[{"data":1,"prerenderedAt":589},["ShallowReactive",2],{"blog-screaming-frog-large-crawl-config":3,"blog-prev-screaming-frog-large-crawl-config":103,"blog-next-screaming-frog-large-crawl-config":177},{"id":4,"title":5,"author":6,"body":7,"canonical":87,"category":88,"cover":87,"date":89,"description":90,"draft":91,"extension":92,"keywords":87,"meta":93,"navigation":94,"noindex":91,"ogImage":87,"path":95,"seo":96,"seoDescription":87,"seoTitle":87,"stem":97,"tags":98,"updated":87,"__hash__":102},"blog\u002Fblog\u002Fscreaming-frog-large-crawl-config.md","爬十万个页面，Screaming Frog 配置先改五处","元析智能",{"type":8,"value":9,"toc":77},"minimark",[10,14,19,22,25,29,32,35,39,42,45,49,52,55,59,62,65,68,71,74],[11,12,13],"p",{},"2025 年 12 月我们爬一个文档站，12 万个 URL，第一次跑完导出结果一看，8 万条是带参数的重复页，真正要看的没几页。不是工具不行，是配置默认值适合小站，大站照默认跑就是浪费时间。",[15,16,18],"h2",{"id":17},"第一处渲染方式","第一处：渲染方式",[11,20,21],{},"默认不执行 JavaScript，抓的是源码。内容靠前端渲染的站，这么抓回来一片空白。开渲染之前先确认必要性，开了之后速度会慢好几倍，十几万页能跑到第二天。",[11,23,24],{},"我的做法是先关掉渲染快速跑一遍，看 HTML 里有没有正文。有就不开，没有才开，并且把渲染超时从默认的几秒调到 10 秒左右，否则异步加载的内容抓不全。",[15,26,28],{"id":27},"第二处抓取范围","第二处：抓取范围",[11,30,31],{},"用包含和排除规则把范围框住，比爬完再筛有效。要排除的东西很固定：会话 ID、排序参数、日历归档、筛选参数、文件下载、后台路径。",[11,33,34],{},"我们那 12 万 URL 里，光带排序参数的就有 5 万多。加一条正则排除之后，实际需要爬的只剩 3 万出头。",[15,36,38],{"id":37},"第三处限速和并发","第三处：限速和并发",[11,40,41],{},"默认并发对生产站有压力，尤其是共享服务器的小站。爬之前先跟运维确认能不能爬、什么时段爬。我们通常把并发降到 2，加上延迟，放在夜里跑。",[11,43,44],{},"曾经有一次没降速，把一个客户的电商站爬得响应变慢，第二天被运维投诉。从那以后，我们的规矩是先小样本跑 200 页，看服务器负载再放量。",[15,46,48],{"id":47},"第四处robots-与-ua","第四处：robots 与 UA",[11,50,51],{},"要不要遵守 robots，取决于目的。做诊断时遵守，能真实反映搜索引擎看到的东西；做覆盖度检查时可以临时不遵守，但要标记出来，别把结果混着看。",[11,53,54],{},"UA 也建议改一下，带上自己的标识，万一爬出问题对方能找到人。",[15,56,58],{"id":57},"第五处导出和存储","第五处：导出和存储",[11,60,61],{},"十万页以上，内存吃紧，结果文件也大。分批爬，按目录拆开，每批单独导出。导出时只留要看的列：状态码、标题、字数、内链数、canonical、索引设置。",[15,63,64],{"id":64},"爬完先看什么",[11,66,67],{},"结果出来别急着导全表，先看四组数：状态码分布、重复标题数、canonical 指向不一致的页数、内链深度大于 4 的页数。",[11,69,70],{},"第一次爬那个文档站，状态码分布里 200 占 78%，剩下的 22% 全是 301。301 集中在两个旧目录，说明改版时跳转没做干净。这种结论从全表里翻是翻不出来的。",[11,72,73],{},"还有个习惯，每次爬完把结果存一份带日期的 CSV，下次对比用。三个月后能看出抓取问题是在变好还是变坏，这比单次报告有用。",[11,75,76],{},"我现在的习惯是爬之前先写一句话：这次要回答什么问题。写不出来就别爬。