawl-delay 认不认、屏蔽后重新放开要多久、sitemap 声明怎么读。我们两边各踩过一次坑，也写了一版两边都能用的写法，11 个站跑了一年多没出问题。照着抄之前，先看这三处差异。","2026-04-23","\u002Fblog\u002Frobots-txt-google-baidu",{"title":1006,"description":1007,"date":1008,"category":37,"path":1009},"G2、Capterra 这些评测站，跨境软件的第二个官网","海外买家选软件先看评测站，但很多中国团队的 listing 是空的。我们把三个 SaaS 的评测站页面补齐，来自评测站的线索占了新客的三成。","2026-04-22","\u002Fblog\u002Freview-site-listings-crossborder",{"title":1011,"description":1012,"date":1013,"category":31,"path":1014},"页脚那行版权年份和备案号，别随手写","页脚是网站上被复制最多、也最容易被忽略的一块。版权年份停在 2023、备案号做成图片、公司名和备案主体不一致，这三类问题我们每季度都能在客户站上查到。","2026-04-21","\u002Fblog\u002Ffooter-legal-details",{"title":1016,"description":1017,"date":1018,"category":31,"path":1019},"站点地图和 RSS 对 AI 抓取还有多少用：六个站三个月的日志","六个站三个月的日志里，AI 爬虫请求 sitemap 的比例只有 11%，RSS 更低到 2%。这两个文件是发现入口，不是引用入口。真正影响 AI 抓取的是内链深度、更新频率和页面能不能被读到。","2026-04-20","\u002Fblog\u002Fsitemap-rss-ai-crawl",{"title":1021,"description":1022,"date":1023,"category":8,"path":1024},"场景词矩阵：行业乘角色乘流程，一张表挖出 300 个词","场景词比品类词好打，但没人知道从哪挖。我们用一张三列表：行业、角色、流程，交叉组合出 300 个候选词，筛完剩 40 个进排期。","2026-04-19","\u002Fblog\u002Fscenario-keyword-matrix",{"title":1026,"description":1027,"date":1028,"category":31,"path":1029},"BreadcrumbList 标了没生效，九成是层级写错","面包屑结构化数据报错通常不在语法，在层级。我们用 2026 年 1 月一个 4000 页文档站的修复过程，讲清 position 连续性、item 对齐、最后一级这三个坑。","2026-04-18","\u002Fblog\u002Fbreadcrumb-jsonld-position-errors",{"title":1031,"description":1032,"date":1033,"category":37,"path":1034},"月报里最该写的，是这个月没做成什么，为什么","客户流失很少因为数字难看，多半因为不知道你在干嘛、卡在哪。我们的月报固定留一页写没做成的事、原因和需要客户配合的动作，附一份可直接套用的写法。","2026-04-17","\u002Fblog\u002Fmonthly-report-what-failed",{"title":1036,"description":1037,"date":1038,"category":37,"path":1039},"核心更新恢复周期：从掉 40% 到回原位用了 5 个月","核心更新后的恢复不是线性，也不是所有人都能恢复。2025 年 3 月核心更新后，我们一个站掉了 40%，第 5 个月回到原位，中间经历两次假恢复。这 5 个月每周做了什么、哪些动作有用、两次假恢复怎么识别，都记下来了。","2026-04-16","\u002Fblog\u002Fcore-update-recovery-cycle",{"title":1041,"description":1042,"date":1043,"category":19,"path":1044},"卖家工具做内容，别只盯着“亚马逊运营”","跨境卖家工具的搜索词极度分散，按平台、按环节、按站点分开，每个都小。我们做了 40 个细分页，半年自然访问从 2400 到 1.9 万。","2026-04-15","\u002Fblog\u002Famazon-seller-tool-content",{"title":1046,"description":1047,"date":1048,"category":31,"path":1049},"页脚和关于页里的工商信息，怎么写才有用","统一社会信用代码要不要放页脚、品牌名和公司全称不一致怎么办？这篇讲工商信息的三个落点、四种常见不一致，以及我们给客户定的一页纸规范。","2026-04-14","\u002Fblog\u002Fbusiness-registration-footer",{"title":1051,"description":1052,"date":1053,"category":25,"path":1054},"AI 可见性分三层：被读到、被提到、被推荐，三种算法不同","被读到看日志，被提到看回答原文，被推荐看位置和链接，三层的数据来源完全不同。一个客户读到 72%、提到 34%、推荐只有 9%，卡在哪一层就修哪一层，别混着算。","2026-04-13","\u002Fblog\u002Fai-visibility-three-levels",{"title":1056,"description":1057,"date":1058,"category":19,"path":1059},"用自家数据写文章，取数、脱敏和口径怎么处理","第三方数据谁都能引，自家数据只有你有。我们每年用客户站的后台数据写 8 到 10 篇，取数、脱敏、口径三件事都做对，文章才会有信息增量。","2026-04-12","\u002Fblog\u002Ffirst-party-data-article",{"title":1061,"description":1062,"date":1063,"category":31,"path":1064},"Googlebot 和百度蜘蛛的抓取习惯差在哪","同一个站，Googlebot 一天来 1.2 万次，百度蜘蛛来 800 次，这不是资源分配的问题。我们从 5 个站的日志里对比出三个行为差异，以及一个站该怎么分别对待。","2026-04-11","\u002Fblog\u002Fgooglebot-vs-baiduspider",{"title":1066,"description":1067,"date":1068,"category":25,"path":1069},"SEO 数据看板我们做了八个，活下来的只有两个","我们做过 8 个 SEO 看板，最后活下来 2 个。GSC 连接器慢、口径没在数据层统一、图表太多、没人维护，这四件事决定看板能不能用下去。","2026-04-10","\u002Fblog\u002Flooker-studio-seo-dashboard",{"title":1071,"description":1072,"date":1073,"category":31,"path":1074},"死链提交没那么重要，但这三类死链必须处理","死链提交不是高优先级，但有三类必须处理：数量占比高的、有外链指向的、还有搜索流量的。我们处理过 3000 条死链的站，写清楚怎么找、怎么分级、怎么提交才有效，处理完抓取频次两周就回到原来的水平。","2026-04-09","\u002Fblog\u002Fdead-link-submission",{"title":1076,"description":1077,"date":1078,"category":8,"path":1079},"Shopify 应用商店的排名，和你想的不一样","应用商店的搜索排序里，安装量和评论权重远高于关键词密度。我们帮一个库存管理 App 调整 listing，两个月自然安装从月均 120 到 340。","2026-04-08","\u002Fblog\u002Fshopify-app-store-seo",{"title":1081,"description":1082,"date":1083,"category":31,"path":1084},"香港公司和内地公司做官网，备案与信任怎么处理","我们客户里有一半注册在香港或新加坡，官网放哪、要不要备案、页面上公司信息怎么写，是每次都要解释的问题。这篇讲三种主体形态各自的做法、代价和取舍。","2026-04-07","\u002Fblog\u002Fhk-vs-mainland-company-site",{"title":1086,"description":1087,"date":1088,"category":37,"path":1089},"训练数据有截止时间：改完内容要等多久才进 AI 回答","大模型有两套知识：参数里的旧知识和检索到的实时网页。改内容走检索 2 到 6 周见效，走参数要 6 到 12 个月。分清这两条路，才知道该等还是该改。","2026-04-06","\u002Fblog\u002Ftraining-cutoff-content-lag",{"title":1091,"description":1092,"date":1093,"category":19,"path":1094},"内容简报该写什么：写手拿到就能动笔的那一页","简报写得太详细，写手没空间；写得太简单，写出来跑偏。我们的一页纸简报固定七项，其中最关键的是“这篇的独家信息是什么”，写不出来就退回去。","2026-04-05","\u002Fblog\u002Fcontent-brief-template",{"title":1096,"description":1097,"date":1098,"category":31,"path":1099},"GSC 网址检查的五个用法，多数人只用了第一个","网址检查不只是看\"有没有收录\"。我们用它做实时抓取、看渲染后的 HTML、比对 canonical、排查改版、验证修复，这篇给五个具体用法和对应判读，附一个 2026 年 2 月的案例。","2026-04-04","\u002Fblog\u002Fgsc-url-inspection-uses",{"title":1101,"description":1102,"date":1103,"category":8,"path":1104},"周报没人看，问题多半在指标口径不统一","六页周报没人读，压到一页之后客户开始主动问数字。这篇写口径不统一的四个表现、一张指标定义表怎么建，以及周报只该写什么，看完能改。","2026-04-03","\u002Fblog\u002Fweekly-report-metrics-alignment",{"title":1106,"description":1107,"date":1108,"category":37,"path":1109},"百度惊雷、飓风、清风算法，软件站最容易踩哪个","百度几个老算法到现在还在用，软件公司踩的坑也集中在三处。2025 年 8 月2025 年 8 月我们复盘了 12 个被处理的站：清风最多，飓风其次，惊雷踩的人少但后果最重。每个算法的触发条件和自查方法都写了，被处理的站里一半是外包闹的。","2026-04-02","\u002Fblog\u002Fbaidu-jinglei-hurricane-algorithms",{"title":1111,"description":1112,"date":1113,"category":31,"path":1114},"跨境软件做多语言站，hreflang 错了等于白做","我们接手过三个跨境 SaaS 的多语言站，两个的 hreflang 配置是错的。常见的三种错法、怎么用 Search Console 验证，以及什么情况干脆别分语言站。","2026-04-01","\u002Fblog\u002Fcross-border-software-hreflang",{"title":1116,"description":1117,"date":1118,"category":31,"path":1119},"中国站点信任信号清单：备案、工商信息和真人","我们给 40 多个国内软件站做过信任页评审，ICP 备案、公安备案、工商信息、公开电话四项不全的占六成。这篇是一份可逐项打勾的清单，每项都写了怎么查、查到什么算过关。","2026-03-31","\u002Fblog\u002Fchina-site-trust-signals",{"title":1121,"description":1122,"date":1123,"category":88,"path":1124},"品牌语料盘点：五个客户平均 240 个来源，官网只占 30 个","我们把 AI 能读到的关于品牌的页面全列出来，五个客户平均 240 个来源，自家官网只有 30 个。盘点要回答三个问题：它读了什么、读错了什么、缺了什么。","2026-03-30","\u002Fblog\u002Fbrand-corpus-audit",{"title":1126,"description":1127,"date":1128,"category":88,"path":1129},"AI 引用怎么统计：我们手工查了 60 个问题四个月","AI 引用没有现成报表，只能手工查。我们固定 60 个问题，每月丢给 ChatGPT、Perplexity 和国内的几个助手，记录有没有引用我们的页面。四个月下来，规律比数字更有用。","2026-03-29","\u002Fblog\u002Fai-citation-tracking",{"title":1131,"description":1132,"date":1133,"category":31,"path":1134},"改目录结构不掉排名：301 映射表怎么做","从 \u002Fp\u002F12345 改成 \u002Fproducts\u002Fdata-sync，几千个 URL 要映射。我们做过 3 次这类改版，关键是映射表的结构和上线顺序。这篇给可直接套用的模板和检查项。","2026-03-28","\u002Fblog\u002Fdirectory-url-change-301-map",{"title":1136,"description":1137,"date":1138,"category":8,"path":1139},"SEO 脚本没人维护，三个月后一定会崩","我们盘过 11 个自建脚本，4 个已经跑不动。崩的方式有静默出错、部分出错和彻底不跑三种。这篇写维护规矩、交接办法，以及哪些活干脆别自己写。","2026-03-27","\u002Fblog\u002Fseo-scripts-maintenance",{"title":1141,"description":1142,"date":1143,"category":8,"path":1144},"算法更新后别急着改站，先做这三件事","算法更新后第一件事不是改站，是确认三件事：是不是真的更新、掉的是哪类页面、竞品有没有同样掉。这三件事没做完就动手，大概率改错方向。我们踩过这个坑，写了份排查顺序，现在团队按这套顺序走，返工少了很多。","2026-03-26","\u002Fblog\u002Fpost-update-first-three-things",{"title":1146,"description":1147,"date":1148,"category":31,"path":1149},"设计系统文档公开，能带来开发者流量","把组件库文档从内网搬到公开域名后，一个设计系统工具每月多 6000 次自然访问。哪些页面值得公开、怎么组织，以及 token 页为什么最难做好。","2026-03-25","\u002Fblog\u002Fdesign-system-docs-seo",{"title":1151,"description":1152,"date":1153,"category":19,"path":1154},"联系页改版复盘：表单从九个字段砍到四个之后","2025 年 12 月我们给一个客户改联系页，字段从 9 个砍到 4 个，加了真人头像和预约入口。三周后提交率从 2.1% 涨到 5.4%，无效询盘也涨到 31%，这篇讲怎么处理这个副作用。","2026-03-24","\u002Fblog\u002Fcontact-page-rebuild-review",{"title":1156,"description":1157,"date":1158,"category":37,"path":1159},"豆包、元宝、文心更爱引平台内容：一次 300 条引用的统计","统计国内三个引擎 300 条引用，官网只占 21%，知乎 18%、CSDN 14%、公众号 12%。国内做 GEO，官网是事实源，平台才是入口。","2026-03-23","\u002Fblog\u002Fdomestic-ai-citation-sources",{"title":1161,"description":1162,"date":1163,"category":8,"path":1164},"意图错配让一个站两个月掉 34% 流量，我们这样修的","2026 年 1 月一个客户自然流量两个月掉 34%，技术没动、外链没少、文章也没删。问题出在 9 个页面承接了不属于它的意图，拆开之后六周恢复。","2026-03-22","\u002Fblog\u002Fintent-mismatch-fix",{"title":1166,"description":1167,"date":1168,"category":31,"path":1169},"一个官网用了 11 个字重，字体治理比压缩更省","开发者工具站首屏要下 5 个字体文件，总体积 1.1MB。砍到 3 个字重加子集化之后，请求降到 1 个、体积 310KB。附四条判断清单，以及可变字体值不值的实测。","2026-03-21","\u002Fblog\u002Ffont-weight-bloat",{"title":1171,"description":1172,"date":1173,"category":31,"path":1174},"预渲染救不了百度收录？我们测了 4 个站","预渲染在 Google 上通常有效，在百度上不一定。我们用 4 个客户站做了 3 个月对照，2 个收录明显改善，2 个几乎没变化，差异出在三个可查的地方。","2026-03-20","\u002Fblog\u002Fprerender-baidu-indexing-test",{"title":1176,"description":1177,"date":1178,"category":31,"path":1179},"批量改标题和描述，脚本能做和不能做的事","400 个页面批量改标题，60 个改完点击率反而降了。脚本能读、能写、能回滚，但判断哪个标题更好它做不了。这篇写我们的流程和那次回滚。","2026-03-19","\u002Fblog\u002Fbulk-meta-rewrite-script",{"title":1181,"description":1182,"date":1183,"category":37,"path":1184},"视频结果在百度搜索结果里越来越占地方","我们监测的 60 个软件类关键词里，21 个首屏出现视频卡片，比一年前多了 9 个。软件公司现在不做视频，等于把这块位置让给同行。视频怎么被收录、怎么做、成本多少，这篇写清楚。60 个词、22 轮记录，数据都在正文里。","2026-03-18","\u002Fblog\u002Fbaidu-video-serp",{"title":1186,"description":1187,"date":1188,"category":19,"path":1189},"协作工具的用例页，别写成客户行业大杂烩","35 个行业解决方案页，每页 200 字，谁都没排上去。我们把它砍成 8 个团队角色页，每页写清工作流和配置示例，自然流量和试用申请同时涨了。","2026-03-17","\u002Fblog\u002Fcollaboration-tool-use-case-pages",{"title":1191,"description":1192,"date":1193,"category":19,"path":1194},"服务条款不是法律装饰，它是软件公司的信任信号","我们看过 20 多个软件站的服务条款，一半是从同行那里复制改名的。这篇讲条款里真正影响信任和转化的四处：定价与续费、数据归属、SLA、退出机制，附一次丢单的教训。","2026-03-16","\u002Fblog\u002Fterms-of-service-trust-signal",{"title":1196,"description":1197,"date":1198,"category":88,"path":1199},"中英文语料不对称，AI 会把中国软件当成国外产品","一家公司英文文档 200 页、中文官网 12 页，结果中文 AI 回答把它描述成海外工具，报价还写美元。语料结构决定 AI 怎么理解你。","2026-03-15","\u002Fblog\u002Fchinese-english-corpus-imbalance",{"title":1201,"description":1202,"date":1203,"category":19,"path":1204},"对比页怎么写才不像拉踩：先给结论，再写边界","对比页的目的不是证明自己更好，是帮用户做决定。我们改过一个对比页，只是把结构从“我们更强”改成“谁适合谁”，试用率从 1.1% 到 2.6%。","2026-03-14","\u002Fblog\u002Fcomparison-page-tone",{"title":1206,"description":1207,"date":1208,"category":31,"path":1209},"字体放 Google Fonts 还是自托管，国内首屏差了一秒多","同一页面在北京上海广州各测 50 次：Google Fonts 字体请求 P75 1.2 秒、6 次超时 3 秒，自托管 P75 180 毫秒。自托管要做的三件事，以及一个把字体放在源站的坑。","2026-03-13","\u002Fblog\u002Fself-host-fonts-china",{"title":1211,"description":1212,"date":1213,"category":31,"path":1214},"hreflang 和 canonical 打架，多语言站最常见的错","英文页 canonical 指向中文页、中文页 hreflang 指向英文页，这种配置会让两个版本互相抵消。我们用一个 2026 年 1 月的客户站讲清正确关系、三种冲突形态和验证方法。","2026-03-12","\u002Fblog\u002Fhreflang-canonical-conflict",{"title":1216,"description":1217,"date":1218,"category":8,"path":1219},"跑 SEO 杂活的六个定时任务，怎么排才不出错","六个定时任务每周省下六小时，但静默失败比不做更危险。