[{"data":1,"prerenderedAt":301},["ShallowReactive",2],{"blog-ahrefs-backlink-filter":3,"blog-prev-ahrefs-backlink-filter":79,"blog-next-ahrefs-backlink-filter":184},{"id":4,"title":5,"author":6,"body":7,"canonical":63,"category":64,"cover":63,"date":65,"description":66,"draft":67,"extension":68,"keywords":63,"meta":69,"navigation":70,"noindex":67,"ogImage":63,"path":71,"seo":72,"seoDescription":63,"seoTitle":63,"stem":73,"tags":74,"updated":63,"__hash__":78},"blog\u002Fblog\u002Fahrefs-backlink-filter.md","从竞品外链里挑可复制的，我们只留三类","元析智能",{"type":8,"value":9,"toc":55},"minimark",[10,14,19,22,25,28,31,34,37,40,43,46,49,52],[11,12,13],"p",{},"2025 年 10 月我们把 6 个竞品站的外链全部导进 Ahrefs，一共 3.2 万条，按页面类型分桶再筛，最后值得动手联系的只有 340 条。外链这事做错方向，比不做的伤害大，因为低质链接现在会被识别成操纵行为。",[15,16,18],"h2",{"id":17},"先按来源页面分桶别看域名","先按来源页面分桶，别看域名",[11,20,21],{},"新手常按域名 DR 排序，挑高 DR 的站去联系，结果一堆新闻门户和目录站。我按链接所在的页面类型分：工具清单页、评测对比页、行业目录、博客正文、论坛签名、付费新闻稿。前三种有争取空间，后三种基本是浪费时间。",[11,23,24],{},"分桶之后再看页面本身：那篇工具清单还在不在更新，文章有没有被删，链接是正文里还是页脚。页脚链接和正文链接的价值不一样，这个判断工具不会替你做完。",[15,26,27],{"id":27},"三类值得花时间",[11,29,30],{},"第一类，工具清单和资源页。GitHub 上的 awesome 列表、开发者工具合集、开源项目推荐页，这类页面接受提交，通常走一个 pull request 或者邮件。它们的价值在于稳定和相关性，我的经验是这类链接带来的排名帮助比新闻稿大得多。",[11,32,33],{},"第二类，竞品评测和对比文章。写这类文章的作者需要把你的产品补进去，前提是你提供清楚的信息：定价、免费额度、和竞品的功能差异、一句话定位。发一封只有链接的邮件过去，没人会理。",[11,35,36],{},"第三类，行业协会和开源社区目录。这类要一次性提交，审核慢，但一旦上去几年不掉。",[15,38,39],{"id":39},"主动放弃的四类",[11,41,42],{},"批量目录站提交，一次几千个，三个月掉一半，还留下操纵痕迹。论坛签名和评论区，nofollow 居多，现在也不缺这点权重。付费链接包和新闻稿分发，Google 从 2024 年起对买卖链接的处理更坚决，一旦被判定，前面的工作全白费。最后是和你业务八竿子打不着的综合门户，相关性为零的链接，多一条都是负担。",[15,44,45],{"id":45},"我们的执行流程",[11,47,48],{},"导出竞品外链，按来源 URL 去重，去掉已经联系过的，剩下按相关度和页面活跃度打分。打分只看三项：页面主题是否和软件业务相关、页面最近 12 个月有没有更新、链接是编辑推荐还是自动收录。",[11,50,51],{},"联系模板我写得很短，三段：我是谁、你的页面里哪一段适合补我们的产品、我可以提供什么素材。发送记录进表格，3 周后没回复就放弃，不追第二封。",[11,53,54],{},"外链这件事最怕的是失控。我们每季度回查一次已获得的外链，掉链的记下来，掉得多说明对方在清理，以后就别再花力气。",{"title":56,"searchDepth":57,"depth":57,"links":58},"",2,[59,60,61,62],{"id":17,"depth":57,"text":18},{"id":27,"depth":57,"text":27},{"id":39,"depth":57,"text":39},{"id":45,"depth":57,"text":45},null,"工具评测","2025-10-30","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。",false,"md",{},true,"\u002Fblog\u002Fahrefs-backlink-filter",{"title":5,"description":66},"blog\u002Fahrefs-backlink-filter",[75,76,77],"外链建设","Ahrefs","竞品分析","2O-L1sZbD7YKu88IgC5VkaaP7cn0CrUQ_JbXA0mTQoA",{"id":80,"title":81,"author":6,"body":82,"canonical":63,"category":172,"cover":63,"date":173,"description":174,"draft":67,"extension":68,"keywords":63,"meta":175,"navigation":70,"noindex":67,"ogImage":63,"path":176,"seo":177,"seoDescription":63,"seoTitle":63,"stem":178,"tags":179,"updated":63,"__hash__":183},"blog\u002Fblog\u002Fwechat-search-ranking-logic.md","微信搜一搜的排名逻辑，和百度根本不是一回事",{"type":8,"value":83,"toc":164},[84,87,90,93,96,99,102,121,124,127,130,133,136,139,142,145,149,152,155,158,161],[11,85,86],{},"微信搜一搜的排名，账号权重占一半，内容传播数据占三成，关键词匹配只占两成。