工具跑得快不代表有意义，配置和目的不清，爬一百万页也只是生成一个巨大的 CSV。",{"title":78,"searchDepth":79,"depth":79,"links":80},"",2,[81,82,83,84,85,86],{"id":17,"depth":79,"text":18},{"id":27,"depth":79,"text":28},{"id":37,"depth":79,"text":38},{"id":47,"depth":79,"text":48},{"id":57,"depth":79,"text":58},{"id":64,"depth":79,"text":64},null,"技术SEO","2025-12-21","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。",false,"md",{},true,"\u002Fblog\u002Fscreaming-frog-large-crawl-config",{"title":5,"description":90},"blog\u002Fscreaming-frog-large-crawl-config",[99,100,101],"Screaming Frog","爬虫配置","技术检测","DCwAhp0UzCMLPG83vMS3jN3KCdRE8kuHWlMloMd0GYM",{"id":104,"title":105,"author":6,"body":106,"canonical":87,"category":165,"cover":87,"date":166,"description":167,"draft":91,"extension":92,"keywords":87,"meta":168,"navigation":94,"noindex":91,"ogImage":87,"path":169,"seo":170,"seoDescription":87,"seoTitle":87,"stem":171,"tags":172,"updated":87,"__hash__":176},"blog\u002Fblog\u002Fbaidu-ranking-weight-changes.md","百度排名机制里，我们测出的三个权重变化",{"type":8,"value":107,"toc":159},[108,111,115,118,121,124,128,131,134,137,141,144,147,150,153,156],[11,109,110],{},"2025 年我们盯了 28 个软件公司的站点，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。不是猜的，是同一个词逐月追出来的。",[15,112,114],{"id":113},"变化一整站评价压过单页优化","变化一：整站评价压过单页优化",[11,116,117],{},"以前一个页面优化到位，能单独把排名顶上去。2025 年下半年开始，这件事越来越难。",[11,119,120],{},"我们有个客户做了个对比：同一个产品词，A 页面优化了标题、正文、内链，但整站有 30% 的页面是空壳；B 页面只做了基础优化，整站结构干净。三个月后 B 排到第 6，A 在第 14 徘徊。",[11,122,123],{},"判断依据：百度现在给的更像是\"站点级\"的评价，单页只占一部分。所以我们给客户做诊断，先看整站有多少页面是\"删掉也没损失\"的，比例超过 30% 就先做减法。",[15,125,127],{"id":126},"变化二外链权重继续让位给品牌搜索","变化二：外链权重继续让位给品牌搜索",[11,129,130],{},"这件事我们从 2024 年就在观察，2025 年更明显。",[11,132,133],{},"监测数据：12 个站里，品牌词月搜索量超过 500 的 5 个站，核心词平均排名第 7；品牌词低于 100 的 7 个站，核心词平均排名第 19。两组站的内容质量、外链数量差不多。这个对比我们去年 11 月又做了一遍，结论没变。",[11,135,136],{},"差异在品牌。用户主动搜你的名字，百度会把这个信号当成\"这个站有人认\"，反过来给核心词加权。买外链做不出这个信号。",[15,138,140],{"id":139},"变化三老内容不更新就往下掉","变化三：老内容不更新就往下掉",[11,142,143],{},"2025 年 9 月我们抽查了 6 个站的 200 个老页面，其中 74 个页面在过去一年没有任何更新，这 74 个里有 51 个排名掉了 5 名以上。",[11,145,146],{},"不是所有内容都要更新。我们的做法是分两类：产品功能页每季度对一遍，价格、版本号、截图必须是最新的；教程类文章每年补一次案例，标题加年份。",[11,148,149],{},"一个做 API 文档的客户，把 30 篇核心教程加了\"2025 年更新\"的版本说明和一段新案例，两个月后这批页面平均排名从 11 升到 8。",[15,151,152],{"id":152},"我们怎么应对",[11,154,155],{},"三件事：砍掉没价值的页面、把品牌词做起来、给老内容定更新节奏。第一件见效最快，通常一个月内能看到收录和抓取的变化。",[11,157,158],{},"品牌这件事最慢，也最难外包。我们的建议是从行业媒体曝光和社区内容开始，先让一部分人搜你的名字，再谈排名。这三个变化，我们每季度会重新验证一次，有变化再更新。