这篇写任务清单、排期原则、心跳告警，以及我们每季度砍任务的规矩，看完能照着搭。","2026-03-11","\u002Fblog\u002Fseo-cron-jobs",{"title":1221,"description":1222,"date":1223,"category":8,"path":1224},"外链权重在降，但这三种链接还在起作用","买的外链基本清零，但有三类链接还在起作用：行业媒体的真实报道、开源项目和文档站的引用、客户和合作伙伴的官网互链。我们按这三类做了 14 个月，附判断一条外链值不值的标准，以及我们不碰的四件事。","2026-03-10","\u002Fblog\u002Fbacklinks-still-work",{"title":1226,"description":1227,"date":1228,"category":8,"path":1229},"设计插件在 Figma 市场里怎么被搜到","Figma 社区的插件列表页，标题和描述就是全部关键词资产，写错了就没人点。我们改了 3 个插件的标题和描述前两行，安装量在两个月里涨了 1.8 倍。","2026-03-09","\u002Fblog\u002Ffigma-plugin-listing-seo",{"title":1231,"description":1232,"date":1233,"category":19,"path":1234},"软件站隐私政策的最小写法：这六段必须写清","隐私政策抄模板，问题不在法律风险，在用户和搜索引擎都能看出你没认真写。这篇给六段必须自己写的内容：收集什么、为什么、存多久、给谁、怎么删、Cookie 怎么管。","2026-03-08","\u002Fblog\u002Fprivacy-policy-minimum-software",{"title":1236,"description":1237,"date":1238,"category":37,"path":1239},"同一篇文章在 6 个 AI 引擎里的引用差异：一张对照表","20 篇内容、60 个查询、6 个引擎。ChatGPT 引文档，Perplexity 引对比，AI Overviews 引表格，豆包引知乎，元宝引公众号。一份内容通吃是幻觉。","2026-03-07","\u002Fblog\u002Fcross-engine-citation-gap",{"title":1241,"description":1242,"date":1243,"category":31,"path":1244},"删掉的文章该 301 还是 410，我们的判断只有三条","2025 年我们给客户删了 88 篇文章，其中只有 12 篇走 410，其余全部 301。判断依据就三条：有没有外链、有没有排名、有没有历史流量。判断错了就是白丢权重。","2026-03-06","\u002Fblog\u002Fdelete-post-301-or-410",{"title":1246,"description":1247,"date":1248,"category":31,"path":1249},"font-display 选 swap 还是 optional，两周实测的数据","两个站各跑两周：文档站用 optional，LCP 1.5 秒、CLS 0.01；营销站用 swap，LCP 1.9 秒、CLS 0.14，加上 preload 和 size-adjust 后 CLS 降到 0.03。","2026-03-05","\u002Fblog\u002Ffont-display-swap-vs-optional",{"title":1251,"description":1252,"date":1253,"category":31,"path":1254},"URL 参数治理：筛选、排序、追踪参数要分开处理","三种参数混在一起治理，是软件站重复内容的源头。我们把参数分三类，给每类定规则：哪些进 canonical、哪些交给服务器清洗、哪些在链接生成时就掐掉。","2026-03-04","\u002Fblog\u002Furl-parameter-governance",{"title":1256,"description":1257,"date":1258,"category":8,"path":1259},"用 GSC API 拉全量数据，别再手点导出了","界面一次只给 1000 行、只留 16 个月，API 每天能拉 2.5 万行、四个维度自由组合。这篇写服务账号怎么配、拉什么、存哪，以及三个坑。","2026-03-03","\u002Fblog\u002Fgsc-api-export-script",{"title":1261,"description":1262,"date":1263,"category":19,"path":1264},"做海外市场的软件公司，别用国内那套内容节奏","把国内的内容节奏搬到海外，是我们在客户身上见过最多的浪费。国内一周 3 篇中等长度有效，海外两周 1 篇深度长文更有效。原因、排期表和去年我们踩过的一个反例，都在这篇。我们给 4 个出海站排过这套节奏，最长跑了 14 个月。","2026-03-02","\u002Fblog\u002Foverseas-content-rhythm",{"title":1266,"description":1267,"date":1268,"category":31,"path":1269},"设计工具官网放交互 demo，代价是搜不到","设计协作工具的官网喜欢用可交互 demo 当首屏，但页面正文常常只剩 80 个字。我们给一个白板工具改了两版，索引文本从 80 字到 1400 字。","2026-03-01","\u002Fblog\u002Fdesign-tool-demo-page-seo",{"title":1271,"description":1272,"date":1273,"category":19,"path":1274},"关于页必须回答的六个问题，少一个都要扣分","关于页被当成公司简介写，所以没人看。这篇列六个必须回答的问题：做什么、不做什么、谁负责、在哪、做了多久、怎么联系，附一次改版前后停留时长与询盘来源的变化。","2026-02-28","\u002Fblog\u002Fabout-page-six-questions",{"title":1276,"description":1277,"date":1278,"category":88,"path":1279},"AI 搜索的第二问最容易丢客户：我们跑了 60 条追问链","第一问被引用不代表第二问还在。60 条追问链实测，出现率从 34% 掉到 19%，第三问只剩 8%。丢的不是流量，是用户的决策过程。附三个接住追问的改法。","2026-02-27","\u002Fblog\u002Ffollowup-question-chain",{"title":1281,"description":1282,"date":1283,"category":31,"path":1284},"簇页内链的规矩：每页链几个、锚文本怎么写","只动内链，不动一个字，六周后 6 个簇的平均排名从 26 到 13。这篇写我们固定用的四条规矩：链回支柱、同簇互链、锚文本唯一、单页出站 5 到 8 个。","2026-02-26","\u002Fblog\u002Fcluster-internal-links",{"title":1286,"description":1287,"date":1288,"category":31,"path":1289},"字体子集上线三个月就失效，问题出在维护上","子集按生成那天的文案切，三个月后新文案缺 17 个字，前端只能回退全量包，首屏字体从 280KB 涨回 2.3MB。把字符集变成构建产物，再加一条缺字监控。","2026-02-25","\u002Fblog\u002Ffont-subset-maintenance",{"title":1291,"description":1292,"date":1293,"category":31,"path":1294},"软件站栏目怎么分：按产品、按场景还是按行业","三种分法我们都试过。软件站的结构要按用户找东西的路径分，不是按公司组织架构分。这篇用两个客户的栏目改造，给出可执行的判断标准和混合结构。","2026-02-24","\u002Fblog\u002Fsite-architecture-by-intent",{"title":1296,"description":1297,"date":1298,"category":25,"path":1299},"把抓取日志做成每日简报，六十行脚本的事","日志看板没人打开，群里每天一条简报却有人讨论。这篇写简报的四个数字、脚本结构、异常判定阈值，以及为什么别做交互看板。文末附阈值设置。","2026-02-23","\u002Fblog\u002Flog-daily-digest-script",{"title":1301,"description":1302,"date":1303,"category":19,"path":1304},"多引擎策略的预算怎么分：7:2:1 还是 5:3:2","按固定比例分预算基本是拍脑袋。我们的分法按“这个引擎能不能验证转化”来：能验证的多投，不能验证的只做基础适配。国内软件公司和出海公司两套分法，附一个失败的分配案例和每季度的调整方法。","2026-02-22","\u002Fblog\u002Fmulti-engine-budget-split",{"title":1306,"description":1307,"date":1308,"category":19,"path":1309},"AI 工具的模板库：做 500 个不如做 50 个","prompt 模板库是 AI 产品最常见的引流页，但大部分是机器生成的空壳。我们做了 50 个可复用的模板，单页平均带来的注册是批量页的 14 倍。","2026-02-21","\u002Fblog\u002Fai-prompt-template-library-seo",{"title":1311,"description":1312,"date":1313,"category":88,"path":1314},"补完作者页和编辑政策，三个月后 AI 引用变了","2026 年 1 月我们给一个客户补了作者页、编辑政策和事实核查说明。三个月后同一组问题在 AI 搜索里的品牌提及从 1 次涨到 6 次，官网引用从 0 到 3 次。这篇讲改了什么、怎么测。","2026-02-20","\u002Fblog\u002Feeat-author-signal-review",{"title":1316,"description":1317,"date":1318,"category":88,"path":1319},"「哪个好」类问题：B2B 软件怎么在 AI 回答里占位","AI 回答对比类问题时最多挂 3 个品牌。不写对比内容，位置就是别人的。自己写对比页的三条规矩：给维度、给场景、明确说不适合谁。","2026-02-19","\u002Fblog\u002Fcomparison-query-geo-playbook",{"title":1321,"description":1322,"date":1323,"category":19,"path":1324},"多语言内容不是翻译，是重写，翻两百篇只带来 3% 流量","一个客户把 200 篇中文博客机翻成英文，一年后英文自然流量只占全站 3%。改成针对本地市场重写 20 篇，六个月后到 11%。翻译和重写是两件事。","2026-02-18","\u002Fblog\u002Fmultilingual-not-translation",{"title":1326,"description":1327,"date":1328,"category":31,"path":1329},"模糊占位图很好看，但它会让 LCP 数字骗人","一个设计工具站 GSC 里 LCP 0.9 秒，用户却反馈打开要三秒。原因是首图用了 20×20 的 base64 模糊图，浏览器把 LCP 记在了低清图上。附判断标准和替代方案。","2026-02-17","\u002Fblog\u002Flqip-blur-lcp-trap",{"title":1331,"description":1332,"date":1333,"category":31,"path":1334},"死链清理：410 比 404 快，但不是所有页都该删","死链不是删了就完事。404、410、301 各对应不同意图。我们把站上死链分成五类按处理方式分组，配一个 2025 年 8 月清理 6000 条死链、404 请求占比从 27% 降到 6% 的记录。","2026-02-16","\u002Fblog\u002Fdead-link-cleanup-404-410",{"title":1336,"description":1337,"date":1338,"category":8,"path":1339},"软件公司写 SEO 内容:别写\"什么是 CRM\",写\"CRM 怎么解决具体问题\"","软件公司写\"什么是 CRM\"这类科普内容,排名起不来,竞争不过百科和媒体,流量进来也不转化。该写解决具体问题的内容,匹配购买意图搜索。这篇讲搜索意图分析和内容选题方法。","2026-02-15","\u002Fblog\u002Fseo-content-writing-guide",{"title":1341,"description":1342,"date":1343,"category":31,"path":1344},"日志里的 404 和 5xx，处理顺序反过来更省事","一个站日志里 5xx 只占 0.7%，却让抓取量掉了六成。这篇写状态码按什么顺序修：5xx 优先、404 分类、301 看跳转次数，附每周检查表。","2026-02-14","\u002Fblog\u002Flog-status-code-triage",{"title":1346,"description":1347,"date":1348,"category":19,"path":1349},"公众号文章怎么写才被微信搜一搜收录","公众号文章被搜一搜收录，关键不在发布后，在发布前：账号原创比例、标题关键词、正文前 200 字、发布时间。我们从 2025 年 3 月起发了 96 篇原创，收录率 71%，把可复制的动作拆成了清单。关键动作只有四个，照着做收录能快三到四天。","2026-02-13","\u002Fblog\u002Fwechat-article-search-indexing",{"title":1351,"description":1352,"date":1353,"category":88,"path":1354},"AI 产品文档怎么写，才会被别的 AI 引用","做 AI 工具的都希望被回答引擎推荐，但文档写成营销腔就没人引用。我们改了定义段、能力边界、失败模式、对比四类内容，一个 agent 平台被引用的次数翻了 5 倍。","2026-02-12","\u002Fblog\u002Fai-agent-docs-geo",{"title":1356,"description":1357,"date":1358,"category":19,"path":1359},"技术作者的履历怎么写：证书、项目、可查痕迹","技术博客的作者页最怕写成简历。这篇讲三样东西怎么写才可验证：证书列几个、项目经历写到什么颗粒度、公开痕迹放哪几个平台，附一个改坏了的反面例子。","2026-02-11","\u002Fblog\u002Ftechnical-author-credentials",{"title":1361,"description":1362,"date":1363,"category":88,"path":1364},"从客服工单里挖 AI 查询：1400 条工单找出 96 个问法","最好的查询库不在关键词工具里，在客服工单里。翻 1400 条工单挖出 96 个真实问法，41 个在 AI 搜索里已有回答但没我们的份。筛选条件只有三条。","2026-02-10","\u002Fblog\u002Fcustomer-ticket-query-mining",{"title":1366,"description":1367,"date":1368,"category":8,"path":1369},"判断搜索意图的五分钟方法：搜一遍看五个位置","意图不用猜，搜索结果前十的页面类型就是答案。我们固定看五个位置：前三条、富媒体、价格、相关问题、文档站，五个位置看完，页面该长什么样就定了。","2026-02-09","\u002Fblog\u002Fserp-intent-reverse-check",{"title":1371,"description":1372,"date":1373,"category":88,"path":1374},"让 AI 引擎记住你的品牌:不只是内容好,还得被反复提及","GEO 不只是把官网内容结构做好。AI 引擎判断一个品牌值不值得引用,看的是多个独立来源有没有反复提及你。第三方提及比自家内容权重高,这篇讲怎么在知乎掘金 CSDN 等平台被提及。","2026-02-08","\u002Fblog\u002Fgeo-brand-mention",{"title":1376,"description":1377,"date":1378,"category":31,"path":1379},"产品演示视频别一进页面就自动播放，流量代价算给你看","一个 SaaS 首页三个播放器脚本 690KB，移动端 86% 的加载是白费的，播放率只有 14%。改成封面图加点击加载后 LCP 从 3.2 秒到 1.8 秒，播放率反而涨了。","2026-02-07","\u002Fblog\u002Fproduct-demo-video-weight",{"title":1381,"description":1382,"date":1383,"category":31,"path":1384},"重复内容自查：先分清四种重复再动手","一发现重复就急着上 canonical，结果越改越乱。我们把重复分成参数型、路径型、跨域型、内容型四种，每种对应不同修法，配一个 2025 年 10 月 4 万 URL 站点的清理记录。","2026-02-06","\u002Fblog\u002Fduplicate-content-four-types",{"title":1386,"description":1387,"date":1388,"category":31,"path":1389},"GSC 抓取统计和服务器日志对不上，原因有四个","GSC 显示 12 万次抓取，日志只有 8 万，差的四万去哪了。这篇写抽样、CDN 缓存、假蜘蛛和统计口径四个原因，以及怎么对齐两个数。","2026-02-05","\u002Fblog\u002Fgsc-crawl-stats-vs-logs",{"title":1391,"description":1392,"date":1393,"category":8,"path":1394},"360 搜索的排名逻辑：老站、备案、地域词","360 搜索的量集中在 PC 端，排名偏好和百度不一样。我们拿 4 个站从 2025 年 10 月测到 2026 年 2 月：备案是门槛，老域名是加分，PC 下载类词是主战场，移动端和新品类词基本没戏。哪个品类能做、哪个别碰，看我们五个月的测试结果。","2026-02-04","\u002Fblog\u002F360-search-ranking-logic",{"title":1396,"description":1397,"date":1398,"category":31,"path":1399},"数据工具的 SQL 示例页，比产品页更容易排上去","分析型数据库的竞争在商业词上很挤，但“怎么写 SQL 查留存率”这类任务页几乎没人认真做。我们做了 60 个带表结构和执行计划的示例页，半年自然访问涨了 3 倍。","2026-02-03","\u002Fblog\u002Fsql-example-pages-seo",{"title":1401,"description":1402,"date":1403,"category":19,"path":1404},"AI 批量生成署名文章，我们为什么不接这种单子","2026 年 1 月有客户拿来 60 篇 AI 稿，要求挂我们团队的名字。我们拒了。这篇讲 AI 署名在作者信号上留下的三个痕迹，以及什么情况下用 AI 辅助写作没问题。","2026-02-02","\u002Fblog\u002Fai-generated-byline-risk",{"title":1406,"description":1407,"date":1408,"category":88,"path":1409},"对话式查询改写：12 个关键词扩成 68 个整句问题","关键词是搜索框的产物，AI 搜索收的是整句。把 12 个核心词按四问法改写，得到 68 个真实问法，其中 57 个现有页面答不了。改写表可以直接抄。","2026-02-01","\u002Fblog\u002Fconversational-query-rewrite",{"title":1411,"description":1412,"date":1413,"category":19,"path":1414},"两篇相似文章合并，301 该怎么做才不漏权重","合并文章最贵的错误是把 301 指向首页。我们合并过一次，第一周流量掉了 40%，第二周才回来。这篇写清判断标准、主文选择、301 顺序和验证方法。","2026-01-31","\u002Fblog\u002Fmerge-similar-posts-301",{"title":1416,"description":1417,"date":1418,"category":31,"path":1419},"懒加载不是越多越好，首屏和折叠线要分开处理","删掉三张首屏图的 loading=\"lazy\"，LCP 从 2.9 秒降到 2.0 秒。首屏立即加载、折叠线下提前 400 像素预载、视频 iframe 等点击再加载，三句话覆盖九成情况。","2026-01-30","\u002Fblog\u002Flazy-loading-placement",{"title":1421,"description":1422,"date":1423,"category":31,"path":1424},"多语言站 sitemap：一个索引文件管 40 个语言版本","40 个语言版本、每个几千页，sitemap 怎么组织才不会互相拖累。我们用 2025 年 12 月一个出海 SaaS 的改造，讲清 sitemapindex 的结构、URL 级 hreflang 写法，以及三个常见错。","2026-01-29","\u002Fblog\u002Fmultilingual-sitemap-structure",{"title":1426,"description":1427,"date":1428,"category":8,"path":1429},"软件公司 SEO 技术检查清单:15 个常见问题","软件公司网站常见的 15 个技术 SEO 问题,按渲染、收录、速度、结构化数据、URL 五类整理。每个写清楚查什么、怎么查、怎么修,照着这份清单逐条过一遍,基本覆盖主要技术坑。","2026-01-28","\u002Fblog\u002Fseo-technical-audit-checklist",{"title":1431,"description":1432,"date":1433,"category":25,"path":1434},"日志分析工具选型：小站别买贵的那套","每天 80 万行日志，云端方案一个月两百美元，本地方案零成本。