这跟百度那套外链加页面权重的逻辑没有可比性。我们运营 3 个公众号、测了 8 个月，把因素按重要性排了个序。",[15,88,89],{"id":89},"我们怎么测的",[11,91,92],{},"2025 年 3 月起，我们给一个 SaaS 客户运营 3 个公众号，累计发 96 篇原创。3 个号注册时长、粉丝量、原创比例都不一样，内容用同一个模板写。",[11,94,95],{},"同样关键词\"项目管理工具推荐\"：A 号（注册 6 年、粉丝 2.1 万、原创 180 篇）排第 4；B 号（注册 1 年、粉丝 800）排到 30 名开外。两篇文章结构几乎一样，差异只在账号。",[15,97,98],{"id":98},"影响排名的五个因素",[11,100,101],{},"按重要性排：",[103,104,105,109,112,115,118],"ol",{},[106,107,108],"li",{},"账号权重：注册时长、原创比例、违规记录，这三项决定基础分",[106,110,111],{},"传播数据：阅读完成率、在看、转发、收藏，发布后 48 小时是关键窗口",[106,113,114],{},"关键词匹配：标题和前 200 字，只占两成",[106,116,117],{},"时效：超过 1 年的文章排名下滑明显，老内容要更新",[106,119,120],{},"互动：评论区活跃、作者回复，有加分但不决定性",[15,122,123],{"id":123},"我们不做的三件事",[11,125,126],{},"不在文章里堆关键词。搜一搜的匹配比百度克制，堆了反而影响阅读完成率，而阅读完成率是传播数据里权重最高的。",[11,128,129],{},"不买阅读量。微信对异常传播的识别很严，我们见过一个号刷了两个月，被降权后再没起来。",[11,131,132],{},"不指望一篇爆款吃长期排名。搜一搜的排名对时效敏感，一篇 10 万+ 的文章，半年后基本掉出前 20。",[15,134,135],{"id":135},"软件公司该怎么做搜一搜",[11,137,138],{},"目标定成\"搜你名字能出来\"，不是\"搜品类词排第一\"。",[11,140,141],{},"具体三件事：账号名和产品名一致；每篇文章标题带一个用户会搜的问题（比如\"API 限流怎么排查\"）；文章里出现产品名时链到小程序或官网。这三件事做好，品牌词和长尾问题词基本能覆盖。",[11,143,144],{},"品类大词让给百度。搜一搜的用户在微信里，是要解决问题，不是逛着找工具。",[15,146,148],{"id":147},"搜一搜和百度-seo-的关系","搜一搜和百度 SEO 的关系",[11,150,151],{},"两套体系别混着投预算。百度那套是“页面进来、排名上去、点击进来”，搜一搜是“账号有分量、内容被传播、搜索能搜到”。",[11,153,154],{},"我们给客户做分工：官网和文档站打百度、Google；公众号打搜一搜和私域。同一个案例，两边各写一版，素材复用但结构完全不同。",[15,156,157],{"id":157},"一个常被问的问题",[11,159,160],{},"有人问：公众号文章能不能做外链？不能，搜一搜不认外链。它的权重体系里根本没有这一项，所以别指望靠外部推广把文章顶上去——唯一的杠杆是内容和账号本身。",[11,162,163],{},"我们现在的做法是每个号每周 2 篇，宁少不滥。质量不够的文章，发了反而拉低账号权重。",{"title":56,"searchDepth":57,"depth":57,"links":165},[166,167,168,169,170,171],{"id":89,"depth":57,"text":89},{"id":98,"depth":57,"text":98},{"id":123,"depth":57,"text":123},{"id":135,"depth":57,"text":135},{"id":147,"depth":57,"text":148},{"id":157,"depth":57,"text":157},"行业观察","2025-10-29","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。",{},"\u002Fblog\u002Fwechat-search-ranking-logic",{"title":81,"description":174},"blog\u002Fwechat-search-ranking-logic",[180,181,182],"微信搜一搜","公众号运营","排名因素","sxQsYt-IC7VRxEQaxJi0hL65r6tXM5bZ2db0idItt1A",{"id":185,"title":186,"author":6,"body":187,"canonical":63,"category":290,"cover":63,"date":291,"description":292,"draft":67,"extension":68,"keywords":63,"meta":293,"navigation":70,"noindex":67,"ogImage":63,"path":294,"seo":295,"seoDescription":63,"seoTitle":63,"stem":296,"tags":297,"updated":63,"__hash__":300},"blog\u002Fblog\u002Furl-naming-rules-software.md","软件公司 URL 怎么起名，我们的 6 条规矩",{"type":8,"value":188,"toc":280},[189,192,196,207,211,228,232,238,242,248,252,255,259,262,265,268,271,277],[11,190,191],{},"URL 不是技术细节，是内容资产。