顺序别反：先有内容和曝光，再有品牌搜索，最后才是排名加权。",{"title":78,"searchDepth":79,"depth":79,"links":160},[161,162,163,164],{"id":113,"depth":79,"text":114},{"id":126,"depth":79,"text":127},{"id":139,"depth":79,"text":140},{"id":152,"depth":79,"text":152},"行业观察","2025-12-20","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。",{},"\u002Fblog\u002Fbaidu-ranking-weight-changes",{"title":105,"description":167},"blog\u002Fbaidu-ranking-weight-changes",[173,174,175],"百度排名","权重变化","监测数据","cUAjsZiBuKrBxPK2fjLE8cSZqCjp81APwyso2dMspT4",{"id":178,"title":179,"author":6,"body":180,"canonical":87,"category":576,"cover":87,"date":577,"description":578,"draft":91,"extension":92,"keywords":87,"meta":579,"navigation":94,"noindex":91,"ogImage":87,"path":580,"seo":581,"seoDescription":87,"seoTitle":87,"stem":582,"tags":583,"updated":87,"__hash__":588},"blog\u002Fblog\u002Fcross-border-software-seo.md","跨境电商软件的 SEO:出海和国内双市场怎么打",{"type":8,"value":181,"toc":566},[182,185,189,192,195,198,201,233,236,239,245,256,261,275,278,282,285,363,366,392,395,398,404,410,413,431,434,437,440,445,459,464,481,487,490,493,496,502,508,514,520,523,526,557,560,563],[11,183,184],{},"跨境电商软件(ERP、listing 工具、广告管理工具)的客户群体特殊:既要做国内市场(服务跨境卖家),又要做海外市场(直接卖海外卖家)。SEO 要双线作战,语言、搜索引擎、外链策略都不一样。这篇文章写我们做跨境软件 SEO 的经验,包括踩过的坑。",[15,186,188],{"id":187},"双语-seo-的难点","双语 SEO 的难点",[11,190,191],{},"先说最难的部分:不是翻译,是关键词对不上。",[11,193,194],{},"中文\"亚马逊 listing 优化工具\",英文对应词可能是\"Amazon listing optimization tool\",也可能是\"Amazon SEO tool\",也可能是\"listing optimizer\"。这三个词在 Google 上的搜索量差 3 倍,意图也有差别。",[11,196,197],{},"我们做过的踩坑:给一个 listing 工具客户做英文站,直接翻译中文关键词,3 个月流量起不来。后来用 Ahrefs 重新挖词,发现真正的高搜索量词是\"Amazon product description generator\"和\"listing builder\",我们一开始没用。",[11,199,200],{},"双语 SEO 的关键:",[202,203,204,212,227],"ul",{},[205,206,207,211],"li",{},[208,209,210],"strong",{},"不要翻译关键词,要重新调研",":中英文用户搜索习惯完全不同",[205,213,214,217,218,222,223,226],{},[208,215,216],{},"不同语言独立站",":不要一个站多语言切换,子目录(",[219,220,221],"code",{},"\u002Fen\u002F","、",[219,224,225],{},"\u002Fzh\u002F",")或子域名(ccTLD)都行",[205,228,229,232],{},[208,230,231],{},"hreflang 标签必须正确",":错的 hreflang 比 没有 hreflang 还糟,会让 Google 把英文页当中文页索引",[15,234,235],{"id":235},"关键词调研工具的差异",[11,237,238],{},"国内市场和海外市场的关键词工具完全是两套:",[11,240,241,244],{},[208,242,243],{},"国内市场",":",[202,246,247,250,253],{},[205,248,249],{},"5118:百度关键词,中文最准",[205,251,252],{},"百度关键词规划师:官方数据,但搜索量范围",[205,254,255],{},"站长工具:查中文词还行",[11,257,258,244],{},[208,259,260],{},"海外市场",[202,262,263,266,269,272],{},[205,264,265],{},"Ahrefs:英文词最准,数据全",[205,267,268],{},"SEMrush:欧洲小语种覆盖好",[205,270,271],{},"Google Keyword Planner:免费但数据粗",[205,273,274],{},"Ubersuggest:入门级,数据不够细",[11,276,277],{},"我们双线做时,英文用 Ahrefs,中文用 5118,两边数据对不上是常态。