这篇按日志量分三档给方案，写清 GoAccess、桌面工具和云端各自的适用边界。","2026-01-27","\u002Fblog\u002Flog-analysis-tool-choice",{"title":1436,"description":1437,"date":1438,"category":31,"path":1439},"IndexNow 值不值得接：我们接了 11 个站后的结论","IndexNow 值得接，但它只解决“发现”，不解决“收录”，更不影响百度和 Google。我们 11 个站从 2025 年 7 月接起，Bing 抓取时间从 26 小时降到 4 小时，百度数据没任何变化。接法和无效场景都写了，如果你只做国内，看完能省下两小时接它的时间。","2026-01-26","\u002Fblog\u002Findexnow-worth-it",{"title":1441,"description":1442,"date":1443,"category":88,"path":1444},"开发者工具做 GEO:开发者问 AI 比问百度多","开发者工具公司需要 GEO 多于传统 SEO。开发者查技术方案直接问 ChatGPT 和 Copilot,不搜百度。这篇讲开发者工具怎么做 GEO,让 AI 引擎在推荐工具时引用你,包括文档结构和社区参与方法。","2026-01-25","\u002Fblog\u002Fdevtools-geo-strategy",{"title":1446,"description":1447,"date":1448,"category":88,"path":1449},"AI 搜索带来的流量怎么算：我们用了三个土办法","分析工具里 AI 来源的会话大多归到 direct 或 referral。我们用日志、问卷和品牌词三条线交叉验证，摸清了一个 AI 工具客户 11% 的注册来自回答引擎。","2026-01-24","\u002Fblog\u002Fai-tool-citation-tracking",{"title":1451,"description":1452,"date":1453,"category":8,"path":1454},"博客作者页该写的七个字段，别只放一张头像","作者页是 E-E-A-T 里最容易补、也最容易被糊弄的一块。这篇给七个必填字段：真实姓名、职位、履历、可验证痕迹、社交账号、负责范围、联系方式，附两条验收标准。","2026-01-23","\u002Fblog\u002Fauthor-page-eeat-fields",{"title":1456,"description":1457,"date":1458,"category":88,"path":1459},"改了 200 个段落的主题句，AI 引用率从 14% 涨到 26%","40 篇文章、三种改法对照 8 周。把段尾总结搬到段首那组涨 9 个百分点，单纯加一句主题句那组几乎没动。差别在句子位置，不在句子质量。","2026-01-22","\u002Fblog\u002Fparagraph-topic-sentence-test",{"title":1461,"description":1462,"date":1463,"category":19,"path":1464},"选题池每周维护 30 分钟，超过 60 条就该冻结","选题池不是待办清单，是会腐烂的库存。我们固定每周五花 30 分钟做三件事：加新词、补状态、删过期。池子从 210 条砍到 45 条之后，才真正有人用它。","2026-01-21","\u002Fblog\u002Ftopic-backlog-grooming",{"title":1466,"description":1467,"date":1468,"category":8,"path":1469},"SEO 报告该怎么看:别被漂亮的图表骗了","SEO 公司的报告经常注水:挑好看的关键词、不显示搜索量、用品牌词充数。真正该看的指标和判断 SEO 公司是否摸鱼的方法,以及我们自己的报告怎么写。","2026-01-20","\u002Fblog\u002Fseo-reporting-honest",{"title":1471,"description":1472,"date":1473,"category":31,"path":1474},"把站内 4200 张图转成 WebP 和 AVIF，我们记下了三个数","实测 PNG 转 WebP 小 62%、JPEG 转 WebP 小 34%、再转 AVIF 只多小 18% 但编码慢五倍。三个必踩的坑：质量参数照搬、透明通道丢失、兼容兜底。","2026-01-19","\u002Fblog\u002Fwebp-avif-image-compression",{"title":1476,"description":1477,"date":1478,"category":31,"path":1479},"robots 屏蔽和 noindex，选错会让页面彻底消失","两个指令管的事不一样：robots 拦抓取，noindex 拦索引。我们用 2025 年 11 月一个文档页同时加上两者的案例，讲清选择标准、叠加的风险，以及替代方案。","2026-01-18","\u002Fblog\u002Frobots-block-vs-noindex",{"title":1481,"description":1482,"date":1483,"category":31,"path":1484},"前端渲染的站，工具爬不到的内容怎么查","一个 React 文档站首屏 HTML 里一个标题都没有，渲染后有 42 个。这篇写源码和渲染后 DOM 怎么对比、三种常见缺失，以及改架构还是改内容。","2026-01-17","\u002Fblog\u002Fjs-rendered-site-crawl-check",{"title":1486,"description":1487,"date":1488,"category":31,"path":1489},"Google 和百度的抓取预算，逻辑完全不同","Google 的抓取预算按站点健康度和页面价值分配，百度按站点评价和更新频率分配。两边对低价值页面的处理方式几乎相反，同一套优化动作在两边效果可能相反。我们做了个对照观察，结果和很多教程说的不一样。","2026-01-16","\u002Fblog\u002Fgoogle-baidu-crawl-budget",{"title":1491,"description":1492,"date":1493,"category":19,"path":1494},"安全厂商的技术博客，别写成产品软文","漏洞分析写得好能带来客户，写成“我们的产品可以防”就没人愿意看。我们复盘了 3 个安全博客两年里阅读量最高的文章，找出一个可以复用的内容结构和比例。","2026-01-15","\u002Fblog\u002Fsecurity-blog-vulnerability-analysis",{"title":1496,"description":1497,"date":1498,"category":8,"path":1499},"品牌词从查无此名到日均 40 次，我们做了五个月","一个 B2B 数据工具，2025 年 9 月开始补品牌信号，此前搜公司名只有 3 条结果。五个月后日均品牌搜索 40 次，第一页有 7 条自有或第三方内容。这篇按时间线复盘每一步。","2026-01-14","\u002Fblog\u002Fbrand-search-zero-to-forty",{"title":1501,"description":1502,"date":1503,"category":88,"path":1504},"内容分块怎么做：AI 按块抓取，先弄清你的块边界","手切 40 个页面之后发现，62% 的内容块跨了主题。AI 抓到的块主题混杂，就不敢引用。块边界由 H2 决定，一段不超过 400 字。","2026-01-13","\u002Fblog\u002Fcontent-chunking-for-ai",{"title":1506,"description":1507,"date":1508,"category":8,"path":1509},"工程设计软件的 SEO 痛点:图片和 PDF 抓不到怎么办","CAD、BIM、仿真、PLM 这类工程软件的产品页多是图片和 PDF,搜索引擎抓不到内容。技术债根因、图片文本化、PDF 处理、采购决策路径、长尾词策略,我们做工程软件 SEO 的方法。","2026-01-12","\u002Fblog\u002Fengineering-software-seo",{"title":1511,"description":1512,"date":1513,"category":8,"path":1514},"软件公司的 E-E-A-T 靠三样东西证明，不靠头衔","作者简介写得再漂亮，正文没有可核对的东西，E-E-A-T 就是空的。我们把落地拆成三件事：作者是谁、文章里有一手内容、外部有人提你。","2026-01-11","\u002Fblog\u002Feeat-software-proof",{"title":1516,"description":1517,"date":1518,"category":31,"path":1519},"SSR 和静态生成哪个 TTFB 更低，我们测了三类软件站","同一页面，静态生成 TTFB 90 毫秒、SSR 480 毫秒、纯客户端渲染首字节 120 毫秒但 LCP 差一截。文档站、营销站、后台各该用哪种，附一条判断清单。","2026-01-10","\u002Fblog\u002Fssr-vs-static-ttfb",{"title":1521,"description":1522,"date":1523,"category":31,"path":1524},"结构化数据检测报的错，一半不用改","一个站 340 条标记报错，真正影响展示的只有 27 条。这篇写三类报错怎么区分、两个检测工具为什么结论不同，以及我们按什么顺序修。","2026-01-08","\u002Fblog\u002Fstructured-data-tool-errors",{"title":1526,"description":1527,"date":1528,"category":37,"path":1529},"Google 垃圾更新打的三类站，软件站最易中招","Google 的垃圾内容政策这几年越收越紧，打的不是 AI，是规模化产出没有增量价值的内容。我们复盘了 2025 年被处理的 9 个站，集中在三类，软件公司最容易踩的是寄生页面这一类。每类长什么样、怎么自查、怎么清理都写了。","2026-01-07","\u002Fblog\u002Fgoogle-spam-update-targets",{"title":1531,"description":1532,"date":1533,"category":19,"path":1534},"安全术语页：写得深才有流量，写得浅就是废页","每个安全公司都想做术语库，大部分做成 200 字一条的百科复制，永远排不上去。我们只做 30 条，每条 1500 字，带配置示例和检测方法，反而吃到了搜索流量。","2026-01-06","\u002Fblog\u002Fsecurity-glossary-pages-seo",{"title":1536,"description":1537,"date":1538,"category":88,"path":1539},"让 AI 引擎引用你的 6 个内容结构技巧","GEO 的核心是让内容容易被 AI 引用。结论前置、可引用事实句、数据带出处、FAQ 结构、段落主题句、表格列表,6 个技巧带 before\u002Fafter 例子。","2026-01-05","\u002Fblog\u002Fgeo-content-structure",{"title":1541,"description":1542,"date":1543,"category":37,"path":1544},"品牌搜索结果首页全是负面，先别急着找删帖","负面信息处理的顺序错了，钱就白花。这篇用问答讲清楚：能删的只有哪三类、先做什么后做什么、为什么\"占位\"比\"删帖\"现实，以及踩过的坑。","2026-01-04","\u002Fblog\u002Fserp-reputation-repair-order",{"title":1546,"description":1547,"date":1548,"category":25,"path":1549},"AI 引用率怎么算：分母选错，同一个团队能得出两个结论","同一份数据，用品牌词做分母算出 23%，换成品类词只有 9%。分母决定结论。我们固定 60 个品类词和对比词，每月同一批，口径写进报告。","2026-01-03","\u002Fblog\u002Fai-citation-rate-denominator",{"title":1551,"description":1552,"date":1553,"category":8,"path":1554},"问题词从客服工单里挖，比关键词工具靠谱","最准的词表不在工具里，在工单里。2025 年 11 月我们导了 1200 条工单，抽出 87 个真实搜索词，其中 31 个月搜不到 50，但搜索后台有稳定曝光。","2026-01-02","\u002Fblog\u002Fquestion-words-support-tickets",{"title":1556,"description":1557,"date":1558,"category":31,"path":1559},"静态资源缓存头配错的四种样子，我们逐个拆开","一个软件站的资源审计，四种缓存头错误全占了：HTML 和 hash 资源共用一套头、改文件没改 hash、CDN 上用 no-store、多机 ETag 不一致。附十分钟能跑完的检查清单。","2026-01-01","\u002Fblog\u002Fcache-control-misconfig",{"title":1561,"description":1562,"date":1563,"category":31,"path":1564},"文章和问答的结构化数据，五个字段最容易写错","FAQ 富媒体展示早就收紧了，Article 标记却一直值得做。我们按真实报错整理出五个最常写错的字段：headline、日期、author、FAQ 主体、语言，每个给正确和错误写法。","2025-12-31","\u002Fblog\u002Fjson-ld-article-faq-fields",{"title":1566,"description":1567,"date":1568,"category":25,"path":1569},"几千个页面的死链，我们为什么自己写脚本查","免费版爬虫有 500 URL 上限，付费版能爬但只报状态码。站上一万页、要按流量和权重排优先级时，八十行脚本更管用，这篇写脚本做什么和三个坑。","2025-12-30","\u002Fblog\u002Fcustom-crawler-dead-links",{"title":1571,"description":1572,"date":1573,"category":31,"path":1574},"百度普通收录 API：从手动提交到脚本自动跑","手动提交 URL 费时还容易漏。我们把百度普通收录 API 接进发布流程，每天自动推 30 条，半年省下约 40 小时。接口用法、提交策略、常见错误码和踩过的坑，这篇一次讲完，最后一节是我们半年后的复盘，照着做能省出每周一小时。","2025-12-29","\u002Fblog\u002Fbaidu-api-index-submission",{"title":1576,"description":1577,"date":1578,"category":31,"path":1579},"合规页怎么做才能带来 B2B 搜索流量","SOC 2、ISO 27001 这类合规页，多数公司只放一个 PDF 下载。我们把它做成搜索落地页，一个客户的合规页半年带来 30 多个企业线索。","2025-12-28","\u002Fblog\u002Fsoc2-compliance-page-seo",{"title":1581,"description":1582,"date":1583,"category":8,"path":1584},"品牌搜索占比：最被低估的指标和我们看的那条线","我们给客户做季度复盘，第一个看的不是排名也不是总流量，是品牌词点击占自然点击的比例。2026 年 2 月复盘 11 个站，占比低于 15% 的 4 个站，三个月后自然流量全在跌。","2025-12-27","\u002Fblog\u002Fbrand-search-share-metric",{"title":1586,"description":1587,"date":1588,"category":25,"path":1589},"查自己被不被 AI 引用：五个能手工复现、不花钱的方法","不买工具也能测 AI 可见性。2026 年 1 月我们手测了 12 个站，五个方法里第三个最准、第四个最容易被忽略。每次 40 分钟，关键是抄原文。","2025-12-26","\u002Fblog\u002Fai-visibility-manual-check",{"title":1591,"description":1592,"date":1593,"category":8,"path":1594},"把三个竞品的博客目录做成一张表，能看到什么","2025 年 12 月我们把三个竞品的 431 篇文章做成一张表，客户看完改了自己的计划：不自卑也不加人。表里的六列，比篇数有用得多。","2025-12-25","\u002Fblog\u002Fcompetitor-catalog-benchmark",{"title":1596,"description":1597,"date":1598,"category":31,"path":1599},"新站三个月不收录，八成不是\"权重\"问题","新站收录慢，最常听到的解释是\"新站没权重，等等就好\"。我们查过 11 个三个月以上不收录的新站，7 个是技术问题：robots、JS 渲染、模板 noindex、sitemap 无效。附排查顺序。","2025-12-23","\u002Fblog\u002Fnew-site-indexing-slow",{"title":1601,"description":1602,"date":1603,"category":8,"path":1604},"跨境电商软件的 SEO:出海和国内双市场怎么打","跨境电商软件(ERP、listing、广告工具)要同时打国内和海外市场。双语 SEO、Google 和百度算法差异、外链建设,我们做跨境软件 SEO 的踩坑经验。","2025-12-22","\u002Fblog\u002Fcross-border-software-seo",{"title":1606,"description":1607,"date":1608,"category":31,"path":1609},"爬十万个页面，Screaming Frog 配置先改五处","第一次爬一个 12 万 URL 的文档站，跑完发现 8 万条是重复参数页，等于白爬。这篇写大站爬取前必须改的五个配置，以及限速怎么设才不压垮生产环境。","2025-12-21","\u002Fblog\u002Fscreaming-frog-large-crawl-config",{"title":1611,"description":1612,"date":1613,"category":37,"path":1614},"百度排名机制里，我们测出的三个权重变化","2025 年盯了 28 个软件站，百度排名机制有三个变化比较确定：整站评价压过单页优化、外链权重继续让位给品牌搜索、老内容不更新就往下掉。每个变化都有我们自己的监测数据，也都对应了具体的操作调整。","2025-12-20","\u002Fblog\u002Fbaidu-ranking-weight-changes",{"title":1616,"description":1617,"date":1618,"category":37,"path":1619},"安全软件做内容，先过“可验证”这一关","安全产品的读者会验证你说的每句话，形容词越多信任越少。我们把客户官网的形容词删掉 60%，换成可验证的数字、方法和第三方记录，三个月后询盘质量明显变好。","2025-12-19","\u002Fblog\u002Fsecurity-software-content-trust",{"title":1621,"description":1622,"date":1623,"category":8,"path":1624},"开发者工具怎么做 SEO:和普通软件不一样的地方","开发者工具的目标用户是开发者,他们搜东西不去百度。技术文档、GitHub、Stack Overflow 才是主战场。我们做 dev tools SEO 的经验。","2025-12-18","\u002Fblog\u002Fdevtools-seo-guide",{"title":1626,"description":1627,"date":1628,"category":8,"path":1629},"品牌词的搜索量怎么查：三个数据源对不上怎么办","同一个品牌词，GSC、百度指数、第三方工具给出的数能差五倍。这篇讲三个源各自在量什么、什么时候信哪个，以及我们判断品牌搜索趋势时固定用的那套口径。","2025-12-17","\u002Fblog\u002Fbrand-query-volume-measure",{"title":1631,"description":1632,"date":1633,"category":37,"path":1634},"竞品在 AI 回答里占多少位：一次 320 条回答的人工盘点","给 6 个客户跑了 80 个查询，竞品平均出现在 43% 的 AI 回答里，客户自己只有 19%。先建参照系再谈优化，附查询分类和记录表。","2025-12-16","\u002Fblog\u002Fcompetitor-ai-share-audit",{"title":1636,"description":1637,"date":1638,"category":19,"path":1639},"月度内容复盘只看四个数，其他都是自娱自乐","我们的月度复盘只有四个指标：曝光、点击率、滚动到 75% 的比例、内容带来的注册或咨询。平均排名和字数早就不看了，看了也做不出动作。","2025-12-15","\u002Fblog\u002Fmonthly-content-review-metrics",{"title":1641,"description":1642,"date":1643,"category":31,"path":1644},"TTFB 高先分清是网络还是服务端，别急着换机房","一个 SaaS 首页移动端首字节 1.4 秒，团队准备换云厂商，其实 800 毫秒花在服务端串行接口上。给一条 curl 拆时间的命令，和三个才该怀疑网络的情况。","2025-12-14","\u002Fblog\u002Fttfb-network-vs-server",{"title":1646,"description":1647,"date":1648,"category":31,"path":1649},"日志分析入门：三个命令看蜘蛛在抓什么","不用装工具也能做日志分析。