三年后它还要出现在你的 API 文档、客户邮件、微信群里。改一次的成本远高于起名时多想十分钟。下面 6 条是我们给软件公司定 URL 规矩时最常用的。",[15,193,195],{"id":194},"_1-全小写短横线分隔","1. 全小写，短横线分隔",[11,197,198,202,203,206],{},[199,200,201],"code",{},"\u002Fdocs\u002Fapi-rate-limit"," 对，",[199,204,205],{},"\u002FDocs\u002FAPI_RateLimit"," 错。大小写混写和下滑线不会让页面打不开，但会让同一个页面出现两个地址，多一层重复内容要处理。",[15,208,210],{"id":209},"_2-用英文关键词不要拼音不要数字-id","2. 用英文关键词，不要拼音、不要数字 ID",[11,212,213,202,216,219,220,223,224,227],{},[199,214,215],{},"\u002Fproducts\u002Fdata-sync",[199,217,218],{},"\u002Fproduct\u002F12345","、",[199,221,222],{},"\u002Fchanpin\u002Fshujutongbu"," 错。数字 ID 对搜索引擎和用户都没有信息量；拼音对英文搜索没有帮助，对海外用户不可读。软件公司的产品名如果是自造词，就把功能词放进去，比如 ",[199,225,226],{},"\u002Fproducts\u002Fedge-cache","。",[15,229,231],{"id":230},"_3-层级不超过四层","3. 层级不超过四层",[11,233,234,237],{},[199,235,236],{},"\u002Fdocs\u002Fapi\u002Frate-limit"," 够了。再往下一层，点击深度和抓取效率都会变差。语言和版本不要塞进路径主层级，宁可放子域或参数。",[15,239,241],{"id":240},"_4-版本号别进主路径","4. 版本号别进主路径",[11,243,244,247],{},[199,245,246],{},"\u002Fdocs\u002Fv2\u002Frate-limit"," 会让每个版本变成一批独立页面，升级大版本时整批 301。2025 年 3 月一个 API 工具站把 v2 挪到主路径，1800 个文档页全换了 URL，三个月排名都没回到原来的位置。建议路径保持稳定，版本用页面内切换，旧版单独放一个归档区。",[15,249,251],{"id":250},"_5-参数只留功能性的","5. 参数只留功能性的",[11,253,254],{},"筛选、排序、分页参数可以存在，但要配 canonical 或 robots 规则处理。追踪参数（utm 那一堆）不要出现在可索引的链接里，服务端能拦就拦。参数越多，重复内容治理越麻烦。",[15,256,258],{"id":257},"_6-改了就必须-301","6. 改了就必须 301",[11,260,261],{},"URL 一改，旧地址 301 到新地址，sitemap 同步更新，站内链接一次性替换干净。别留旧链接挂着 404，也别让两个地址同时可访问。",[15,263,264],{"id":264},"改名的操作顺序",[11,266,267],{},"已经上线的 URL 要改，顺序别乱：先盘点旧 URL 和流量（GSC 里按点击排序，取前 200 个），再定新 URL 规则，然后写 301 映射表并上线，同时改站内链接和 sitemap，最后在 GSC 提交改版。这个客户改 800 个 URL 用了三天，映射表核对占了两天，上线本身半小时。上线后前两周流量会掉 10% 到 20%，属于正常波动，别急着回滚。",[15,269,270],{"id":270},"一个反例",[11,272,273,276],{},[199,274,275],{},"\u002Fblog\u002F2025\u002F03\u002F12\u002Fseo-tips"," 这种带日期的路径，除非你真的按时间归档，否则别用。用户和蜘蛛都会觉得旧文章过期，日历层级又多两层。我们给客户改版时，带日期的博客路径是最常被顺手收拾的一类。",[11,278,279],{},"起名的时候把 URL 念一遍，看能不能在电话里说清楚。念不清的，重起。",{"title":56,"searchDepth":57,"depth":57,"links":281},[282,283,284,285,286,287,288,289],{"id":194,"depth":57,"text":195},{"id":209,"depth":57,"text":210},{"id":230,"depth":57,"text":231},{"id":240,"depth":57,"text":241},{"id":250,"depth":57,"text":251},{"id":257,"depth":57,"text":258},{"id":264,"depth":57,"text":264},{"id":270,"depth":57,"text":270},"技术SEO","2025-10-31","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。",{},"\u002Fblog\u002Furl-naming-rules-software",{"title":186,"description":292},"blog\u002Furl-naming-rules-software",[298,299,290],"URL设计","站点结构","gKD3HPdkPazEvttQVeqFrANB7XHDD0cFy85m6dJajBA",1794088111230]