比如同功能的工具,中文搜索量 300\u002F月,英文可能 3000\u002F月。预算分配要跟着搜索量走,不能平均分。",[15,279,281],{"id":280},"google-和百度算法差异对内容策略的影响","Google 和百度算法差异对内容策略的影响",[11,283,284],{},"Google 和百度的算法差异,直接影响内容怎么写:",[286,287,288,304],"table",{},[289,290,291],"thead",{},[292,293,294,298,301],"tr",{},[295,296,297],"th",{},"维度",[295,299,300],{},"Google",[295,302,303],{},"百度",[305,306,307,319,330,341,352],"tbody",{},[292,308,309,313,316],{},[310,311,312],"td",{},"外链权重",[310,314,315],{},"高",[310,317,318],{},"2025 年后明显降低",[292,320,321,324,327],{},[310,322,323],{},"内容长度偏好",[310,325,326],{},"长内容(2000+ 字)排名好",[310,328,329],{},"中等长度(800-1500 字)够用",[292,331,332,335,338],{},[310,333,334],{},"AI 内容态度",[310,336,337],{},"看质量,不一刀切",[310,339,340],{},"严格识别,批量降权",[292,342,343,346,349],{},[310,344,345],{},"新站沙盒期",[310,347,348],{},"6-9 个月",[310,350,351],{},"3-6 个月",[292,353,354,357,360],{},[310,355,356],{},"E-E-A-T",[310,358,359],{},"权威度信号重要",[310,361,362],{},"看重备案、ICP",[11,364,365],{},"实操建议:",[202,367,368,374,380,386],{},[205,369,370,373],{},[208,371,372],{},"英文站写长内容",":2000-3000 字,深入展开,Google 偏好",[205,375,376,379],{},[208,377,378],{},"中文站写短内容",":800-1500 字,直击重点,百度偏好",[205,381,382,385],{},[208,383,384],{},"同一主题写两版",":不要翻译,要按各自的算法偏好重写",[205,387,388,391],{},[208,389,390],{},"作者署名和权威度",":英文站重视 E-E-A-T,作者要有 LinkedIn、GitHub 等可信背书;中文站备案和 ICP 是基础",[15,393,394],{"id":394},"竞品分析怎么做",[11,396,397],{},"跨境软件的竞品分析比纯国内或纯海外复杂,要分两层:",[11,399,400,403],{},[208,401,402],{},"国内市场竞品",":用 5118 查竞品的关键词、流量、外链。看竞品在百度上的排名词,找空位。",[11,405,406,409],{},[208,407,408],{},"海外市场竞品",":用 Ahrefs 或 SEMrush 查竞品的 Google 排名、外链来源、内容策略。",[11,411,412],{},"我们的做法:",[414,415,416,419,422,425,428],"ol",{},[205,417,418],{},"列出国内外各 5-10 个直接竞品",[205,420,421],{},"拉竞品的 organic keywords(自然排名词)",[205,423,424],{},"找出竞品有排名你却没有的词(content gap)",[205,426,427],{},"分析竞品内容类型,看哪些类型他们做你没做",[205,429,430],{},"分析竞品外链来源,看能不能找同类渠道(不是抄,是找同类)",[11,432,433],{},"跨境软件的竞品分析有个特殊性:国内竞品和海外竞品经常不是同一批。国内做得好的(比如店小秘),海外不一定强;海外强的(Helium 10),国内不一定有。这意味着双线都有空位,关键是判断哪边机会大。",[15,435,436],{"id":436},"外链建设在出海场景的特殊性",[11,438,439],{},"跨境软件的外链建设是国内和海外两套玩法,难度差很多:",[11,441,442,244],{},[208,443,444],{},"国内市场外链",[202,446,447,450,453,456],{},[205,448,449],{},"2025 年后买卖外链基本失效",[205,451,452],{},"主要靠行业媒体(亿邦动力、雨果网)发软文",[205,454,455],{},"知乎、小红书做内容,带官网链接",[205,457,458],{},"ERP 工具可以做行业榜单收录",[11,460,461,244],{},[208,462,463],{},"海外市场外链",[202,465,466,469,472,475,478],{},[205,467,468],{},"买卖外链 Google 还能识别一些,但风险高",[205,470,471],{},"主要靠 guest post(行业博客投稿)",[205,473,474],{},"Reddit、HackerNews、Product Hunt 发起讨论",[205,476,477],{},"行业媒体(EcomCrew、Practical Ecommerce)发深度内容",[205,479,480],{},"工具类网站(G2、Capterra)做产品收录",[11,482,483,486],{},[208,484,485],{},"踩坑经验",":我们给一个跨境 ERP 客户做外链,一开始想用国内的\"行业媒体发稿\"思路打海外,在几个跨境电商博客买了 guest post。