我们用 awk 和 grep 从 nginx 日志里拉出蜘蛛请求量、状态码分布和最费预算的路径，附一个 2025 年 10 月的诊断记录，日抓 1.2 万次里 41% 是无效请求。","2025-12-13","\u002Fblog\u002Fcrawl-log-analysis-commands",{"title":1651,"description":1652,"date":1653,"category":19,"path":1654},"让 AI 写出软件行业的细节，靠三份材料","同一套提示词，给不给材料，出来的稿子差一个档。我们每篇给 AI 准备三份材料：产品文档、真实问答、术语禁区，文章才能有具体参数和真实场景。","2025-12-12","\u002Fblog\u002Fai-writing-software-details",{"title":1656,"description":1657,"date":1658,"category":37,"path":1659},"流量掉 62%，我们用 9 天定位到根因","2025 年 9 月一个客户站流量 9 天掉 62%，所有人都以为是算法更新。我们排了 9 天，根因是改版时换了 1200 个页面 URL、301 链丢了。排查顺序、修复动作和恢复曲线都在这篇，流量掉的时候可以直接照着走，我们把九天的判断都记了下来。","2025-12-11","\u002Fblog\u002Ftraffic-drop-diagnosis-9-days",{"title":1661,"description":1662,"date":1663,"category":88,"path":1664},"GEO 和 SEO 的区别:为什么软件公司两个都要做","GEO 是面向 AI 搜索引擎的优化,和 SEO 互补。SEO 让你排在百度和 Google 前面,GEO 让 ChatGPT 和 Perplexity 引用你。软件公司尤其需要两个都做。","2025-12-10","\u002Fblog\u002Fgeo-vs-seo",{"title":1666,"description":1667,"date":1668,"category":8,"path":1669},"开源版和商业版之间，缺的是三层页面","开源项目转商业版，最大的流失发生在“我知道有企业版，但不知道要不要买”。我们补了三层页面，企业咨询量从月均 4 条到 17 条。","2025-12-09","\u002Fblog\u002Fopen-source-commercial-funnel",{"title":1671,"description":1672,"date":1673,"category":37,"path":1674},"海外发布和国内通稿：开发者工具的钱该投哪个","我们陪 4 个开发者工具做过海外发布，2 个做过国内通稿。两边成本差 10 倍，效果差在别处。这篇对比发布节奏、素材要求和可归因数据，给一个按阶段分配的组合建议。","2025-12-08","\u002Fblog\u002Fproduct-hunt-vs-china-pr",{"title":1676,"description":1677,"date":1678,"category":31,"path":1679},"FAQPage 该不该做：40 篇内容对照实验，差别只有 2 个百分点","20 篇加 FAQPage 结构化数据，20 篇只改 Q&A 格式。六周后引用率 27% 对 25%，差别在噪声范围。真正起作用的是问答格式，不是那段 JSON。","2025-12-07","\u002Fblog\u002Ffaqpage-schema-ai-citation",{"title":1681,"description":1682,"date":1683,"category":8,"path":1684},"商业调研词最容易被浪费，别只写产品页","搜“怎么选”“哪个好”“价格”的人在准备做决定，你给的却是产品功能页。我们改过一个客户的替代页，试用申请从每月 3 个到 11 个，改的不是文案，是信息种类。","2025-12-06","\u002Fblog\u002Fcommercial-investigation-pages",{"title":1686,"description":1687,"date":1688,"category":31,"path":1689},"CLS 是最容易被放过的一项，代价却落在转化上","三项 Core Web Vitals 里 CLS 总被排最后，因为老板感知不到。但它是唯一直接造成误点击的指标：一个 B2B 注册表单页从 0.34 修到 0.06，提交率从 11.2% 到 13.8%。","2025-12-05","\u002Fblog\u002Fcls-ignored-metric",{"title":1691,"description":1692,"date":1693,"category":31,"path":1694},"GSC 索引报告别只看数字，四个分区怎么读","索引报告里\"已编入索引\"只是结果，\"已发现未编入\"和\"已抓取未编入\"才是原因。我们按四个分区讲读法，并用一个 2025 年 11 月的客户站，说明怎么从 1.8 万收录里挑出该修的 600 页。","2025-12-04","\u002Fblog\u002Fgsc-index-report-reading",{"title":1696,"description":1697,"date":1698,"category":25,"path":1699},"内容工具我们试了十几款，最后只留四个","从关键词工具到内容评分平台，一年多试了十几款，砍掉一半。这篇写我们留下哪四个、各自解决什么问题、选型的四条标准，以及为什么评分工具最先被砍。","2025-12-03","\u002Fblog\u002Fcontent-tools-we-keep",{"title":1701,"description":1702,"date":1703,"category":8,"path":1704},"精选摘要怎么抢：一段 40 字的定义就够","精选摘要不是靠写得长，是靠一段能直接抄走的话。我们 2025 年拿下 37 个精选摘要，其中 29 个靠页面里一段 40 到 60 字的定义句。做法、案例和踩过的坑都在这篇。方法不复杂，难的是一个词一个词地抢，我们做了 9 个月才攒到 37 个。","2025-12-02","\u002Fblog\u002Ffeatured-snippet-40-words",{"title":1706,"description":1707,"date":1708,"category":8,"path":1709},"SaaS 公司怎么做 SEO:我们建议的 4 步","SaaS 公司的 SEO 和普通企业站不同,产品文档是核心资产,商业词竞争激烈。我们给 SaaS 客户做的 4 步方法,从技术诊断到内容建设。","2025-12-01","\u002Fblog\u002Fsaas-seo-guide",{"title":1711,"description":1712,"date":1713,"category":19,"path":1714},"GitHub Issues 是被浪费的长尾关键词内容库","一个开源项目 1800 个 issue 里，藏着大量用户真实搜索的问法。我们挑了 40 个改写成文档页，半年带来 2.6 万次自然访问。","2025-11-30","\u002Fblog\u002Fgithub-issues-longtail-content",{"title":1716,"description":1717,"date":1718,"category":19,"path":1719},"媒体投稿老被拒稿，八成问题出在选题和结构上","我们帮客户投过 11 篇行业媒体稿，采用 7 篇。这篇讲编辑真正在筛什么、选题从哪里捞、结构怎么搭，以及投稿邮件里千万别写的那三句话。","2025-11-29","\u002Fblog\u002Ftech-media-pitch-structure",{"title":1721,"description":1722,"date":1723,"category":31,"path":1724},"B2B 软件公司到底要不要屏蔽 AI 爬虫：三类页面的取舍账","内容站屏蔽 AI 爬虫是理智的，软件公司不一定。一个客户 12 月上了一键拦截，三周后品牌词回答里消失。文档页、定价页、白皮书要分开处理。","2025-11-28","\u002Fblog\u002Fblock-ai-crawlers-tradeoff",{"title":1726,"description":1727,"date":1728,"category":88,"path":1729},"问答式内容怎么写才容易被 AI 引用","AI 引擎喜欢问答，不是因为格式好看，是问题句本身就是检索式。问题要照抄用户问法，答案第一句就给结论。我们给一个开发者工具站加了 47 个问题句，三个月里被引用从 2 次到 9 次。","2025-11-27","\u002Fblog\u002Fqa-format-ai-citation",{"title":1731,"description":1732,"date":1733,"category":31,"path":1734},"客服组件把 CLS 顶到零点二八，我们最后是这么治的","一个 B2B SaaS 移动端 CLS 0.28，补了图片宽高完全没用。真凶是 cookie 同意条和客服窗口两个第三方组件，三周压到 0.03，附上我们总结的第三方组件验收规矩。","2025-11-26","\u002Fblog\u002Fcls-chat-widget-fix",{"title":1736,"description":1737,"date":1738,"category":31,"path":1739},"换域名的 12 步清单，少一步掉一批词","域名迁移我们做过 9 次，最惨一次掉了 62% 的自然流量，原因是旧域名 301 只做了一半。这份 12 步清单按时间顺序排，从迁移前 4 周的准备到上线后 8 周的观察。","2025-11-25","\u002Fblog\u002Fdomain-migration-12-steps",{"title":1741,"description":1742,"date":1743,"category":31,"path":1744},"AI 批量生成落地页，我们踩过的两个收录坑","180 个集成页上线 6 周，GSC 只收了 41 页。模板化太重、内链缺失，这两个坑让批量生成的内容白做。这篇写我们后来改的三处和放量前的验证法。","2025-11-24","\u002Fblog\u002Fai-generated-pages-indexing",{"title":1746,"description":1747,"date":1748,"category":8,"path":1749},"排名因素里真正可控的，只有这 6 个","排名因素被讲得太玄。我们把能动手的归成 6 个：内容质量、页面结构、内链、外链来源、站点速度、品牌搜索。其余三个不可控，预算别花在那上面。这篇写清楚每个怎么改，也按投入产出给了个排序，预算紧的时候照着砍。","2025-11-23","\u002Fblog\u002Fcontrollable-ranking-factors",{"title":1751,"description":1752,"date":1753,"category":37,"path":1754},"开源项目的内容该放 GitHub 还是官网","两边都在发内容，结果互相打架。我们给 6 个开源项目重新分工：GitHub 管使用和信任，官网管选型和采购，流量涨得最明显的是对比页。","2025-11-22","\u002Fblog\u002Fopen-source-site-vs-github",{"title":1756,"description":1757,"date":1758,"category":37,"path":1759},"花三万块买行业榜单排名，我们为什么劝客户放弃","去年 11 月一个客户收到某行业榜单的报价，三万块保前五。我们拦下来了，把这三万花在别处，四个月后品牌词月搜索从 130 涨到 250。这篇讲榜单的三种收费形态和判断方法。","2025-11-21","\u002Fblog\u002Fpaid-industry-award-worth-it",{"title":1761,"description":1762,"date":1763,"category":31,"path":1764},"服务器日志里的真 GPTBot：四个识别特征和一次全量统计","2025 年 11 月一个月，四个站共有 4.2 万次 AI 爬虫请求，冒充 GPTBot 的占了 9000 次。四个特征能分辨真假，光看 user-agent 会被骗。","2025-11-20","\u002Fblog\u002Fgptbot-log-analysis",{"title":1766,"description":1767,"date":1768,"category":8,"path":1769},"把两万个关键词压成四张表，我们的词库分层法","词库的价值在删不在存。两万一千个词经过三层过滤，最后只留四张表：核心词、页面词、长尾池、排除词。每张表有明确用途和负责人，每周五更新一次。","2025-11-19","\u002Fblog\u002Fkeyword-library-four-tables",{"title":1771,"description":1772,"date":1773,"category":31,"path":1774},"给图片和嵌入组件写死尺寸，CLS 一周从零点三降到零点零九","一个开发者工具站 CLS 0.31，不是靠玄学优化的：168 个图片补宽高、6 个 iframe 用 aspect-ratio 占位、骨架屏写死真实高度。附查询方法和两段可直接抄的代码。","2025-11-18","\u002Fblog\u002Fcls-image-dimension-fix",{"title":1776,"description":1777,"date":1778,"category":31,"path":1779},"单页应用做 SEO，服务端渲染和预渲染怎么选","三种方案解决同一个问题——让蜘蛛拿到有正文的 HTML，但成本和适用场景差很远。我们用三个客户站的改造结果，给出按站点类型排序的选型顺序和判断清单。","2025-11-17","\u002Fblog\u002Fspa-rendering-ssr-ssg-prerender",{"title":1781,"description":1782,"date":1783,"category":19,"path":1784},"AI 写初稿可以，但有三件事必须人来做","我们从 2025 年 3 月开始用 AI 写初稿，八成文章有它参与，但没有一篇直接发。观点、具体操作、一手数据这三件事决定了文章值不值得被引用。","2025-11-16","\u002Fblog\u002Fai-draft-human-edits",{"title":1786,"description":1787,"date":1788,"category":8,"path":1789},"2025 年百度 SEO 算法变化的 3 个观察","2025 年百度搜索算法持续打压 AI 生成内容和买卖外链。我们基于自身和同行站点的监测,总结 3 个关键变化和应对方法。","2025-11-15","\u002Fblog\u002Fbaidu-seo-algorithm-2025",{"title":1791,"description":1792,"date":1793,"category":8,"path":1794},"国际 SEO 和国内 SEO 的五个根本差异","国际 SEO 和国内 SEO 最大的差别不在算法，在信任从哪来。我们做过 20 多个出海站，从信任来源、关键词工具、外链玩法、内容节奏到技术基础，五个差异一次讲透，附上我们最贵的一次踩坑记录，以及那件事之后定下的规矩。","2025-11-14","\u002Fblog\u002Finternational-vs-domestic-seo",{"title":1796,"description":1797,"date":1798,"category":31,"path":1799},"GitHub README 其实是你流量最大的落地页","一个开源工具的 README 每月被看 4 万次，官网首页才 9000 次。我们把 README 按落地页重写，官网注册涨了 37%，方法就四步。","2025-11-13","\u002Fblog\u002Fgithub-readme-seo",{"title":1801,"description":1802,"date":1803,"category":8,"path":1804},"软件公司目录站清单：十二个值得提交，三类碰不得","这两年我们给 9 家软件公司提过目录站，累计 60 多个提交位。这篇列出 12 个还值得提交的目录、3 类不该碰的站，以及提交资料必须统一的三处写法，附验收方法。","2025-11-12","\u002Fblog\u002Fsaas-directory-submission-list",{"title":1806,"description":1807,"date":1808,"category":31,"path":1809},"llms.txt 上线三个月实测：四个站零抓取，AI 爬虫压根不读它","2025 年 9 月给四个客户站加了 llms.txt，三个月 nginx 日志里 0 次请求。这个文件现在写了也没坏处，但把它当 GEO 主线的人会失望。","2025-11-11","\u002Fblog\u002Fllms-txt-three-month-test",{"title":1811,"description":1812,"date":1813,"category":8,"path":1814},"信息增益不是玄学，它落在一段具体的话里","判断一篇稿子能不能发，我只看有没有一句话是同行写不出来的。实测数据、客户现场、成本和失败记录是四种能造出增益的原料，凑不齐就别写，换题比硬写便宜。","2025-11-10","\u002Fblog\u002Finformation-gain-paragraph",{"title":1816,"description":1817,"date":1818,"category":31,"path":1819},"CLS 高不一定是图片没尺寸，我们查到的四类真凶","今年查的 14 个软件站里，CLS 超 0.25 的 5 个有 4 个主因是字体切换和动态组件，图片没尺寸只占 1 个。四类真凶按实际占比排序，附上每类的查法。","2025-11-09","\u002Fblog\u002Fcls-real-causes",{"title":1821,"description":1822,"date":1823,"category":31,"path":1824},"hreflang 配错比不配更糟，三个必查点","hreflang 不是翻译插件自动生成就完事。我们查过 20 多个多语言软件站，最常见的错是缺自引用、缺 x-default、语言代码写错。这篇给三个必查点和一份验证方法。","2025-11-08","\u002Fblog\u002Fhreflang-config-three-checks",{"title":1826,"description":1827,"date":1828,"category":8,"path":1829},"关键词难度分不能当排期依据，它只算了一半","我们按 KD 排过一个季度的内容计划，12 篇只上去 3 篇。难度分没算站点权重、没有算 SERP 结果类型。这篇写我们改用的评估法和排期规则。","2025-11-07","\u002Fblog\u002Fkeyword-difficulty-not-schedule",{"title":1831,"description":1832,"date":1833,"category":19,"path":1834},"一套内容兼顾百度和 Google，我们只改三处","中英文双站各写一套成本太高。我们给软件公司的做法是：一套底稿只改标题、首段、结构化数据三处，其余复用，内容成本降六成，Google 收录还涨了两三成。三处分别改什么、为什么不用全改，下面按顺序说。","2025-11-06","\u002Fblog\u002Fone-content-baidu-google",{"title":1836,"description":1837,"date":1838,"category":31,"path":1839},"同一套 API 六种语言 SDK，文档别复制六份","SDK 文档按语言复制粘贴，会出现六个内容几乎一样的页面互相抢排名。我们用一个主页面加语言参数的做法，把索引页从 620 个压到 190 个。","2025-11-05","\u002Fblog\u002Fsdk-multilanguage-docs-seo",{"title":1841,"description":1842,"date":1843,"category":31,"path":1844},"我们撤掉 23 条友链之后：排名没掉，抓取频次涨了四成","2025 年 7 月我们建议一个客户撤掉 23 条存量友链，客户最怕排名崩。撤完 6 周的数据：核心词前 10 数量没掉，百度蜘蛛日抓取从 1400 次涨到 2000 次，首页快照从 9 天缩到 2 天。","2025-11-04","\u002Fblog\u002Fremoved-reciprocal-links-review",{"title":1846,"description":1847,"date":1848,"category":88,"path":1849},"产品实体和品牌实体混在一页，AI 会把公司当产品","产品名和公司名写在同一个标题里，AI 会把两条实体记录合并。后果是品牌词回答错、产品对比缺席。我们拆开三个客户，六周后对比类查询开始出现他们。","2025-11-03","\u002Fblog\u002Fproduct-entity-vs-brand-entity",{"title":1851,"description":1852,"date":1853,"category":19,"path":1854},"旧文翻新前先判断：这篇到底值不值得重写","翻新旧文先问三个问题：还有搜索需求吗、有没有排名基础、原文有没有别人没有的东西。40 篇老文里我们只重写了 12 篇，8 周后 9 篇回到前 20，其余直接删或并。","2025-11-02","\u002Fblog\u002Fold-post-rewrite-judgment",{"title":1856,"description":1857,"date":1858,"category":31,"path":1859},"INP 是什么、怎么测、怎么改：客户最常问的八个问题","INP 是替代 FID 的交互指标，200 毫秒合格、500 毫秒算差。八个高频问题：阈值、和 FID 区别、移动端为什么更差、PSI 为什么测不出、加服务器有没有用、多久见效。","2025-11-01","\u002Fblog\u002Finp-common-questions",{"title":1861,"description":1862,"date":1863,"category":31,"path":1864},"软件公司 URL 怎么起名，我们的 6 条规矩","URL 是会被写进文档、发到群里的长期资产，改一次成本极高。