[{"data":1,"prerenderedAt":386},["ShallowReactive",2],{"blog-ai-crawler-frequency-signal":3,"blog-prev-ai-crawler-frequency-signal":135,"blog-next-ai-crawler-frequency-signal":268},{"id":4,"title":5,"author":6,"body":7,"canonical":119,"category":120,"cover":119,"date":121,"description":122,"draft":123,"extension":124,"keywords":119,"meta":125,"navigation":126,"noindex":123,"ogImage":119,"path":127,"seo":128,"seoDescription":119,"seoTitle":119,"stem":129,"tags":130,"updated":119,"__hash__":134},"blog\u002Fblog\u002Fai-crawler-frequency-signal.md","AI 爬虫来得勤不勤，能看出内容有没有人用","元析智能",{"type":8,"value":9,"toc":108},"minimark",[10,14,18,21,24,27,30,33,36,39,42,45,48,51,54,57,60,63,66,69,72,75,78,81,84,87,90,93,96,99,102,105],[11,12,13],"p",{},"2026 年 10 月我们对比了两个规模相近的软件站。一个 GPTBot 日均抓取一百二十次，另一个只有八次。后者上线两年，内容数量是前者的两倍。频次差距说明的不是站大站小，是内容有没有被用。",[15,16,17],"h2",{"id":17},"先看频次怎么分布",[11,19,20],{},"按 UA 分组，把主流 AI 爬虫分开：OpenAI、Anthropic、Google 的 AI 抓取、国内的几家。",[11,22,23],{},"按路径分组，看它抓的是产品页、文档页还是博客。",[11,25,26],{},"按日期分组，看是持续抓还是集中在某几天。",[11,28,29],{},"这三组数据拉出来，才能谈频次高低。",[15,31,32],{"id":32},"频次高说明什么",[11,34,35],{},"第一，你的内容被反复检查。AI 需要新数据时会回来。",[11,37,38],{},"第二，站点在训练或索引的候选池里。不是所有站都会进入。",[11,40,41],{},"第三，某些页面被高频抓取，通常是被引用的候选。",[11,43,44],{},"频次高不代表被引用多，但它是前置信号。",[15,46,47],{"id":47},"频次低说明什么",[11,49,50],{},"可能是内容没被发现。检查站点地图、内链、外链，看爬上来的路径通不通。",[11,52,53],{},"可能是内容没价值。两年没更新的博客，AI 抓一次就知道不用再来。",[11,55,56],{},"也可能是被 robots 挡住了。检查 robots.txt 里有没有误屏蔽 AI 爬虫。",[15,58,59],{"id":59},"怎么用这个数据",[11,61,62],{},"把高频抓取的页面列出来，再去问 AI 相关问题，看这些页面有没有被引用。",[11,64,65],{},"两种结果都有意义：高频且被引用，说明路径对了，可以复制；高频但没被引用，说明内容抓到了但没被采用，要改结构。",[11,67,68],{},"低频但被引用的页面，值得重点维护，因为它是少数有效的页面。",[15,70,71],{"id":71},"一个真实的对比",[11,73,74],{},"2026 年 8 月对比的两个站，频次低的那个内容数量多一倍，但七成页面是两年前的。",[11,76,77],{},"我们把其中二十篇更新之后，GPTBot 的抓取频次在四周内从日均八次涨到三十一次。",[11,79,80],{},"频次变化比引用变化来得早，这就是它的价值。",[15,82,83],{"id":83},"两个注意",[11,85,86],{},"第一，日志里的 AI 爬虫有假货。UA 谁都能写，先做 IP 验证再统计。",[11,88,89],{},"第二，CDN 缓存会让日志少记请求。如果站前有 CDN，抓取频次会比真实值低。",[11,91,92],{},"这两点不处理，频次数据就没法用。",[11,94,95],{},"我们通常先跑一周原始数据，确认没有异常再开始看趋势。单日数据波动太大，看不出东西。",[15,97,98],{"id":98},"一个态度",[11,100,101],{},"AI 爬虫的抓取频次是目前最便宜的可观测信号。",[11,103,104],{},"不用买工具，从服务器日志里就能拉出来。每周看一次，比每月猜一次强。",[11,106,107],{},"我们的做法是把频次做成周报里的一个小块，和引用记录放在一起看。两个数据一对，方向就清楚了。",{"title":109,"searchDepth":110,"depth":110,"links":111},"",2,[112,113,114,115,116,117,118],{"id":17,"depth":110,"text":17},{"id":32,"depth":110,"text":32},{"id":47,"depth":110,"text":47},{"id":59,"depth":110,"text":59},{"id":71,"depth":110,"text":71},{"id":83,"depth":110,"text":83},{"id":98,"depth":110,"text":98},null,"技术SEO","2026-10-11","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。",false,"md",{},true,"\u002Fblog\u002Fai-crawler-frequency-signal",{"title":5,"description":122},"blog\u002Fai-crawler-frequency-signal",[131,132,