3 个月后查外链,Google 把其中 2 个判定为 paid link,权重清零。后来改成做 Product Hunt 发布 + Reddit 真实分享,效果反而好。",[11,488,489],{},"跨境软件出海的外链建设,不要图快。Google 对 paid link 的识别越来越强,真实的内容和外链才是长期资产。",[15,491,492],{"id":492},"具体踩坑经验",[11,494,495],{},"总结几个我们踩过的坑:",[11,497,498,501],{},[208,499,500],{},"坑 1:双语站用翻译插件","\n客户图省事,用 WPML 自动翻译英文站。3 个月后流量起不来。原因:翻译质量太差,关键词不对。改回人工重写后,流量 2 个月涨 4 倍。",[11,503,504,507],{},[208,505,506],{},"坑 2:中文策略直接搬到英文","\n中文站写\"亚马逊运营技巧\"这类宽泛词做内容。搬到英文站做\"Amazon selling tips\",竞争激烈排名起不来。改成做\"how to optimize Amazon listing for A9 algorithm\"这种长尾,1 个月进前 10。",[11,509,510,513],{},[208,511,512],{},"坑 3:海外外链买 PBN","\n图便宜买了 PBN(私人博客网络)外链。2 个月后 Google 算法更新,整个站排名掉了 30 名以上。清理掉 PBN 外链、提交 reconsideration,3 个月才恢复。",[11,515,516,519],{},[208,517,518],{},"坑 4:忽视 hreflang","\n一个客户英文站和中文站内容相似度高,Google 把英文页当中文页的副本索引,中文站排名被压。修正 hreflang 后,中文站 1 个月恢复。",[15,521,522],{"id":522},"我们不接的跨境软件",[11,524,525],{},"不是所有跨境软件都适合做 SEO。我们不接:",[202,527,528,534,540,549],{},[205,529,530,533],{},[208,531,532],{},"国内站没建好的",":先做好一边再做另一边",[205,535,536,539],{},[208,537,538],{},"英文站纯翻译的",":不愿人工重写,做不出效果",[205,541,542,545,548],{},[208,543,544],{},"预算不够做双线的",[546,547],"seo",{}," 预算砍半,双线都做不起来,不如集中一边",[205,550,551,554,556],{},[208,552,553],{},"产品没 PMF 的",[546,555],{}," 之前先解决产品问题",[15,558,559],{"id":559},"总结",[11,561,562],{},"跨境软件 SEO 的核心:双线独立做,不要套模板。国内用 5118 挖词、百度偏好短内容、外链靠行业媒体;海外用 Ahrefs 挖词、Google 偏好长内容、外链靠真实 guest post 和社区。踩过的坑:hreflang、PBN、翻译插件、关键词直译。",[11,564,565],{},"如果你做跨境软件,想知道你的双线 SEO 现状,联系我们做个诊断。",{"title":78,"searchDepth":79,"depth":79,"links":567},[568,569,570,571,572,573,574,575],{"id":187,"depth":79,"text":188},{"id":235,"depth":79,"text":235},{"id":280,"depth":79,"text":281},{"id":394,"depth":79,"text":394},{"id":436,"depth":79,"text":436},{"id":492,"depth":79,"text":492},{"id":522,"depth":79,"text":522},{"id":559,"depth":79,"text":559},"SEO","2025-12-22","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。",{},"\u002Fblog\u002Fcross-border-software-seo",{"title":179,"description":578},"blog\u002Fcross-border-software-seo",[584,585,586,587],"跨境电商SEO","出海","双语SEO","软件出海","7ayDI_UFLfbHcY4_m5_xBtF3j8A_U3DLrZQK8DBWWXg",1794088109851]