这 6 条规矩来自我们改过的三万多条软件站 URL：全小写、短横线、能看出内容、不随版本变、改了必须 301。","2025-10-31","\u002Fblog\u002Furl-naming-rules-software",{"title":1866,"description":1867,"date":1868,"category":25,"path":1869},"从竞品外链里挑可复制的，我们只留三类","3.2 万条竞品外链，筛完值得联系的只有 340 条。这篇写我们的分桶方法、三类值得争取的页面，以及哪些外链碰都别碰，附完整工作流。","2025-10-30","\u002Fblog\u002Fahrefs-backlink-filter",{"title":1871,"description":1872,"date":1873,"category":37,"path":1874},"微信搜一搜的排名逻辑，和百度根本不是一回事","微信搜一搜不看外链、不看域名年龄，看的是账号权重和内容传播。我们运营 3 个公众号测了 8 个月，把影响排名的因素按重要性排了序，也说清哪些事做了没用。3 个号、96 篇原创，每个关键词排第几都有记录。","2025-10-29","\u002Fblog\u002Fwechat-search-ranking-logic",{"title":1876,"description":1877,"date":1878,"category":31,"path":1879},"更新日志页要不要给搜索引擎看：我的答案是分版本","一个 SaaS 的 changelog 有 600 多条记录，全索引等于制造 600 个薄页面。我们把它拆成三类，只放其中一类进索引，结果反而更好。","2025-10-28","\u002Fblog\u002Fchangelog-page-seo",{"title":1881,"description":1882,"date":1883,"category":8,"path":1884},"互链谈判怎么做：从发邮件到撤链的完整流程","今年我们替客户谈过 14 次互链，成 5 次，撤 2 次。这篇给完整流程：第一封邮件怎么写、遇到高价互换怎么拒、问题链接怎么在两周内撤掉，附三段能直接改着用的邮件话术。","2025-10-27","\u002Fblog\u002Freciprocal-link-negotiation",{"title":1886,"description":1887,"date":1888,"category":88,"path":1889},"品牌实体页怎么写：About 页要当成知识图谱的正式入口","客户被 AI 说成\"做 BI 的公司\"，实际做数据管道。问题出在 About 页：没写清做什么、给谁做、在哪。改完 3 周，品牌词回答里的错误描述从 4 处降到 1 处。","2025-10-26","\u002Fblog\u002Fbrand-entity-page-guide",{"title":1891,"description":1892,"date":1893,"category":19,"path":1894},"内容审计的四个判断：哪些该删，哪些该并，哪些该改","312 篇文章审下来，建议删 88 篇、并 41 篇、改 63 篇，剩下 120 篇不动。判断标准不是流量高低，而是曝光、外链、承接页和意图匹配四件事的组合。","2025-10-25","\u002Fblog\u002Fcontent-audit-decide-rules",{"title":1896,"description":1897,"date":1898,"category":31,"path":1899},"B2B 软件站先修 INP 还是 LCP，我的答案是先修交互","统计 9 个 B2B 软件站的用户投诉，7 个指向点了没反应，只有 2 个抱怨打开慢。营销站看 LCP，后台看 INP，混合站按核心页面访问占比决定先修哪个。","2025-10-24","\u002Fblog\u002Finp-before-lcp-priority",{"title":1901,"description":1902,"date":1903,"category":31,"path":1904},"面包屑不只是导航，它决定蜘蛛怎么理解栏目","面包屑是站内结构最便宜的一份说明书。我们给一个 3000 页文档站重做面包屑，内链层级从 6 层压到 3 层，收录率两个月从 41% 涨到 78%。这篇写怎么做、标什么、错在哪。","2025-10-23","\u002Fblog\u002Fbreadcrumb-seo-structure",{"title":1906,"description":1907,"date":1908,"category":25,"path":1909},"用 5118 和爱站查中文词，先知道这两个坑","5118 长尾多但脏数据多，爱站权重方便但滞后一两周。我们挖过一次 5 万词的库，人工筛完能用不到 8000 个。这篇写我们的筛法和三步验证流程。","2025-10-22","\u002Fblog\u002F5118-aizhan-chinese-keywords",{"title":1911,"description":1912,"date":1913,"category":37,"path":1914},"搜狗站长平台还在更新，但这三类站不值得投入","搜狗被腾讯收购后搜索流量还在，但结构变了。我们拿 5 个站试了半年：三类站别在搜狗上花时间，两类可以顺手做，也写了搜狗站长平台的具体操作和真实的收录周期，省得你重复踩坑。数据来自我们自己 5 个在管站点，不是行业报告。","2025-10-21","\u002Fblog\u002Fsogou-webmaster-platform-value",{"title":1916,"description":1917,"date":1918,"category":31,"path":1919},"API 版本化文档的 SEO 处理：旧版该留还是该删","接口升到 v2 之后，v1 文档删了排名会掉，留着又怕重复内容。我们给三个 API 产品做版本迁移，最后都选了同一套处理方式：留页面、加迁移提示、只合并差异小的页。","2025-10-20","\u002Fblog\u002Fapi-versioning-docs-seo",{"title":1921,"description":1922,"date":1923,"category":8,"path":1924},"互链换量可以做，站群那套东西碰都别碰","今年 6 月我们拒了一个互链联盟的合作，对方手里 200 个站，要求三家互链。这篇讲互链的三条红线、为什么三角链最危险，以及软件公司换链的正确姿势。","2025-10-19","\u002Fblog\u002Freciprocal-link-red-lines",{"title":1926,"description":1927,"date":1928,"category":88,"path":1929},"事实句密度不能太高：每千字放 5 条时 AI 引用率最高","30 篇文章按事实句密度分三组测了 6 周。每千字 8 条以上那组引用率最低，只有 11%，因为 AI 会挑走不重要的一句。每千字 3 到 5 条是甜区，引用率 34%。","2025-10-18","\u002Fblog\u002Ffact-density-citation-test",{"title":1931,"description":1932,"date":1933,"category":19,"path":1934},"支柱页加簇页，软件站的博客目录该怎么搭","博客目录不该按发布时间排，该按问题域排。我们给一家安全软件公司把 200 篇文章重组成 6 个簇，支柱页从第 40 名爬到第 9 名，簇页平均曝光涨了 2.3 倍。","2025-10-17","\u002Fblog\u002Fpillar-cluster-blog-structure",{"title":1936,"description":1937,"date":1938,"category":31,"path":1939},"把后台筛选的 INP 从四百二十毫秒压到一百八十","一个安全审计 SaaS 的客户列表页，1200 条数据七个筛选条件，点一下停半秒。虚拟化、把计算挪出事件处理器、埋点改异步三步做完，P75 从 420 毫秒降到 180 毫秒。","2025-10-16","\u002Fblog\u002Finp-improvement-case-study",{"title":1941,"description":1942,"date":1943,"category":31,"path":1944},"重定向链超过两次，抓取预算就白扔了","一个链接从 HTTP 跳 HTTPS、再跳 www、再跳新路径，四次重定向才到目的地。我们清过一条六跳的链，改直连后目标页抓取频次翻了四倍。这篇写链怎么长出来、怎么查、怎么改。","2025-10-15","\u002Fblog\u002Fredirect-chain-cleanup",{"title":1946,"description":1947,"date":1948,"category":25,"path":1949},"关键词工具给的搜索量，软件行业普遍偏低","同一批词，三个工具给出 90、140 和 420 三种量级。软件行业的词太小众，工具样本撑不起来。我们现在的做法是工具只排优先级，绝对值一律用 GSC 曝光校准。","2025-10-14","\u002Fblog\u002Fkeyword-volume-underestimated",{"title":1951,"description":1952,"date":1953,"category":25,"path":1954},"Bing 站长工具的三个设置，每个新站我们都会做","Bing 在国内份额不大，但它的索引还喂给 Copilot 这类产品。我们每个新站上线第一周固定做三个设置：验证并提交 sitemap、接 IndexNow、看搜索性能挖长尾词，20 分钟能做完。今年给 7 个新站配过，收录周期平均快一周。配置不难，难在有人告诉你先做哪一步。","2025-10-13","\u002Fblog\u002Fbing-webmaster-tools-setup",{"title":1956,"description":1957,"date":1958,"category":31,"path":1959},"API 参考页别按模块拆，按用户任务拆","一个支付 API 文档有 210 个端点，原来按模块分了 12 页，改成一个端点一页后自然流量涨了 1.4 倍。具体怎么拆、URL 怎么起、哪些页不该单独做。","2025-10-12","\u002Fblog\u002Fapi-reference-seo-structure",{"title":1961,"description":1962,"date":1963,"category":8,"path":1964},"友情链接尽调：六个字段决定这条链要不要换","今年我们帮 3 个软件站审过 63 条存量友链，退掉了 41 条。这篇给一张六字段核对表：出站链接数、收录、历史快照、主题相关、更新频率、交换方式，每项都附红线值。","2025-10-11","\u002Fblog\u002Ffriend-link-due-diligence-checklist",{"title":1966,"description":1967,"date":1968,"category":88,"path":1969},"首段结论写法：前 80 字答不完一个问题就别想被引用","抽出 62 个客户博客首段，只有 11 个能在 80 字内回答标题承诺的问题。我们重写了 40 个首段，9 周后 AI 引用从 9 次到 27 次。附三种能用的开头句式。","2025-10-10","\u002Fblog\u002Ffirst-paragraph-conclusion-geo",{"title":1971,"description":1972,"date":1973,"category":19,"path":1974},"一个人一年写 50 篇，我们的选题和排期是这么排的","一个人写 50 篇不靠写得更快，靠提前半年把选题排成表。我们 1.5 个人一年发 48 篇，平均每篇 2.5 天，每周只做一件确定的事：把当周那篇写完。","2025-10-09","\u002Fblog\u002Fsolo-writer-yearly-plan",{"title":1976,"description":1977,"date":1978,"category":31,"path":1979},"INP 高但不知道怎么改，先查这四类长任务占了多少","查过 7 个软件站的 INP，四类长任务能解释其中 6 个：框架水合、输入处理器里的同步计算、第三方脚本抢线程、大列表一次渲染。附带定位方法和每类的改法，附实测数字。","2025-10-08","\u002Fblog\u002Finp-long-task-sources",{"title":1981,"description":1982,"date":1983,"category":31,"path":1984},"同一内容多个地址，canonical 和 301 怎么分工","三个手段被混用，是因为没分清\"内容重复\"\"永久搬家\"\"不想被收录\"。我们用一张对照表和 2025 年 9 月一个产品页的案例，讲清各自管什么，以及混用的代价。","2025-10-07","\u002Fblog\u002Fcanonical-vs-301-vs-noindex",{"title":1986,"description":1987,"date":1988,"category":31,"path":1989},"GSC 收录报告里已排除的页面，一半不用处理","已排除听着吓人，其实多数是正常状态。真正要动手的只有已抓取未编入索引和误伤两类。我们按四周对比一次，先说清哪四类不用管，再说该管的两类怎么处理。","2025-10-06","\u002Fblog\u002Fgsc-excluded-pages-triage",{"title":1991,"description":1992,"date":1993,"category":31,"path":1994},"同一篇内容，Google 收录了百度没收，差在哪","同一批 120 篇技术文档，Google 三天收了 118 篇，百度两周只收 21 篇。我们对比了已收录和未收录的页面，差异集中在入口、语言和判重三处，处理完百度收录涨到 76 篇。英文文档站想拿国内流量，得先把中文落地页补上，否则收录再多也没入口。","2025-10-05","\u002Fblog\u002Fgoogle-indexed-baidu-not",{"title":1996,"description":1997,"date":1998,"category":37,"path":1999},"品牌词搜索量才是 SaaS 最该看的 SEO 指标","商业词排名可以买、可以堆内容，品牌词搜索量骗不了人。我们看 SaaS 客户的 SEO 健康度，先看品牌词月度搜索量的曲线，再看别的。","2025-10-04","\u002Fblog\u002Fsaas-brand-term-vs-generic",{"title":2001,"description":2002,"date":2003,"category":8,"path":2004},"客座投稿和目录站：软件公司的预算别平均分","一家 SaaS 去年花 2.6 万买目录套餐，换回 0 个询盘；另一家花 1.1 万投 3 篇行业媒体，来了 21 个注册。这篇对比两种渠道的成本、见效周期和风险，给出 7:3 的预算分法。","2025-10-03","\u002Fblog\u002Fguest-post-vs-directory-submission",{"title":2006,"description":2007,"date":2008,"category":37,"path":2009},"AI 搜索吃掉点击之后，B2B 软件站怎么算这笔账","一个客户半年自然流量掉 18%，注册量反而涨 9%。AI 搜索把点击变成了\"看过就走\"。我们换了一套三指标的算法，比盯会话数靠谱。","2025-10-02","\u002Fblog\u002Fai-search-zero-click-accounting",{"title":2011,"description":2012,"date":2013,"category":19,"path":2014},"竞品内容差距分析，我们从抓竞品目录开始做表","抓三个竞品的文章目录，打四个标签，做一张 400 行的表，空位就出来了：题材空位、角度空位、时间空位。这篇写我们每月做一次的具体步骤和筛选标准。","2025-10-01","\u002Fblog\u002Fcompetitor-content-gap-audit",{"title":2016,"description":2017,"date":2018,"category":31,"path":2019},"定位 LCP 元素的三步查法，找错了改什么都白费","软件站 LCP 优化的常见浪费，是把首屏产品截图当成 LCP 元素，其实真正的是标题。实验室看元素、瀑布图看请求、真实数据看设备差异，三步十分钟定下来。","2025-09-30","\u002Fblog\u002Flcp-element-diagnosis-steps",{"title":2021,"description":2022,"date":2023,"category":31,"path":2024},"sitemap 提交了没反应，先查这 5 个地方","提交 sitemap 不等于收录。我们用 2025 年 7 月一个 8500 条 URL 的站点复盘，其中 3200 条是 302 追踪链接。下面 5 个检查点按顺序过一遍，能排掉九成\"提交无效\"的情况。","2025-09-29","\u002Fblog\u002Fsitemap-submit-no-response",{"title":2026,"description":2027,"date":2028,"category":25,"path":2029},"GA4 里看不出自然搜索的效果，问题在埋点","默认报告只给会话数，给不了注册和试用。我们配一套能看 SEO 转化的 GA4 报表要两小时，配错的话后面每个月都在解释数据为什么对不上，这里写清步骤。","2025-09-28","\u002Fblog\u002Fga4-seo-events-setup",{"title":2031,"description":2032,"date":2033,"category":37,"path":2034},"2025 年 3 月核心更新复盘：12 个站掉了 7 个","这次核心更新不是\"打击 AI 内容\"，是重新分配谁值得被信任。我们 12 个在管站点 7 个流量下滑、5 个上涨，最大跌幅 46%。哪些站掉了、我们改了什么、半年后恢复多少，都在这篇。核心更新是站点评级重排，不是技术修补。","2025-09-27","\u002Fblog\u002Fgoogle-march-2025-core-update-review",{"title":2036,"description":2037,"date":2038,"category":19,"path":2039},"SaaS 做免费小工具引流：算得清账的只有三种","免费工具页能带流量，但大部分做出来只有流量没有注册。我们复盘了 12 个工具页，发现只有算钱的计算器、出结果的生成器、给判断的检测器这三类值得做。","2025-09-26","\u002Fblog\u002Fsaas-free-tool-seo",{"title":2041,"description":2042,"date":2043,"category":8,"path":2044},"外链验收：怎么确认一条链接真的被搜索引擎收录","我们上个月替客户验了 120 条外链，只有 68 条真能被抓到。这篇给一套四步验收流程：查收录、看 dofollow、看引荐流量、30 天复查，附每步用的工具和判断标准。","2025-09-25","\u002Fblog\u002Fbacklink-verification-workflow",{"title":2046,"description":2047,"date":2048,"category":88,"path":2049},"Perplexity 引用偏好：它比另外几家更认第三方","测了 40 个查询，Perplexity 平均每条答案挂 6.4 个来源，官网只占 1.9 个。它按句子挂引用，最认交叉验证过的说法。想被它引用，得先被别人写对。","2025-09-24","\u002Fblog\u002Fperplexity-citation-preferences",{"title":2051,"description":2052,"date":2053,"category":8,"path":2054},"长尾词别按搜索量排序，我们改按四个维度打分","我们把长尾词拆成四个维度打分，给搜索量的权重只有 10 分。改完之后选题结果变了：月搜 30 的词排进计划，月搜 5000 的词被砍掉，这不是勇敢，是算过账。","2025-09-23","\u002Fblog\u002Flongtail-keyword-scoring",{"title":2056,"description":2057,"date":2058,"category":31,"path":2059},"LCP 一点二秒客户还说卡，我们查到了指标外的问题","一个 B2B 数据工具的 LCP 只有 1.2 秒，用户却抱怨打开很慢。原因是骨架屏先渲染、真实数据 2.4 秒才回来。我们补了两条自建埋点，才把用户嘴里的\"慢\"量化出来。","2025-09-22","\u002Fblog\u002Flcp-metric-vs-user-wait",{"title":2061,"description":2062,"date":2063,"category":25,"path":2064},"百度统计和 GA4 的流量差 30%，问题常在这四处","两个统计工具口径天生不同，差 20% 到 40% 都正常。我们做过一次完整对账，差异集中在时区、过滤规则、会话归因和站内引流四个地方，查完就不用再纠结谁在说谎。","2025-09-20","\u002Fblog\u002Fbaidu-tongji-vs-ga4-diff",{"title":2066,"description":2067,"date":2068,"category":25,"path":2069},"百度搜索资源平台 8 个入口，我们每周只看 5 个","平台功能不少，真正影响收录和排名判断的只有几个。我们每周固定看 5 个入口，每个看什么数据、哪些数字是噪音、哪 3 个后台工具出了事才点，一次说清楚。平台后台的数字，和你自己统计里的数对不上是正常的。","2025-09-19","\u002Fblog\u002Fbaidu-search-resource-platform-tools",{"title":2071,"description":2072,"date":2073,"category":31,"path":2074},"SaaS 文档页怎么把人送到注册：我们改过的一条路径","文档带来的是技术评估者，不是最终买单的人。我们给一个 API 监控 SaaS 改了文档页的 4 个位置，90 天里文档到注册的转化从 0.6% 提到 1.9%。","2025-09-18","\u002Fblog\u002Fsaas-docs-signup-path",{"title":2076,"description":2077,"date":2078,"category":8,"path":2079},"给开发者工具站做了三个月外链，最后只留下 7 条","一个 API 网关工具站，三个月、11 个渠道、87 条候选，最后留下 7 条外链。