[{"data":1,"prerenderedAt":387},["ShallowReactive",2],{"blog-sitemap-rss-ai-crawl":3,"blog-prev-sitemap-rss-ai-crawl":213,"blog-next-sitemap-rss-ai-crawl":321},{"id":4,"title":5,"author":6,"body":7,"canonical":197,"category":198,"cover":197,"date":199,"description":200,"draft":201,"extension":202,"keywords":197,"meta":203,"navigation":204,"noindex":201,"ogImage":197,"path":205,"seo":206,"seoDescription":197,"seoTitle":197,"stem":207,"tags":208,"updated":197,"__hash__":212},"blog\u002Fblog\u002Fsitemap-rss-ai-crawl.md","站点地图和 RSS 对 AI 抓取还有多少用：六个站三个月的日志","元析智能",{"type":8,"value":9,"toc":186},"minimark",[10,14,18,85,88,92,95,98,101,104,108,111,114,117,121,124,131,137,143,146,149,152,155,158,161,165,168,171,174,177,180,183],[11,12,13],"p",{},"站点地图对 AI 抓取有用，但没有你想的那么有用。2026 年 2 月到 4 月，我们统计了六个客户站的日志：AI 爬虫总请求 11.4 万次，其中请求 sitemap 的占 11%，请求 RSS 的占 2%。",[15,16,17],"h2",{"id":17},"三个月的数",[19,20,21,37],"table",{},[22,23,24],"thead",{},[25,26,27,31,34],"tr",{},[28,29,30],"th",{},"项目",[28,32,33],{},"请求数",[28,35,36],{},"占比",[38,39,40,52,63,74],"tbody",{},[25,41,42,46,49],{},[43,44,45],"td",{},"页面请求",[43,47,48],{},"98100",[43,50,51],{},"86%",[25,53,54,57,60],{},[43,55,56],{},"sitemap 请求",[43,58,59],{},"12500",[43,61,62],{},"11%",[25,64,65,68,71],{},[43,66,67],{},"robots.txt 请求",[43,69,70],{},"1200",[43,72,73],{},"1%",[25,75,76,79,82],{},[43,77,78],{},"RSS 请求",[43,80,81],{},"2200",[43,83,84],{},"2%",[11,86,87],{},"sitemap 那 11% 里，大部分集中在月初和站点有新内容发布的那几天。说明它被当作\"发现新内容\"的工具，不是\"读内容\"的通道。",[15,89,91],{"id":90},"sitemap-还有用的地方","sitemap 还有用的地方",[11,93,94],{},"新页发现。刚发的页面，内链还没铺开，sitemap 是最快的通知方式。我们观察到一个规律：新页发布后 24 小时内被 AI 爬虫抓到的，八成是通过 sitemap 或内链。",[11,96,97],{},"大站分批。页面超过 5000 个的站，sitemap 分索引文件，能让爬虫按板块抓，不至于只扫首页附近。",[11,99,100],{},"URL 变更通知。改版之后旧 URL 301 到新 URL，sitemap 同步更新，能加快替换速度。",[11,102,103],{},"这三件事都不是\"提升引用\"，是\"缩短发现时间\"。别把它当优化手段，把它当基础设施。",[15,105,107],{"id":106},"rss-的实际情况","RSS 的实际情况",[11,109,110],{},"RSS 在 AI 抓取里的存在感很低。六个站三个月只有 2200 次 RSS 请求，其中大部分来自聚合器和阅读工具，不是 AI 爬虫。",[11,112,113],{},"但它还有两个用处。一是给内容团队做分发，二是被一些行业聚合站抓取后二次传播。后者的价值在品牌提及，不在 AI 引用。",[11,115,116],{},"我的建议是保留 RSS，但别为它投入。生成 RSS 是建站的基本功能，有就行。",[15,118,120],{"id":119},"真正影响-ai-抓取的是什么","真正影响 AI 抓取的是什么",[11,122,123],{},"按影响大小排：",[11,125,126,130],{},[127,128,129],"strong",{},"内链深度。"