133],"AI爬虫","日志分析","GEO","WZmCv1YmECps7Y7ZtmMIX7qHwvYk3rt__56rLii2Xo8",{"id":136,"title":137,"author":6,"body":138,"canonical":119,"category":257,"cover":119,"date":258,"description":259,"draft":123,"extension":124,"keywords":119,"meta":260,"navigation":126,"noindex":123,"ogImage":119,"path":261,"seo":262,"seoDescription":119,"seoTitle":119,"stem":263,"tags":264,"updated":119,"__hash__":267},"blog\u002Fblog\u002Fcontent-production-pipeline.md","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",{"type":8,"value":139,"toc":248},[140,143,146,149,152,155,158,161,164,167,170,173,176,179,182,185,188,191,194,197,200,203,206,209,212,215,218,221,224,227,230,233,236,239,242,245],[11,141,142],{},"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。",[15,144,145],{"id":145},"七个环节",[11,147,148],{},"选题：确定写什么，产出是一个标题加一句为什么写。",[11,150,151],{},"关键词：确定目标词和词群，产出是一张关键词表。",[11,153,154],{},"大纲：确定结构，产出是 H2\u002FH3 和每节要点。",[11,156,157],{},"初稿：写成可读的稿子，产出是 1500 到 4000 字的 Markdown。",[11,159,160],{},"审稿：事实、结构、语言三层检查，产出是修改意见和终稿。",[11,162,163],{},"发布：排版、配图、内链、提交收录，产出是线上页面。",[11,165,166],{},"复盘：发布后 30 天看数据，产出是一条结论。",[15,168,169],{"id":169},"每个环节的负责人",[11,171,172],{},"选题和关键词：内容负责人。",[11,174,175],{},"大纲：写手和内容负责人一起定。",[11,177,178],{},"初稿：写手。",[11,180,181],{},"审稿：内容负责人加一个业务同事。",[11,183,184],{},"发布：运营。",[11,186,187],{},"复盘：内容负责人。",[11,189,190],{},"关键不是谁做，是每个环节都有明确的人，别出现\"大家都以为别人会做\"。",[15,192,193],{"id":193},"最容易卡住的环节",[11,195,196],{},"第一个卡点，大纲。写手等大纲，内容负责人没时间定，一等就是三天。",[11,198,199],{},"解法：大纲用 30 分钟的会一次定完，当场确认，别来回改。",[11,201,202],{},"第二个卡点，审稿。稿子压在一堆事后面，一周没人看。",[11,204,205],{},"解法：固定审稿时间，比如每周二、周四上午，其他时间不接审稿。",[11,207,208],{},"第三个卡点，发布。排版和配图拖延，稿子写完三天才上线。",[11,210,211],{},"解法：发布包给运营，写手写完就交，别等排版。",[15,213,214],{"id":214},"我们的流水线怎么跑",[11,216,217],{},"每周一上午定选题和大纲，周二到周四写初稿，周五统一审稿，周一批量发布。",[11,219,220],{},"2025 年 10 月我们把流程定下来，一篇文章从选题到上线 7 到 10 天，同时跑 4 到 6 篇，产能稳定在每周 3 篇。",[15,222,223],{"id":223},"一个提醒",[11,225,226],{},"流水线的目的是减少等待，不是把人变成机器。写手需要思考时间，审稿需要判断时间，这些别压缩。",[11,228,229],{},"我们一般给每篇留 1.5 天的缓冲，防止突发情况卡住整条线。",[15,231,232],{"id":232},"我们用什么工具",[11,234,235],{},"不用复杂的系统。一张在线表格管选题和状态，一个共享文档放终稿，一个日历排发布时间。",[11,237,238],{},"工具越简单，越容易坚持。我们试过两个内容管理系统，最后都回到了表格，因为团队宁愿用熟悉的工具。",[15,240,241],{"id":241},"一句判断",[11,243,244],{},"内容产能的瓶颈，往往不在写，在交接。把交接规则定清楚，同样的团队能多产出 30%。",[11,246,247],{},"流程不用复杂，一张表、几个固定时间点就够。",{"title":109,"searchDepth":110,"depth":110,"links":249},[250,251,252,253,254,255,256],{"id":145,"depth":110,"text":145},{"id":169,"depth":110,"text":169},{"id":193,"depth":110,"text":193},{"id":214,"depth":110,"text":214},{"id":223,"depth":110,"text":223},{"id":232,"depth":110,"text":232},{"id":241,"depth":110,"text":241},"内容运营","2026-10-10","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。",{},"\u002Fblog\u002Fcontent-production-pipeline",{"title":137,"description":259},"blog\u002Fcontent-production-pipeline",[265,266,257],"内容流程","内容团队","lOq75rW301Dnk9g-21bHDlwMnf0LYrqihUm274jrrJU",{"id":269,"title":270,"author":6,"body":271,"canonical":119,"category":374,"cover":119,"date":375,"description":376,"draft":123,"extension":124,"keywords":119,"meta":377,"navigation":126,"noindex":123,"ogImage":119,"path":378,"seo":379,"seoDescription":119,"seoTitle":119,"stem":380,"tags":381,"updated":119,"__hash__":385},"blog\u002Fblog\u002Ffunding-press-release-seo.md","融资和产品发布，媒体稿怎么写才有搜索价值",{"type":8,"value":272,"toc":365},[273,276,279,282,285,288,291,294,297,300,303,306,309,312,315,318,321,324,327,330,333,336,339,342,345,348,351,354,356,359,362],[11,274,275],{},"2026 年 10 月一个客户发了融资新闻，稿子铺了十几家媒体，一周后搜索几乎看不到。他说钱花了没效果。新闻稿能不能被搜到，取决于你写的是事件还是问题。",[15,277,278],{"id":278},"为什么新闻稿会沉底",[11,280,281],{},"大部分媒体稿的写法是：某公司完成某轮融资，投资方是谁，创始人说了一句愿景。",[11,283,284],{},"这种稿子只有事件信息，没有可被搜索的问题。用户不会搜\"某公司完成 A 轮融资\"，除非他已经在关注你。",[11,286,287],{},"能长期带来流量的稿子，一定回答了某个具体问题。",[15,289,290],{"id":290},"三个可检索的入口",[11,292,293],{},"第一个，行业问题。比如\"为什么中小团队的日志成本降不下来\"，融资只是背景，主体是问题。",[11,295,296],{},"第二个，产品能力。比如\"我们做了什么事，让 API 监控的告警噪音降低一半\"。",[11,298,299],{},"第三个，数据发现。行业报告类的内容，带独家数据的最容易被引用。",[11,301,302],{},"三个入口里，第一个和第三个的搜索价值最高。",[15,304,305],{"id":305},"标题怎么写",[11,307,308],{},"别用\"某公司完成数千万融资\"当主标题。",[11,310,311],{},"用问题或者结论当标题，把融资放进副标题或者正文第一段。",[11,313,314],{},"比如\"日志成本为什么降不下来，我们拆了三十家公司的账单\"，这样的标题能被搜到，也能被媒体转载。",[15,316,317],{"id":317},"稿子发出去之后",[11,319,320],{},"第一，官网发一份完整版，比媒体版更长，带数据和图表。",[11,322,323],{},"第二，在行业媒体投一到两篇深度稿，不要广撒网。",[11,325,326],{},"第三，把稿子拆成几条社交内容，在领英、知乎、公众号分发。",[11,328,329],{},"第四，观察哪些媒体带来的访问最多，下次重点投。",[15,331,332],{"id":332},"怎么衡量",[11,334,335],{},"看三个数：稿子带来的官网访问、品牌词搜索量、被其他站转载和引用的数量。",[11,337,338],{},"第三个最难量化，但最重要。转载时会带链接，这些链接是新闻稿真正的长期价值。",[11,340,341],{},"搜索品牌名加融资关键词，看第一页有多少是媒体报道。这个能反映铺得够不够。",[15,343,344],{"id":344},"发布前两周该准备什么",[11,346,347],{},"官网的落地页要能承接从媒体过来的访问：产品页、数据页、团队页都得是最新的。",[11,349,350],{},"很多公司的稿子发出去，用户点进官网找不到稿子里说的那个功能，转化就断在这一步。",[11,352,353],{},"把稿子里提到的每个功能、每个数字，在官网上都找到对应页面，这件事比多发五家媒体重要。",[15,355,98],{"id":98},[11,357,358],{},"新闻稿的价值不在发布当天，在之后的半年一年。",[11,360,361],{},"它是一份可以被反复引用的资料。写的时候想着\"半年后有人搜这个问题，这篇能不能被找到\"，角度就不一样了。",[11,363,364],{},"花在写稿上的时间，应该比花在发稿渠道上的时间多。",{"title":109,"searchDepth":110,"depth":110,"links":366},[367,368,369,370,371,372,373],{"id":278,"depth":110,"text":278},{"id":290,"depth":110,"text":290},{"id":305,"depth":110,"text":305},{"id":317,"depth":110,"text":317},{"id":332,"depth":110,"text":332},{"id":344,"depth":110,"text":344},{"id":98,"depth":110,"text":98},"行业观察","2026-10-12","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。",{},"\u002Fblog\u002Ffunding-press-release-seo",{"title":270,"description":376},"blog\u002Ffunding-press-release-seo",[382,383,384],"媒体稿","品牌","内容分发","KIn7EaSyZOMYbaFXPyYMHupdKsiQxkHI1wLnYXf7GRE",1794088100891]