这篇复盘我们的筛选表、被退掉的 80 条长什么样，以及 7 条链三个月后的实际效果。","2025-09-17","\u002Fblog\u002Fdevtools-backlink-audit-review",{"title":2081,"description":2082,"date":2083,"category":88,"path":2084},"ChatGPT 搜索最爱引用的四种页面，产品页排最后","我们手工跑了 120 个 B2B 软件相关查询，记录 ChatGPT 每条回答的来源。文档页、对比页、带更新日期的说明页吃掉了大部分引用，产品页只有 7 条。","2025-09-16","\u002Fblog\u002Fchatgpt-search-cited-page-types",{"title":2086,"description":2087,"date":2088,"category":8,"path":2089},"搜索意图分四类，我们见过软件站三类配错页面","同一个词背后的人想干的是四种不同的事：找答案、找方案、比方案、找产品。四类意图对应四类页面，软件站最常见的错误是配错页面，损失比内容写得不生动大得多。","2025-09-15","\u002Fblog\u002Fsearch-intent-four-types-pages",{"title":2091,"description":2092,"date":2093,"category":31,"path":2094},"官网首屏提速复盘：从四点六秒到一点九秒我们改了什么","一个 Next.js 营销站移动端 LCP 4.6 秒，六周做到 1.9 秒，没换服务器没重写前端。六件事按性价比排序：图片、阻塞请求、字体子集、第三方脚本、首图优先级、边缘缓存。","2025-09-14","\u002Fblog\u002Flcp-optimization-retrospective",{"title":2096,"description":2097,"date":2098,"category":31,"path":2099},"爬虫从发现到入库走四步，卡在哪一步决定怎么修","页面不收录，九成能在发现、抓取、渲染、索引这四步里定位到原因。我们按倒序排查，附每一环的查法，以及 2025 年 6 月一个 1200 页文档站只收录 300 页的复盘。","2025-09-13","\u002Fblog\u002Fcrawler-discovery-to-index",{"title":2101,"description":2102,"date":2103,"category":25,"path":2104},"GSC 平均排名不能直接看，它会掩盖三个问题","平均排名是按曝光加权摊平的一个数，涨了不代表流量涨。我们踩过这个坑之后，月报里改成排名分桶加查询级曝光，这篇文章写清楚怎么读才不会被它误导。","2025-09-12","\u002Fblog\u002Fgsc-average-position-misread",{"title":2106,"description":2107,"date":2108,"category":8,"path":2109},"百度收录慢的三个真实原因，提交不等于收录","百度收录慢，多数不是提交入口的问题。我们今年排查了 43 个软件公司站点，最常见的原因只有三个：抓取频次低、页面质量分不够、站内链接断裂。这篇写清每个原因怎么验证、具体改哪里。提交只是告知，不是收录的开关。","2025-09-11","\u002Fblog\u002Fbaidu-indexing-slow-reasons",{"title":2111,"description":2112,"date":2113,"category":8,"path":2114},"SaaS 定价页 SEO：别把最赚钱的页面做成静态图","我拉了 7 个 SaaS 站的流量数据，定价页只占自然流量 3% 到 5%，却贡献了两成的试用注册。定价页做 SEO 的 5 个动作，以及两个我们踩过的坑。","2025-09-10","\u002Fblog\u002Fsaas-pricing-page-seo",{"title":2116,"description":2117,"date":2118,"category":8,"path":2119},"软件公司买外链：七类渠道里我们只保留了两类","今年 4 月到 9 月我们替 4 家软件公司验了 7 类外链渠道，最后只留 2 类。这篇给出每类渠道的报价区间、验收方式和退掉的具体理由，附我们判断垃圾链的 5 条止损线。","2025-09-09","\u002Fblog\u002Fsoftware-backlink-vetting-rules",{"title":2121,"description":2122,"date":2123,"category":88,"path":2124},"AI Overviews 引用机制：被选中的为什么总是某一段","统计 6 个 B2B 软件客户 47 条 AI Overviews 引用后，我们发现只有 4 条指向首页。AI 是按内容块挑引用，不是按页面挑。这条差异决定你该改哪里。","2025-09-08","\u002Fblog\u002Fai-overview-citation-mechanics",{"title":2126,"description":2127,"date":2128,"category":8,"path":2129},"软件公司做关键词研究，我们固定跑这五步流程","关键词研究不是打开工具导一张表。我们的顺序是来源、去噪、分层、匹配、复检，两万多个词最后进排期的只有 71 个。五步分别怎么做、素材从哪来，都写在下面。","2025-09-07","\u002Fblog\u002Fkeyword-research-five-steps",{"title":2131,"description":2132,"date":2133,"category":31,"path":2134},"首图拖慢首屏，软件公司官网 LCP 卡住的三个真实原因","审计 11 个软件公司官网，8 个的 LCP 元素是首屏产品截图。三个原因最常见：设计稿原图直出、首屏图被加懒加载、首图写在 CSS 背景里。改完平均从 4 秒压到 2.2 秒。","2025-09-06","\u002Fblog\u002Flcp-hero-image-bottleneck",{"title":2136,"description":2137,"date":2138,"category":31,"path":2139},"软件站抓取预算被浪费的 4 个地方和查法","我们给 6 个软件公司站做过日志分析，抓取预算平均有 38% 花在没用的 URL 上。这篇文章列出最常浪费抓取预算的 4 个地方，以及用日志怎么查、怎么改。","2025-09-05","\u002Fblog\u002Fcrawl-budget-diagnosis",{"title":2141,"description":2142,"date":2143,"category":2143,"path":2144},"百度收录怎么查才准，三个口径的差别","[object Object]",null,"\u002Fblog\u002Fbaidu-index-count-methods",{"title":2146,"description":2142,"date":2143,"category":2143,"path":2147},"营销页 HTML 该不该上 CDN 缓存，我们给了一条判断线","\u002Fblog\u002Fcdn-html-cache-rules",{"title":2149,"description":2142,"date":2143,"category":2143,"path":2150},"robots.txt 最常写错的 7 处，每一处都有真实事故","\u002Fblog\u002Frobots-txt-common-mistakes",[2152,2271,2478,2621,2788,2909],{"id":2153,"title":2154,"body":2155,"canonical":2143,"category":2259,"date":13,"description":2260,"draft":2261,"extension":2262,"keywords":2263,"meta":2264,"navigation":2265,"noindex":2261,"ogImage":2143,"path":2266,"seo":2267,"seoDescription":2143,"seoTitle":2143,"source":2268,"stem":2269,"__hash__":2270},"news\u002Fnews\u002Findustry-digest.md","搜索与 AI 行业动态速览（每日自动更新）",{"type":2156,"value":2157,"toc":2252},"minimark",[2158,2172,2176,2213,2216,2246,2249],[2159,2160,2161,2165],"blockquote",{},[2162,2163,2164],"p",{},"本页由官网脚本每天自动汇总外部权威信源，只记录标题、来源、发布日期和原文链接，不改写、不评论原文；内容以链接原文为准。",[2162,2166,2167,2168,2171],{},"原文标题保持原语言（多为英文）。最近更新时间：",[2169,2170,13],"code",{},"（UTC+8，页面每天定时刷新）。",[2173,2174,13],"h3",{"id":2175},"_2026-11-06",[2177,2178,2179,2190,2198,2205],"ul",{},[2180,2181,2182,2189],"li",{},[2183,2184,2188],"a",{"href":2185,"rel":2186},"https:\u002F\u002Fwww.searchenginejournal.com\u002Fgoogle-search-queries-hit-a-5-year-high-and-search-revenue-didnt-follow\u002F595485\u002F",[2187],"nofollow","Google Search Queries Hit A 5-Year High, And Search Revenue Didn’t Follow"," · Search Engine Journal",[2180,2191,2192,2197],{},[2183,2193,2196],{"href":2194,"rel":2195},"https:\u002F\u002Fwww.seroundtable.com\u002Fopenai-updates-chatgpt-ads-bidding-options-42470.html",[2187],"OpenAI Updates ChatGPT Ads Bidding Options For Max Bid or Max CPC"," · Search Engine Roundtable",[2180,2199,2200,2189],{},[2183,2201,2204],{"href":2202,"rel":2203},"https:\u002F\u002Fwww.searchenginejournal.com\u002Fnovember-click-spikes-in-google-search-console-are-bot-traffic-how-to-filter-your-reports\u002F595834\u002F",[2187],"November Click Spikes In Google Search Console Are Bot Traffic – How To Filter Your Reports",[2180,2206,2207,2212],{},[2183,2208,2211],{"href":2209,"rel":2210},"https:\u002F\u002Fblog.google\u002Finnovation-and-ai\u002Ftechnology\u002Fdevelopers-tools\u002Fintroducing-gemini-omni-1-2-flash\u002F",[2187],"Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips"," · Google AI 博客",[2173,2214,18],{"id":2215},"_2026-11-05",[2177,2217,2218,2225,2232,2239],{},[2180,2219,2220,2189],{},[2183,2221,2224],{"href":2222,"rel":2223},"https:\u002F\u002Fwww.searchenginejournal.com\u002Fgoogle-ai-answer-ads-apparel-electronics-report\u002F595984\u002F",[2187],"Ads In Google AI Answers Concentrate In Apparel & Electronics",[2180,2226,2227,2197],{},[2183,2228,2231],{"href":2229,"rel":2230},"https:\u002F\u002Fwww.seroundtable.com\u002Ffacebook-answers-google-42467.html",[2187],"Facebook Answers AI-Generated 6 Million Pages Is Google Scaled Content Abuse",[2180,2233,2234,2189],{},[2183,2235,2238],{"href":2236,"rel":2237},"https:\u002F\u002Fwww.searchenginejournal.com\u002Fgeo-isnt-a-distraction-referral-data-misses-where-ai-shapes-buying-decisions\u002F595730\u002F",[2187],"GEO Isn’t A Distraction – Referral Data Misses Where AI Shapes Buying Decisions",[2180,2240,2241,2197],{},[2183,2242,2245],{"href":2243,"rel":2244},"https:\u002F\u002Fwww.seroundtable.com\u002Fgoogle-ads-expands-ai-content-label-disclosure-42461.html",[2187],"Google Ads Expands AI Content Disclosure Label To More Regions",[2247,2248],"hr",{},[2162,2250,2251],{},"信源简介：Search Engine Roundtable、Search Engine Journal 是海外 SEO 行业媒体；Google AI 博客、OpenAI 官方博客提供 AI 引擎动态。以上链接均指向原文，不转载全文。",{"title":2253,"searchDepth":2254,"depth":2254,"links":2255},"",2,[2256,2258],{"id":2175,"depth":2257,"text":13},3,{"id":2215,"depth":2257,"text":18},"行业动态","每天自动汇总 Search Engine Roundtable、Search Engine Journal、Google AI 博客、OpenAI 官方博客的搜索与 AI 引擎动态。只记录标题、来源、发布日期和原文链接，不改写原文；内容以链接原文为准。",false,"md","SEO动态,GEO动态,搜索引擎算法更新,AI搜索动态",{},true,"\u002Fnews\u002Findustry-digest",{"title":2154,"description":2260},"外部信源自动汇总","news\u002Findustry-digest","OB6v_dirD5pvO3L9MMZeg2T5QQw_-UpmCd2_Kx8RgnE",{"id":2272,"title":2273,"body":2274,"canonical":2143,"category":2470,"date":1458,"description":2471,"draft":2261,"extension":2262,"keywords":2143,"meta":2472,"navigation":2265,"noindex":2261,"ogImage":2143,"path":2473,"seo":2474,"seoDescription":2143,"seoTitle":2143,"source":2475,"stem":2476,"__hash__":2477},"news\u002Fnews\u002Fkimi-web-search.md","Kimi 升级联网搜索,国内 AI 搜索竞争加剧",{"type":2156,"value":2275,"toc":2463},[2276,2279,2284,2287,2301,2304,2308,2311,2393,2396,2399,2402,2408,2413,2418,2423,2427,2430,2433,2439,2445,2451,2454,2457,2460],[2162,2277,2278],{},"Kimi 1 月升级了联网搜索,响应速度和信源覆盖都明显改善。结论先给:国内 AI 搜索市场四家(文心一言、豆包、Kimi、通义)的格局基本成型,但各家抓取偏好差异大,做国内 GEO 不能一套策略打全部。",[2280,2281,2283],"h2",{"id":2282},"kimi-这次升级了什么","Kimi 这次升级了什么",[2162,2285,2286],{},"1 月下旬,月之暗面对 Kimi 的联网搜索做了升级。我们测下来几个变化:",[2177,2288,2289,2292,2295,2298],{},[2180,2290,2291],{},"响应速度比之前快,复杂问题 3-5 秒出结果(之前 8-10 秒)",[2180,2293,2294],{},"信源覆盖更广,知乎、微信公众号、新闻源都进来了",[2180,2296,2297],{},"引用来源比之前多,平均一个回答引用 4-6 个信源(之前 2-3 个)",[2180,2299,2300],{},"多轮追问稳定性提升",[2162,2302,2303],{},"这不是小修小补,是 Kimi 在补联网搜索这块的短板。之前 Kimi 强在长文本理解,弱在实时搜索,这次升级明显是想把搜索这条腿也补上。",[2280,2305,2307],{"id":2306},"国内-ai-搜索四家格局","国内 AI 搜索四家格局",[2162,2309,2310],{},"2025 年下半年到 2026 年初,国内 AI 搜索市场基本形成四家分立的局面:",[2312,2313,2314,2333],"table",{},[2315,2316,2317],"thead",{},[2318,2319,2320,2324,2327,2330],"tr",{},[2321,2322,2323],"th",{},"厂商",[2321,2325,2326],{},"强项",[2321,2328,2329],{},"用户特征",[2321,2331,2332],{},"抓取偏好",[2334,2335,2336,2351,2365,2379],"tbody",{},[2318,2337,2338,2342,2345,2348],{},[2339,2340,2341],"td",{},"文心一言",[2339,2343,2344],{},"百度系资源、资讯全",[2339,2346,2347],{},"大众用户、信息查询",[2339,2349,2350],{},"百度系内容(百家号、百度百科)、新闻源",[2318,2352,2353,2356,2359,2362],{},[2339,2354,2355],{},"豆包",[2339,2357,2358],{},"字节系资源、响应快",[2339,2360,2361],{},"年轻用户、生活查询",[2339,2363,2364],{},"抖音、今日头条、新闻源",[2318,2366,2367,2370,2373,2376],{},[2339,2368,2369],{},"Kimi",[2339,2371,2372],{},"长文本、深度分析",[2339,2374,2375],{},"知识工作者、开发者",[2339,2377,2378],{},"知乎、公众号、技术社区",[2318,2380,2381,2384,2387,2390],{},[2339,2382,2383],{},"通义",[2339,2385,2386],{},"阿里系资源、电商场景",[2339,2388,2389],{},"商务用户、B 端",[2339,2391,2392],{},"淘宝、1688、行业资讯",[2162,2394,2395],{},"四家各有基本盘,短期看格局稳定。Kimi 这次升级后,在\"深度搜索\"这个细分场景上,和文心一言、通义的差距在缩小。",[2280,2397,2398],{"id":2398},"各家抓取偏好差异",[2162,2400,2401],{},"这是做国内 GEO 最要关注的点。四家 AI 引擎的抓取偏好差别很大,同一套内容,在四家被引用的概率差很多。",[2162,2403,2404,2407],{},[2405,2406,2341],"strong",{},"重百度系内容。百家号、百度百科、百度知道在文心一言里权重高。如果你想被文心一言引用,百度系内容布局是必须的。",[2162,2409,2410,2412],{},[2405,2411,2355],{},"重字节系。抖音短视频描述、今日头条文章是主要信源。豆包用户偏生活场景,B 端软件公司在这家能拿到的流量有限。",[2162,2414,2415,2417],{},[2405,2416,2369],{},"重深度内容。知乎高赞回答、公众号长文、技术社区(掘金、CSDN)是主要信源。Kimi 用户里知识工作者和开发者占比高,软件公司在这家值得重点投入。",[2162,2419,2420,2422],{},[2405,2421,2383],{},"重 B 端和电商。1688 商品页、行业资讯站是主要信源。软件公司做 B 端选型内容,通义是个该关注的渠道。",[2280,2424,2426],{"id":2425},"对国内-geo-策略的影响","对国内 GEO 策略的影响",[2162,2428,2429],{},"国内 GEO 和海外 GEO 最大的不同,是引擎多、偏好分散。海外基本就是 ChatGPT + Perplexity 两家,加 Google AI Overviews 三家。