," 从首页点三次以内能到的页面，被抓概率最高。我们查过一个站，深度 5 层以上的 400 个页面，三个月里只有 12 个被 AI 爬虫碰过。",[11,132,133,136],{},[127,134,135],{},"更新频率。"," 每周有更新的板块，抓取频次明显高。停更半年的板块，爬虫基本不来了。",[11,138,139,142],{},[127,140,141],{},"页面可读性。"," 内容在 HTML 里还是靠 JS 渲染，直接决定它能不能读到。我们有个客户的核心文档是纯前端渲染，日志显示抓了，但内容没进去。",[11,144,145],{},"这三件都比 sitemap 重要。",[15,147,148],{"id":148},"一个反常识的做法",[11,150,151],{},"别把所有 URL 都塞进 sitemap。我们见过一个站 sitemap 里有 5 万个 URL，其中 4.6 万个是参数页和筛选页。",[11,153,154],{},"结果是重要页面的信号被稀释，爬虫的抓取预算花在垃圾页上。",[11,156,157],{},"现在我们的做法是分文件：核心内容一个，文档一个，博客一个，参数页不放。每个文件不超过 2000 个 URL。",[11,159,160],{},"站点地图是给爬虫省时间的，不是给自己省事的。",[15,162,164],{"id":163},"一份能用的-sitemap-该怎么写","一份能用的 sitemap 该怎么写",[11,166,167],{},"三个基本要求：只有可索引的 200 状态页、lastmod 是真的、重要页面排在前面。",[11,169,170],{},"lastmod 这块必须说。很多站的 lastmod 是程序生成的，每次构建都更新，等于没有。爬虫看几次发现日期和内容对不上，就不信这个字段了。",[11,172,173],{},"我们的做法是让 lastmod 跟着正文改动走：内容改了就更新，模板换了不动。这需要在发布流程里加一步，不复杂，但大多数站没做。",[11,175,176],{},"顺带说一句，sitemap 里的优先级字段都被忽略，不用纠结填 0.8 还是 1.0。",[15,178,179],{"id":179},"什么时候该停下来别折腾",[11,181,182],{},"站里页面不到 500 个的时候，sitemap 的作用基本可以被内链替代。真正该花时间的是把重要页面点三次能到、每周更新几页。",[11,184,185],{},"我们见过太多人在这上面花两周，然后把\"没效果\"归因到 AI 不认 sitemap。方向一开始就偏了。",{"title":187,"searchDepth":188,"depth":188,"links":189},"",2,[190,191,192,193,194,195,196],{"id":17,"depth":188,"text":17},{"id":90,"depth":188,"text":91},{"id":106,"depth":188,"text":107},{"id":119,"depth":188,"text":120},{"id":148,"depth":188,"text":148},{"id":163,"depth":188,"text":164},{"id":179,"depth":188,"text":179},null,"技术SEO","2026-04-20","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。",false,"md",{},true,"\u002Fblog\u002Fsitemap-rss-ai-crawl",{"title":5,"description":200},"blog\u002Fsitemap-rss-ai-crawl",[209,210,211,198],"sitemap","RSS","AI爬虫","QS9NV_FIQZ4WgeNItEpiAiaZuVFIaylqWFJ6TyfbSng",{"id":214,"title":215,"author":6,"body":216,"canonical":197,"category":309,"cover":197,"date":310,"description":311,"draft":201,"extension":202,"keywords":197,"meta":312,"navigation":204,"noindex":201,"ogImage":197,"path":313,"seo":314,"seoDescription":197,"seoTitle":197,"stem":315,"tags":316,"updated":197,"__hash__":320},"blog\u002Fblog\u002Fscenario-keyword-matrix.md","场景词矩阵：行业乘角色乘流程，一张表挖出 300 个词",{"type":8,"value":217,"toc":299},[218,221,224,227,230,233,236,239,242,245,248,251,254,257,260,263,266,269,272,275,278,281,284,287,290,293,296],[11,219,220],{},"场景词比品类词好打，问题是没人知道从哪挖。我们用一张三列表来生成：行业、角色、流程。三列交叉，一次能出 300 个候选词，筛完能进排期的 40 个左右。",[15,222,223],{"id":223},"三列怎么填",[11,225,226],{},"行业列写你的客户所在行业，8 到 12 个。角色列写用这个产品的人，4 到 6 个，比如运维、开发、采购、财务。流程列写这个角色用产品做的事，6 到 8 个。",[11,228,229],{},"三列相乘，理论上就是几百种组合。不用全填，先填最有把握的两列交叉。",[15,231,232],{"id":232},"一个具体例子",[11,234,235],{},"做日志分析工具的，行业填电商、金融、游戏、SaaS。角色填运维、后端开发、SRE。流程填排障、容量规划、审计、告警配置。",[11,237,238],{},"“电商 运维 排障”就变成一个词：电商大促期间怎么排查接口超时。这就是场景词。",[15,240,241],{"id":241},"生成之后要筛",[11,243,244],{},"第一筛：这个词有人搜吗。用搜索后台和工具过一遍，零曝光的先放一边。",[11,246,247],{},"第二筛：我们有没有素材。没有真实客户场景的，写出来还是空的。",[11,249,250],{},"第三筛：竞品写过吗。写过的看能不能换角度，没写过的优先。",[15,252,253],{"id":253},"为什么场景词更好打",[11,255,256],{},"品类词的前十是官网、媒体、点评站，新页面很难挤进去。场景词的前十常常只有两三个像样的结果，一篇讲透的页面就能排上去。",[11,258,259],{},"2025 年我们给一个客户做了 40 个场景词，12 个月内 22 个进了前三页，品类词一个都没进。",[15,261,262],{"id":262},"一张表的维护",[11,264,265],{},"行业和角色变了就更新，流程列每季度补一次。客户签了新行业，第一件事是往矩阵里加一列。",[11,267,268],{},"这张表比关键词工具导出的表好用，因为它自带语境，写的时候不用再想“这篇谁来读”。",[15,270,271],{"id":271},"一个提醒",[15,273,274],{"id":274},"谁来一起填这张表",[11,276,277],{},"别一个人填。让销售填行业和角色，让技术支持填流程，让市场的人补客户原话。",[11,279,280],{},"我们 2026 年 2 月做过一次，三拨人各填一列，交叉出来的词比一个人拍脑袋多出两倍，而且更接近真实问法。",[15,282,283],{"id":283},"多久更新一次",[11,285,286],{},"半年一次就够。行业和角色的变化慢，流程列跟着产品迭代走。",[11,288,289],{},"更新的标准是：新签了三个同行业客户，就加一列；产品上线了新模块，就补一行。",[11,291,292],{},"矩阵生成的词不要直接当标题。组合词往往很生硬，要改成人话。比如“SaaS 财务 对账”要写成“SaaS 公司财务怎么每月对账”，读起来才正常。",[11,294,295],{},"一个词能不能写，还要看它有没有对应的客户场景可以引用。写不出来的场景，先别排进计划。",[11,297,298],{},"矩阵的价值不是生成词，是把“写什么”和“谁在什么情况下会搜”绑在一起。绑不上，就还是空的。",{"title":187,"searchDepth":188,"depth":188,"links":300},[301,302,303,304,305,306,307,308],{"id":223,"depth":188,"text":223},{"id":232,"depth":188,"text":232},{"id":241,"depth":188,"text":241},{"id":253,"depth":188,"text":253},{"id":262,"depth":188,"text":262},{"id":271,"depth":188,"text":271},{"id":274,"depth":188,"text":274},{"id":283,"depth":188,"text":283},"SEO","2026-04-19","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。",{},"\u002Fblog\u002Fscenario-keyword-matrix