国内四家各有偏好,一套内容打不过来。",[2162,2431,2432],{},"我们建议的国内 GEO 策略分三步:",[2162,2434,2435,2438],{},[2405,2436,2437],{},"第一步,按目标用户选主战场。"," 你的用户是开发者,重点投 Kimi(知乎、技术社区);你的用户是企业采购,重点投通义和文心一言(行业资讯、百度系);你的用户是大众,豆包不能漏。",[2162,2440,2441,2444],{},[2405,2442,2443],{},"第二步,内容按引擎偏好分发。"," 同一个产品的对比内容,知乎版写深一点(给 Kimi),公众号版写全一点(给文心一言),行业媒体版写专业一点(给通义)。不是一稿多发,是按引擎偏好调整。",[2162,2446,2447,2450],{},[2405,2448,2449],{},"第三步,做多引擎可见度监测。"," 国内四家加上 ChatGPT(国内开发者也在用),五个引擎每周跑一批关键词,记录引用情况。没有现成工具,得自己搭脚本。",[2280,2452,2453],{"id":2453},"我们怎么看",[2162,2455,2456],{},"Kimi 这次升级是 2026 年国内 AI 搜索竞争的第一枪。后面几个月,文心一言、豆包、通义都会跟进升级,四家的差距会进一步缩小,竞争会更激烈。",[2162,2458,2459],{},"对做 GEO 的人是好事——引擎竞争意味着各家都会优化抓取和引用机制,内容质量好的站被引用的概率会上升。但难度也在涨——四家偏好分散,一套策略打不过来,必须做差异化。",[2162,2461,2462],{},"我们的判断:2026 年国内 GEO 会从\"可选项\"变成软件公司的标配。但和海外不同,国内没有一家独大的 AI 引擎,做多引擎差异化布局,是这一年软件公司要解决的核心问题。早动手的,能在四家都占住位置;晚动手的,只能挤一家。",{"title":2253,"searchDepth":2254,"depth":2254,"links":2464},[2465,2466,2467,2468,2469],{"id":2282,"depth":2254,"text":2283},{"id":2306,"depth":2254,"text":2307},{"id":2398,"depth":2254,"text":2398},{"id":2425,"depth":2254,"text":2426},{"id":2453,"depth":2254,"text":2453},"行业资讯","Kimi 1 月升级联网搜索,国内 AI 搜索市场四家(文心一言、豆包、Kimi、通义)格局基本成型。各家抓取偏好不同,国内 GEO 策略要做差异化。",{},"\u002Fnews\u002Fkimi-web-search",{"title":2273,"description":2471},"月之暗面","news\u002Fkimi-web-search","EvJbWJF2o7NodVp2fzKNDvn8iYmtfo-Pdw3neJQWgDc",{"id":2479,"title":2480,"body":2481,"canonical":2143,"category":2470,"date":1518,"description":2614,"draft":2261,"extension":2262,"keywords":2143,"meta":2615,"navigation":2265,"noindex":2261,"ogImage":2143,"path":2616,"seo":2617,"seoDescription":2143,"seoTitle":2143,"source":2618,"stem":2619,"__hash__":2620},"news\u002Fnews\u002Fgoogle-core-update.md","Google 2025 年第三次核心算法更新:低质内容站继续被清洗",{"type":2156,"value":2482,"toc":2607},[2483,2486,2489,2492,2495,2509,2512,2515,2518,2524,2530,2536,2542,2546,2549,2563,2566,2569,2572,2578,2584,2590,2596,2598,2601,2604],[2162,2484,2485],{},"Google 2025 年第三次核心算法更新 1 月初完成滚动,SearchEngineLand 的跟踪数据显示,这次更新对低质内容站的打击比前两次都重。结论先给:对认真做内容的软件公司官网是利好,流量会有回升;但靠 AI 批量生成内容和堆外链的站,这次掉得厉害。",[2280,2487,2488],{"id":2488},"这次更新是什么",[2162,2490,2491],{},"Google 核心更新一年通常发 3-4 次,2025 年第三次在 12 月底开始,1 月初完成。核心更新不是针对某个作弊手法,是 Google 对\"什么算高质量内容\"的整体判断调整。",[2162,2493,2494],{},"这次更新的特点,根据 SearchEngineLand 和几家 SEO 工具( Semrush、Sistrix)的数据:",[2177,2496,2497,2500,2503,2506],{},[2180,2498,2499],{},"受影响最大的是内容站,尤其是 AI 生成内容占比高的站",[2180,2501,2502],{},"联盟营销站(affiliate)受打击明显",[2180,2504,2505],{},"老牌媒体和品牌官网多数稳定或上升",[2180,2507,2508],{},"论坛和 UGC 站点(Reddit、Quora 等)继续涨",[2162,2510,2511],{},"Sistrix 的数据显示,这次更新期间,排名前 100 的内容站里有约 18% 出现明显波动,其中下跌的占多数。",[2280,2513,2514],{"id":2514},"低质内容站被洗的标准",[2162,2516,2517],{},"Google 没公开\"低质\"的精确判断标准,但从被洗的站反推,几个特征很明显:",[2162,2519,2520,2523],{},[2405,2521,2522],{},"第一,内容信息密度低。"," 一篇文章 800 字,删掉一半意思不变。这类内容在 AI 生成站里最常见。Google 的 Helpful Content 系统对这种内容的识别更准了。",[2162,2525,2526,2529],{},[2405,2527,2528],{},"第二,内容主题和站点定位不匹配。"," 一个软件公司官网发\"如何选牙刷\"\"十大旅游景点\",这是典型为流量而内容,被识别后整站权重受影响。",[2162,2531,2532,2535],{},[2405,2533,2534],{},"第三,E-E-A-T 信号弱。"," Experience(经验)、Expertise(专业)、Authoritativeness(权威)、Trust(信任)。没有作者署名、没有作者专业背景、没有第一手经验的内容,E-E-A-T 分低,这次更新里掉得明显。",[2162,2537,2538,2541],{},[2405,2539,2540],{},"第四,外链质量差。"," 大量低质外链、买卖外链的站,这次继续被清。",[2280,2543,2545],{"id":2544},"对软件公司官网的影响偏正面","对软件公司官网的影响(偏正面)",[2162,2547,2548],{},"软件公司官网这次多数是受益的。我们监测的几个客户站,1 月初自然流量平均涨了 8%-15%。原因不复杂:",[2177,2550,2551,2554,2557,2560],{},[2180,2552,2553],{},"软件公司官网内容主题集中,围绕产品和技术,匹配度高",[2180,2555,2556],{},"多数有真实作者(技术团队),E-E-A-T 信号强",[2180,2558,2559],{},"外链来源相对真实(行业媒体、社区、合作方)",[2180,2561,2562],{},"受打击的联盟站和 AI 内容站掉下去后,排名位置让出来了",[2162,2564,2565],{},"不是所有软件公司都涨。涨得明显的,是过去半年一直在做内容建设的;没动作的,基本持平。这印证了一件事:核心更新对认真做内容的站是奖励,不是惩罚。",[2280,2567,2568],{"id":2568},"怎么做才能不被洗",[2162,2570,2571],{},"如果你想让自己的软件公司官网在下一波核心更新里也站稳,关注几件事:",[2162,2573,2574,2577],{},[2405,2575,2576],{},"把作者署名和专业背景做出来。"," 技术博客每篇要有真实作者,作者页面要有专业背景介绍。这件事成本低,但 E-E-A-T 信号提升明显。",[2162,2579,2580,2583],{},[2405,2581,2582],{},"内容只写自己懂的领域。"," 软件公司官网就写产品、技术、行业。别为了 SEO 去写泛流量话题。Google 现在对主题集中度的判断比一年前强。",[2162,2585,2586,2589],{},[2405,2587,2588],{},"删掉低质旧内容。"," 官网上有两年前写的、现在看信息密度低的内容,该删就删,该重写就重写。留着会拉低整站质量评分。",[2162,2591,2592,2595],{},[2405,2593,2594],{},"外链只做真实的。"," 行业媒体投稿、社区讨论、合作方互链,这些是真实外链。买外链这条路 2026 年彻底走不通了。",[2280,2597,2453],{"id":2453},[2162,2599,2600],{},"Google 这几次核心更新的方向很一致:奖励认真做内容的,洗掉投机取巧的。对软件公司是好事——本来就在认真做内容,之前被低质站挤掉的排名位置,现在开始往回还。",[2162,2602,2603],{},"但别误以为\"不做 SEO 也会涨\"。涨的站都是过去半年在内容上花了力气的。如果去年你什么都没做,这次更新和你关系不大。核心更新是放大器,放大你做对的事,也放大你做错的事。",[2162,2605,2606],{},"下一波核心更新估计在 2026 年 3-4 月。这段时间把作者署名、内容主题集中度、旧内容清理这三件事做了,下一波还能继续受益。",{"title":2253,"searchDepth":2254,"depth":2254,"links":2608},[2609,2610,2611,2612,2613],{"id":2488,"depth":2254,"text":2488},{"id":2514,"depth":2254,"text":2514},{"id":2544,"depth":2254,"text":2545},{"id":2568,"depth":2254,"text":2568},{"id":2453,"depth":2254,"text":2453},"Google 2026 年 1 月核心更新完成,低质内容站继续被清洗。对认真做内容的软件公司官网是利好,但有几点要注意。",{},"\u002Fnews\u002Fgoogle-core-update",{"title":2480,"description":2614},"SearchEngineLand","news\u002Fgoogle-core-update","tMfYP8p4-KAj0R9KD6i3XyI3a9wqkHNXEjTfIzpY8Lw",{"id":2622,"title":2623,"body":2624,"canonical":2143,"category":2470,"date":1688,"description":2781,"draft":2261,"extension":2262,"keywords":2143,"meta":2782,"navigation":2265,"noindex":2261,"ogImage":2143,"path":2783,"seo":2784,"seoDescription":2143,"seoTitle":2143,"source":2785,"stem":2786,"__hash__":2787},"news\u002Fnews\u002Fbaidu-ai-content-policy.md","百度搜索明确:AI 生成内容必须标注,否则降权",{"type":2156,"value":2625,"toc":2774},[2626,2629,2632,2635,2653,2656,2659,2662,2665,2669,2672,2675,2736,2739,2742,2745,2751,2757,2763,2765,2768,2771],[2162,2627,2628],{},"百度 12 月发了新规:AI 生成内容必须明确标注,不标注的会降权。结论先给:这是国内搜索平台第一次把\"AI 内容标注\"写成强制要求,比 Google 严。还在用 AI 批量生成内容不标注的站,建议立刻改。",[2280,2630,2631],{"id":2631},"政策说了什么",[2162,2633,2634],{},"2025 年 12 月初,百度搜索资源平台发文,明确要求站点对 AI 生成内容做标注。核心几条:",[2177,2636,2637,2640,2643,2650],{},[2180,2638,2639],{},"AI 生成或主要靠 AI 生成的内容,要在页面或结构化数据里标注",[2180,2641,2642],{},"不标注被识别出来,直接降权",[2180,2644,2645,2646,2649],{},"标注方式可以是页面可见的说明文字,也可以是 schema 里的 ",[2169,2647,2648],{},"creativeWorkStatus"," 等字段",[2180,2651,2652],{},"纯人工编辑、AI 仅辅助润色的不算\"AI 生成\"",[2162,2654,2655],{},"这个政策不是\"建议\",是\"要求\"。百度没有给宽限期,发文即生效。",[2280,2657,2658],{"id":2658},"不标注会怎样",[2162,2660,2661],{},"百度的原话是\"采取降权等措施\"。我们监测的两个用 AI 批量生成内容的站,12 月中旬自然流量掉了 40% 和 55%。两个站都没做标注,且内容有明显的 AI 痕迹——句长均匀、信息密度低、人称缺失。",[2162,2663,2664],{},"这里要分清一件事:百度打击的不是\"用了 AI\",是\"用了 AI 还藏着不说\"。你用 AI 写初稿、人工深度改写,这种不算违规。违规的是 AI 生成后直接发,还不标。",[2280,2666,2668],{"id":2667},"和-google-政策对比","和 Google 政策对比",[2162,2670,2671],{},"Google 对 AI 内容的态度,2023 年就定调了:\"不反对 AI 内容,反对的是为操纵排名而批量生产的内容。\"Google 没强制要求标注 AI 内容,但它的 SpamBrain 系统会识别低质 AI 内容并降权。",[2162,2673,2674],{},"两边对比一下:",[2312,2676,2677,2690],{},[2315,2678,2679],{},[2318,2680,2681,2684,2687],{},[2321,2682,2683],{},"维度",[2321,2685,2686],{},"百度",[2321,2688,2689],{},"Google",[2334,2691,2692,2703,2714,2725],{},[2318,2693,2694,2697,2700],{},[2339,2695,2696],{},"标注要求",[2339,2698,2699],{},"强制",[2339,2701,2702],{},"不强制",[2318,2704,2705,2708,2711],{},[2339,2706,2707],{},"打击对象",[2339,2709,2710],{},"未标注的 AI 内容",[2339,2712,2713],{},"为排名批量生产的低质内容",[2318,2715,2716,2719,2722],{},[2339,2717,2718],{},"识别手段",[2339,2720,2721],{},"文本特征 + 投诉",[2339,2723,2724],{},"SpamBrain 算法",[2318,2726,2727,2730,2733],{},[2339,2728,2729],{},"处罚方式",[2339,2731,2732],{},"降权",[2339,2734,2735],{},"降权,严重的手动操作",[2162,2737,2738],{},"百度更严。Google 是\"你写得烂就罚你\",百度是\"你用 AI 不标就罚你\"。对国内做内容的来说,百度的政策更难绕。",[2280,2740,2741],{"id":2741},"软件公司要注意什么",[2162,2743,2744],{},"软件公司官网通常 AI 内容占比不高,但有几类内容容易踩雷:",[2162,2746,2747,2750],{},[2405,2748,2749],{},"第一,API 文档和产品介绍。"," 这类内容不少是用 AI 生成初稿再改的。如果改动幅度小,建议加标注。最稳的做法是页面底部加一行\"本文部分内容由 AI 辅助生成并经人工审核\",并加 schema 标注。",[2162,2752,2753,2756],{},[2405,2754,2755],{},"第二,技术博客。"," 如果团队用 AI 写技术博客,改写幅度不够的,要么加标注,要么人工改到位。改到位的标准是:有第一手经验、有具体数字、有时间锚点、有明确观点。通篇正确的废话不算改到位。",[2162,2758,2759,2762],{},[2405,2760,2761],{},"第三,产品对比页。"," 这类页面 AI 生成比例高,且商业属性强,百度的关注度会更高。建议加标注,且对比信息要保证准确——AI 写的对比很容易编数据,人工核对一遍。",[2280,2764,2453],{"id":2453},[2162,2766,2767],{},"支持百度的这个政策。AI 内容不标注,对认真做内容的站不公平。一个团队花三天写的深度文章,流量被一个 AI 五分钟生成的同主题文章抢走,这种事 2024-2025 年一直在发生。强制标注至少让用户和搜索引擎知道内容来源,这是底线。",[2162,2769,2770],{},"但政策执行有不确定性。百度识别 AI 内容的准确率,目前业内反馈参差——有误伤,也有漏网。我们建议的客户策略是\"宁可标多,不要漏标\"。标了不会扣分,漏标被识别会扣分,这笔账很容易算。",[2162,2772,2773],{},"最后一句给还在用 AI 批量生成内容不做改写的团队:这条路 2026 年走不通了。Google 和百度都已经把识别能力拉起来了,继续这条路,降权只是时间问题。把 AI 当辅助工具用,人工把内容质量扛起来,这是唯一可持续的做法。",{"title":2253,"searchDepth":2254,"depth":2254,"links":2775},[2776,2777,2778,2779,2780],{"id":2631,"depth":2254,"text":2631},{"id":2658,"depth":2254,"text":2658},{"id":2667,"depth":2254,"text":2668},{"id":2741,"depth":2254,"text":2741},{"id":2453,"depth":2254,"text":2453},"百度 12 月新规要求 AI 生成内容必须标注。不标注的站会降权。和 Google 相比,百度更严格,软件公司要重视。",{},"\u002Fnews\u002Fbaidu-ai-content-policy",{"title":2623,"description":2781},"百度搜索资源平台","news\u002Fbaidu-ai-content-policy","c5WUfhci769GOGcocHx1nSjBBnk8gR4y2Elpe7DFzk8",{"id":2789,"title":2790,"body":2791,"canonical":2143,"category":2470,"date":1773,"description":2902,"draft":2261,"extension":2262,"keywords":2143,"meta":2903,"navigation":2265,"noindex":2261,"ogImage":2143,"path":2904,"seo":2905,"seoDescription":2143,"seoTitle":2143,"source":2906,"stem":2907,"__hash__":2908},"news\u002Fnews\u002Fperplexity-shopping.md","Perplexity 推出购物功能,AI 搜索的商业化在加速",{"type":2156,"value":2792,"toc":2895},[2793,2796,2800,2803,2806,2809,2812,2815,2818,2826,2829,2833,2836,2850,2853,2856,2860,2863,2866,2872,2878,2884,2886,2889,2892],[2162,2794,2795],{},"Perplexity 11 月推了 Shopping 功能,Pro 用户能用。结论先给:这件事的意义不在\"又能买东西了\",而在于 AI 搜索正式从信息检索走到消费决策。对软件公司来说,B2B 采购场景里 GEO 的权重会跟着往上涨。",[2280,2797,2799],{"id":2798},"perplexity-shopping-是什么","Perplexity Shopping 是什么",[2162,2801,2802],{},"简单说,Perplexity 在回答里直接给商品卡片,带价格、评分、购买链接,Pro 用户能一键结账。之前是\"AI 告诉你哪几个值得考虑\",现在是\"AI 告诉你选哪个 + 帮你买\"。",[2162,2804,2805],{},"功能本身不复杂,真正值得关注的是它背后那套逻辑——AI 不再只是引用信源,它开始做选择。被 AI 选中的商品拿流量,没被选中的出局。这和 ChatGPT Search 的引用竞争是同一个方向,只是 Perplexity 把\"决策\"这一步也吃进去了。",[2280,2807,2808],{"id":2808},"为什么软件公司要关注一个购物功能",[2162,2810,2811],{},"你可能觉得 Shopping 是消费品的事,和 B2B 软件没关系。短期看是这样,但这个模式会蔓延。",[2162,2813,2814],{},"B2B 软件采购已经在用 AI 搜了。我们接触的几个企业客户,采购前调研工具,直接问 Perplexity\"做 CRM 选哪几家\"\"日志监控工具对比\"。Perplexity 给的答案里,被列进去的那几家就是候选名单,没列进去的直接出局。",[2162,2816,2817],{},"当 Perplexity 把 Shopping 那套\"决策+购买\"的能力往 B2B 延伸(这只是时间问题),会出现两种情况:",[2177,2819,2820,2823],{},[2180,2821,2822],{},"采购方让 AI 帮忙筛供应商,直接进采购短名单",[2180,2824,2825],{},"AI 的筛选依据来自它抓到的内容——你的官网、第三方评测、用户社区讨论",[2162,2827,2828],{},"这意味着,你在 AI 引擎里的可见度,直接影响你能不能进采购短名单。",[2280,2830,2832],{"id":2831},"ai-搜索的商业化路径在加速","AI 搜索的商业化路径在加速",[2162,2834,2835],{},"2025 年下半年看下来,AI 搜索的商业化在明显提速。几个节点排一下:",[2177,2837,2838,2841,2844,2847],{},[2180,2839,2840],{},"10 月底,ChatGPT Search 全量开放",[2180,2842,2843],{},"11 月,Perplexity Shopping 上线",[2180,2845,2846],{},"同期,Google 在 AI Overviews 里加更多商业卡片",[2180,2848,2849],{},"国内这边,Kimi、豆包都在试商业化",[2162,2851,2852],{},"这是一个清晰的信号:AI 搜索厂商都意识到,纯信息检索变现难,得往决策和交易走。谁走得快,谁就能把流量变现。",[2162,2854,2855],{},"对做 GEO 的人来说,这意味着我们要关注的不再只是\"被引用\",还要关注\"被选中\"。",[2280,2857,2859],{"id":2858},"geo-在消费决策里的权重在上升","GEO 在消费决策里的权重在上升",[2162,2861,2862],{},"传统 SEO 时代,购物决策的流量入口是 Google Shopping、比价站、评测站。AI 搜索时代,这个入口在往 AI 引擎挪。",[2162,2864,2865],{},"GEO 在这个场景里要做的几件事变了:",[2162,2867,2868,2871],{},[2405,2869,2870],{},"第一,产品页要有明确的\"为什么选我\"。"," AI 在做决策时,会抓产品页的卖点陈述。模糊的\"提升效率、降低成本\"这种话 AI 用不上,要的是具体的\"支持 50 万 QPS\"\"部署 10 分钟上线\"。",[2162,2873,2874,2877],{},[2405,2875,2876],{},"第二,第三方评测和用户讨论要铺开。"," AI 选商品会综合多个信源,只看你官网不够。掘金、知乎、V2EX、行业媒体上要有真实讨论。注意是真实讨论,不是软文——AI 现在对软文的识别比一年前强很多。",[2162,2879,2880,2883],{},[2405,2881,2882],{},"第三,结构化数据要做对。"," Product schema、Review schema 这些,让 AI 能直接读你的产品信息。这一步很多软件公司没做,等于把决策环节的信源权让给了竞品。",[2280,2885,2453],{"id":2453},[2162,2887,2888],{},"Perplexity Shopping 这一步不算大,但方向明确。AI 搜索的商业化会继续走,而且会从消费品往 B2B 蔓延。",[2162,2890,2891],{},"软件公司现在要做的不是等\"B2B 版 Shopping\"上线再动手,而是先把 GEO 基础打好:产品页信息要可被 AI 读取、第三方讨论要铺开、结构化数据要做对。等 AI 搜索真的开始做 B2B 决策,你已经在候选池里了。",[2162,2893,2894],{},"我们判断,2026 年上半年会有至少一家 AI 搜索厂商推出面向 B2B 采购的决策功能。提前布局的软件公司,会在那一波拿到先发优势。晚动手的,得花两三倍力气补课。",{"title":2253,"searchDepth":2254,"depth":2254,"links":2896},[2897,2898,2899,2900,2901],{"id":2798,"depth":2254,"text":2799},{"id":2808,"depth":2254,"text":2808},{"id":2831,"depth":2254,"text":2832},{"id":2858,"depth":2254,"text":2859},{"id":2453,"depth":2254,"text":2453},"Perplexity Shopping 上线意味着 AI 搜索从\"找信息\"走到\"做决策\"。软件采购场景里,GEO 的权重会进一步上升。",{},"\u002Fnews\u002Fperplexity-shopping",{"title":2790,"description":2902},"Perplexity","news\u002Fperplexity-shopping","HMXlTmaVkpOV7nrlSe9Tdf4xkrLuDA7qNygxMa0XBtY",{"id":2910,"title":2911,"body":2912,"canonical":2143,"category":2470,"date":1863,"description":3062,"draft":2261,"extension":2262,"keywords":2143,"meta":3063,"navigation":2265,"noindex":2261,"ogImage":2143,"path":3064,"seo":3065,"seoDescription":2143,"seoTitle":2143,"source":3066,"stem":3067,"__hash__":3068},"news\u002Fnews\u002Fchatgpt-search-launch.md","ChatGPT 搜索正式上线,对软件公司意味着什么",{"type":2156,"value":2913,"toc":3055},[2914,2917,2920,2923,2926,2929,2989,2992,2996,2999,3002,3013,3016,3020,3023,3026,3032,3038,3044,3046,3049,3052],[2162,2915,2916],{},"ChatGPT Search 全量开放了,免费用户也能用。结论先放这:对软件公司来说,GEO 已经不是\"再做做看\"的事,而是和 SEO 平级的基础工作。如果你的文档和博客在 ChatGPT 里搜不到,2026 年你会丢掉一部分开发者流量。",[2280,2918,2919],{"id":2919},"这次更新到底是什么",[2162,2921,2922],{},"2025 年 10 月底,OpenAI 把 ChatGPT Search(原来叫 SearchGPT)向所有用户开放,免费版也能用,不再需要候补名单。功能上和之前测试版差别不大:实时联网检索、给来源链接、可以追问。",[2162,2924,2925],{},"但真正变了的是\"使用者基数\"。之前是部分用户,现在是所有 ChatGPT 用户。StatCounter 10 月的数据显示,chatgpt.com 的全球访问量已经稳居前列,Search 全量开放后,这个基数还会往上走。",[2280,2927,2928],{"id":2928},"和传统搜索的区别",[2312,2930,2931,2943],{},[2315,2932,2933],{},[2318,2934,2935,2937,2940],{},[2321,2936,2683],{},[2321,2938,2939],{},"Google\u002F百度",[2321,2941,2942],{},"ChatGPT Search",[2334,2944,2945,2956,2967,2978],{},[2318,2946,2947,2950,2953],{},[2339,2948,2949],{},"返回形式",[2339,2951,2952],{},"10 条蓝色链接 + 摘要",[2339,2954,2955],{},"一段综合回答 + 引用来源",[2318,2957,2958,2961,2964],{},[2339,2959,2960],{},"用户行为",[2339,2962,2963],{},"自己挑、自己点",[2339,2965,2966],{},"AI 帮你挑,直接给答案",[2318,2968,2969,2972,2975],{},[2339,2970,2971],{},"流量分配",[2339,2973,2974],{},"前 3 名吃掉大部分点击",[2339,2976,2977],{},"被引用的那几个信源吃流量",[2318,2979,2980,2983,2986],{},[2339,2981,2982],{},"适合内容",[2339,2984,2985],{},"关键词覆盖型",[2339,2987,2988],{},"结论清晰、可被引用型",[2162,2990,2991],{},"差别一句话:传统搜索是\"列表竞争\",ChatGPT 搜索是\"被引用竞争\"。你不需要排第一,你需要被 AI 选为信源。",[2280,2993,2995],{"id":2994},"开发者已经把搜索挪到-chatgpt-了","开发者已经把搜索挪到 ChatGPT 了",[2162,2997,2998],{},"这是我们自己的观察,不一定代表全行业。过去半年我们接触的开发者里,问技术问题直接用 ChatGPT\u002FClaude 的比例明显升高。GitHub Copilot 的调研也提到,2025 年开发者用 AI 搜索替代 Google 的比例在涨。",[2162,3000,3001],{},"具体到几个场景:",[2177,3003,3004,3007,3010],{},[2180,3005,3006],{},"报错信息排查:直接贴报错给 ChatGPT,比翻 Stack Overflow 快",[2180,3008,3009],{},"库的用法对比:问\"React Query 和 SWR 怎么选\",AI 给对比表",[2180,3011,3012],{},"选型调研:问\"做日志聚合选哪家的工具\",AI 列几个候选 + 优劣",[2162,3014,3015],{},"如果你的产品在这些场景里不被 ChatGPT 提到,你就是不在候选名单里。",[2280,3017,3019],{"id":3018},"对软件公司-geo-的影响","对软件公司 GEO 的影响",[2162,3021,3022],{},"这次全量开放把 GEO 的时间窗往前推了至少半年。我们之前说 GEO 见效周期是 2-4 个月,现在这个窗口在缩短,因为用户基数大了,反馈循环快了。",[2162,3024,3025],{},"软件公司要重点关注三件事:",[2162,3027,3028,3031],{},[2405,3029,3030],{},"第一,官方文档被 ChatGPT 抓到。"," ChatGPT Search 走的是 Bing + 自有爬虫的混合检索。你的 docs 站点要让爬虫能抓——检查 robots.txt、sitemap、是否被 Bing 收录。我们见过几个客户的文档站屏蔽了所有爬虫,然后问\"为什么 ChatGPT 不引用我们\",这是低级错误。",[2162,3033,3034,3037],{},[2405,3035,3036],{},"第二,产品词和对比词的内容布局。"," \"XX vs YY\"\"XX 替代品\"\"XX 怎么样\"这类查询,ChatGPT 会综合多个信源回答。你要在外部有内容(掘金、知乎、行业媒体),让 AI 有更多理由提到你。",[2162,3039,3040,3043],{},[2405,3041,3042],{},"第三,品牌词的可见度。"," 直接搜你产品名,ChatGPT 应该能准确说出你是做什么的。如果搜出来是竞品或者信息过时,说明你的品牌信源建设不够。",[2280,3045,2453],{"id":2453},[2162,3047,3048],{},"看好 GEO,但反对把 GEO 神化。ChatGPT Search 全量开放是个节点事件,不代表传统搜索要死了——Google 的流量盘子短期内没人能动。真正的变化是用户行为在分流:深度调研走 AI,快速查询走搜索。",[2162,3050,3051],{},"软件公司现在的合理动作是:SEO 继续做,把 GEO 加进日常。具体说就是文档站点要 AI 友好、对比内容要在外部铺、品牌词要在主流 AI 引擎做可见度监测。我们自己的做法是每周用脚本在 6 个 AI 引擎跑一批关键词,记录引用情况。这套监测没有现成工具,得自己搭。",[2162,3053,3054],{},"如果你还没开始监测你的产品词在 ChatGPT 里被怎么引用,建议现在就开始。免费版开放了,门槛比半年前低多了。",{"title":2253,"searchDepth":2254,"depth":2254,"links":3056},[3057,3058,3059,3060,3061],{"id":2919,"depth":2254,"text":2919},{"id":2928,"depth":2254,"text":2928},{"id":2994,"depth":2254,"text":2995},{"id":3018,"depth":2254,"text":3019},{"id":2453,"depth":2254,"text":2453},"ChatGPT Search 向所有用户开放后,开发者用 AI 搜索技术问题的比例明显上升。软件公司的 GEO 优先级要从\"可选项\"变成\"必选项\"。",{},"\u002Fnews\u002Fchatgpt-search-launch",{"title":2911,"description":3062},"OpenAI","news\u002Fchatgpt-search-launch","yq_EL-A531QdK3punN_k_NaWDSWGKdPtK9ipkRSztBc",[3070],{"id":3071,"title":3072,"body":3073,"canonical":2143,"client":3228,"cover":2143,"date":448,"description":3229,"draft":2261,"extension":2262,"industry":3230,"keywords":2143,"meta":3231,"navigation":2265,"noindex":2261,"ogImage":2143,"path":3232,"seo":3233,"seoDescription":2143,"seoTitle":2143,"service":3234,"stem":3235,"__hash__":3236},"cases\u002Fcases\u002Fyuanxi-self-case.md","元析智能官网:我们用自己的方法给自己做 SEO",{"type":2156,"value":3074,"toc":3220},[3075,3078,3081,3084,3087,3090,3094,3097,3136,3139,3142,3145,3148,3152,3155,3175,3178,3181,3184,3188,3193,3196,3201,3204,3209,3212,3217],[2162,3076,3077],{},"我们目前没有客户案例。公司 2026 年 8 月刚注册,还没签过一单。但我们需要一个案例证明方法有效——所以拿自己的官网当第一个案例。",[2162,3079,3080],{},"这不是嘴上说说。这个站从技术选型到内容写作,全部按我们给客户提案时的方法做。方法有效,几个月后排名会起来;没起来,我们改方法,不嘴硬。这篇文章把我们做了什么、为什么这么做,完整写下来。",[2280,3082,3083],{"id":3083},"技术选型",[2162,3085,3086],{},"用 Nuxt 3,基于 Vue 3。原因只有一个:SEO 需要服务端渲染或静态生成,纯 SPA 的页面搜索引擎抓不到内容。Nuxt 3 的 SSR\u002FSSG 在服务端把页面渲染成完整 HTML,爬虫直接拿到带内容的 DOM,不用等 JS 执行。",[2162,3088,3089],{},"样式用 UnoCSS,原子化、按需生成,产物体积小。内容用 @nuxt\u002Fcontent 模块,Markdown 写文章直接生成页面和路由,不搭 CMS。",[2280,3091,3093],{"id":3092},"技术-seo-做了什么","技术 SEO 做了什么",[2162,3095,3096],{},"具体做了这几件事:",[2177,3098,3099,3109,3115,3124,3130],{},[2180,3100,3101,3104,3105,3108],{},[2405,3102,3103],{},"sitemap.xml",":@nuxtjs\u002Fsitemap 自动从路由生成,排除 \u002Fadmin,加 autoLastmod。博客、资讯、案例、厂商分类页的 URL 通过 \u002Fapi\u002F",[2405,3106,3107],{},"sitemap","\u002Furls 动态注入。",[2180,3110,3111,3114],{},[2405,3112,3113],{},"robots.txt",":允许抓取前台,禁止 \u002Fadmin。明确允许 AI 引擎抓取,这是 GEO 的前提。",[2180,3116,3117,3120,3123],{},[2405,3118,3119],{},"结构化数据",[3121,3122],"nuxt-schema-org",{}," 注入 Organization(公司名、联系方式、服务区域)、BreadcrumbList(面包屑)、FAQPage(FAQ 段落)、Article(博客文章)。让搜索引擎和 AI 引擎能结构化读取。",[2180,3125,3126,3129],{},[2405,3127,3128],{},"canonical",":每个页面有 canonical 标签,避免重复 URL 被当抄袭。",[2180,3131,3132,3135],{},[2405,3133,3134],{},"语义化 HTML",":用 header\u002Fnav\u002Fmain\u002Farticle\u002Fsection,不用全 div 套 div。h1 每页一个,层级清晰。",[2280,3137,3138],{"id":3138},"内容策略",[2162,3140,3141],{},"内容准则写在项目里,明确禁用 AI 味词汇(赋能\u002F打造\u002F助力\u002F一站式\u002F全方位\u002F卓越\u002F领先\u002F极致\u002F深耕\u002F驱动\u002F生态\u002F闭环……)。写作要求:说人话、有数字、有操作、有立场、有节奏。AI 初稿直接发等于对公司的背叛,必须人工去 AI 味。",[2162,3143,3144],{},"厂商分类拆成 10 类:SaaS、开发者工具、跨境电商软件、工程设计软件、设计工具、数据与 AI 工具、安全运维、浏览器扩展、协作办公、教育学习。每个分类对应一组长尾词,用分类页打长尾流量,不硬磕\"SEO 服务\"这种红海词。",[2162,3146,3147],{},"博客写实操:SEO 报告怎么看、GEO 和 SEO 区别、SaaS 怎么做 SEO、跨境软件出海怎么做。不写\"行业洞察\"那种空话。",[2280,3149,3151],{"id":3150},"geo-做了什么","GEO 做了什么",[2162,3153,3154],{},"GEO 是让 AI 引擎(ChatGPT、Perplexity、Kimi、文心)能引用你。做了三件事:",[2177,3156,3157,3163,3169],{},[2180,3158,3159,3162],{},[2405,3160,3161],{},"FAQ 结构",":每篇内容尽量带 FAQ 段落,用 FAQPage 结构化数据标记,方便 AI 提取。",[2180,3164,3165,3168],{},[2405,3166,3167],{},"结论前置",":文章开头直接给结论,不绕三段再点题。AI 抓摘要时,前 200 字决定能不能被引用。",[2180,3170,3171,3174],{},[2405,3172,3173],{},"可引用事实句",":写\"百度 2025 年对 AI 生成内容识别更严\"\"Perplexity 上线购物功能\"这种有具体时间、具体主体的句子,而不是\"AI 正在改变搜索\"这种空话。",[2280,3176,3177],{"id":3177},"待验证",[2162,3179,3180],{},"站刚上线,没有排名数据,没有流量数据,没有收录数据。我们不编。",[2162,3182,3183],{},"验证计划:3 个月后看百度搜索资源平台和 Google Search Console 的收录数;6 个月后看非 brand 关键词排名有没有进 top 20。6 个月没动静,说明方法有问题,我们改。",[2280,3185,3187],{"id":3186},"faq","FAQ",[2162,3189,3190],{},[2405,3191,3192],{},"为什么没有客户案例?",[2162,3194,3195],{},"公司刚注册,还没签过客户。我们不拿同行案例改个名字当自己的,那是骗人。所以第一个案例只能是自己。",[2162,3197,3198],{},[2405,3199,3200],{},"怎么验证你们的方法有效?",[2162,3202,3203],{},"看我们自己官网的排名。\"软件公司 SEO\"\"GEO 优化\"这种非 brand 词能进百度或 Google 前几页,说明方法有效。进不去,我们改方法,不找借口。",[2162,3205,3206],{},[2405,3207,3208],{},"这个站用了什么技术栈?",[2162,3210,3211],{},"Nuxt 3 + @nuxt\u002Fcontent + @nuxtjs\u002Fseo + UnoCSS。没用 WordPress,没用纯 SPA。选型理由就一个:服务端渲染对 SEO 友好。",[2162,3213,3214],{},[2405,3215,3216],{},"自己给自己做案例,怎么保证客观?",[2162,3218,3219],{},"没法完全客观。我们能做的是:数据全公开(收录、排名、流量都摆出来),失败也写出来,不挑好看的数据。客户案例我们也会按这个标准做。",{"title":2253,"searchDepth":2254,"depth":2254,"links":3221},[3222,3223,3224,3225,3226,3227],{"id":3083,"depth":2254,"text":3083},{"id":3092,"depth":2254,"text":3093},{"id":3138,"depth":2254,"text":3138},{"id":3150,"depth":2254,"text":3151},{"id":3177,"depth":2254,"text":3177},{"id":3186,"depth":2254,"text":3187},"元析智能(自身)","这是我们的第一个案例——我们自己的官网。从技术选型到内容策略,把我们做 SEO 和 GEO 的方法完整记录下来。没有客户案例之前,官网就是案例。","SEO\u002FGEO 服务",{},"\u002Fcases\u002Fyuanxi-self-case",{"title":3072,"description":3229},"SEO + GEO","cases\u002Fyuanxi-self-case","CWBGqxarAbzeszFA87x3T0SnZijOWsxV1SXb82MMmQc",1794088098157]