",{"title":215,"description":311},"blog\u002Fscenario-keyword-matrix",[317,318,319],"场景词","关键词矩阵","选题","j0BG51CIzcVmIpUWv16xdjyP0c4CLtwqrdBTETlrGs4",{"id":322,"title":323,"author":6,"body":324,"canonical":197,"category":198,"cover":197,"date":376,"description":377,"draft":201,"extension":202,"keywords":197,"meta":378,"navigation":204,"noindex":201,"ogImage":197,"path":379,"seo":380,"seoDescription":197,"seoTitle":197,"stem":381,"tags":382,"updated":197,"__hash__":386},"blog\u002Fblog\u002Ffooter-legal-details.md","页脚那行版权年份和备案号，别随手写",{"type":8,"value":325,"toc":371},[326,329,332,335,338,341,347,353,359,362,365,368],[11,327,328],{},"页脚有五个字段，加起来不超过四行，但它是最容易被看出\"没人管\"的地方。2026 年 3 月我们评审一个客户的站，页脚写着\"© 2023\"，备案号是一张图片，公司名还少了一个\"科技\"。这三处，任何一处都会让人怀疑整站信息是不是三年没维护过。",[15,330,331],{"id":331},"页脚该写的五个字段",[11,333,334],{},"版权年份，写当前年份，不用 JavaScript 自动生成，直接写死，每年改一次。公司全称，和营业执照一致。ICP 备案号，用文字加链接，链到工信部查询页。公安备案号，有就放，没有就去办。最后一个联系入口，邮箱或电话都行。",[11,336,337],{},"五个字段之外的东西尽量别加。合作伙伴 logo、支付图标、各种认证徽章堆在页脚，会让页面变脏，而且图标本身就是一堆图片请求。我们测量过一个客户的页脚，12 个徽章图标加起来 380KB，为了页面上几行装饰，付出的是移动端首屏时间。",[15,339,340],{"id":340},"三类高频错误",[11,342,343,346],{},[127,344,345],{},"年份没更新。"," 这是最常见的，也最容易被客户一眼看到。我们查过的客户站里，页脚年份滞后的占四成。",[11,348,349,352],{},[127,350,351],{},"备案号做成图片。"," 有些站为了排版整齐，把备案号做成一张小图。用户读不到，也没法点，等于写了但没用。",[11,354,355,358],{},[127,356,357],{},"公司名和备案主体不一致。"," 这种情况在用过品牌名做站名、后来没同步全称的站上特别常见。工商、备案、合同三处名字对不上，采购流程里一定会被问。",[15,360,361],{"id":361},"怎么维护",[11,363,364],{},"把页脚放进两件事里：一是新站上线检查表，逐项签字；二是年度更新计划，每年 1 月改年份、查备案状态。",[11,366,367],{},"我们给客户的做法是列一张\"站点基础信息表\"，页脚、关于页、备案系统、地图商户四处共用同一份信息，改一处必须同步另外三处。表格看着笨，但两年下来，因为信息不一致被客户挑出来的问题从每月两三起降到零。",[11,369,370],{},"另外提醒一句：页脚里的链接别全用 nofollow。备案查询链接、隐私政策、关于页这些是该被蜘蛛跟的，全加 nofollow 等于自己把内页藏起来。我们就见过一个站，页脚 8 个链接全加了 nofollow，内页收录慢得离谱。",{"title":187,"searchDepth":188,"depth":188,"links":372},[373,374,375],{"id":331,"depth":188,"text":331},{"id":340,"depth":188,"text":340},{"id":361,"depth":188,"text":361},"2026-04-21","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。",{},"\u002Fblog\u002Ffooter-legal-details",{"title":323,"description":377},"blog\u002Ffooter-legal-details",[383,384,385],"页脚","备案","站点细节","MrRqwORPJWKqP2loD2i-vZDF7RnM81HbEI607-YAH9c",1794088106613]