[{"data":1,"prerenderedAt":369},["ShallowReactive",2],{"blog-content-production-pipeline":3,"blog-prev-content-production-pipeline":146,"blog-next-content-production-pipeline":247},{"id":4,"title":5,"author":6,"body":7,"canonical":131,"category":132,"cover":131,"date":133,"description":134,"draft":135,"extension":136,"keywords":131,"meta":137,"navigation":138,"noindex":135,"ogImage":131,"path":139,"seo":140,"seoDescription":131,"seoTitle":131,"stem":141,"tags":142,"updated":131,"__hash__":145},"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","元析智能",{"type":8,"value":9,"toc":120},"minimark",[10,14,18,21,24,27,30,33,36,39,42,45,48,51,54,57,60,63,66,69,72,75,78,81,84,87,90,93,96,99,102,105,108,111,114,117],[11,12,13],"p",{},"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。",[15,16,17],"h2",{"id":17},"七个环节",[11,19,20],{},"选题：确定写什么，产出是一个标题加一句为什么写。",[11,22,23],{},"关键词：确定目标词和词群，产出是一张关键词表。",[11,25,26],{},"大纲：确定结构，产出是 H2\u002FH3 和每节要点。",[11,28,29],{},"初稿：写成可读的稿子，产出是 1500 到 4000 字的 Markdown。",[11,31,32],{},"审稿：事实、结构、语言三层检查，产出是修改意见和终稿。",[11,34,35],{},"发布：排版、配图、内链、提交收录，产出是线上页面。",[11,37,38],{},"复盘：发布后 30 天看数据，产出是一条结论。",[15,40,41],{"id":41},"每个环节的负责人",[11,43,44],{},"选题和关键词：内容负责人。",[11,46,47],{},"大纲：写手和内容负责人一起定。",[11,49,50],{},"初稿：写手。",[11,52,53],{},"审稿：内容负责人加一个业务同事。",[11,55,56],{},"发布：运营。",[11,58,59],{},"复盘：内容负责人。",[11,61,62],{},"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\"。",[15,64,65],{"id":65},"最容易卡住的环节",[11,67,68],{},"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。",[11,70,71],{},"解法：大纲用 30 分钟的会一次定完，当场确认，别来回改。",[11,73,74],{},"第二个卡点，审稿。稿子压在一堆事后面，一周没人看。",[11,76,77],{},"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。",[11,79,80],{},"第三个卡点，发布。排版和配图拖延，稿子写完三天才上线。",[11,82,83],{},"解法：发布包给运营，写手写完就交，别等排版。",[15,85,86],{"id":86},"我们的流水线怎么跑",[11,88,89],{},"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。",[11,91,92],{},"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。",[15,94,95],{"id":95},"一个提醒",[11,97,98],{},"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。",[11,100,101],{},"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。",[15,103,104],{"id":104},"我们用什么工具",[11,106,107],{},"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。",[11,109,110],{},"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。",[15,112,113],{"id":113},"一句判断",[11,115,116],{},"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%。",[11,118,119],{},"流程不用复杂，一张表、几个固定时间点就够。",{"title":121,"searchDepth":122,"depth":122,"links":123},"",2,[124,125,126,127,128,129,130],{"id":17,"depth":122,"text":17},{"id":41,"depth":122,"text":41},{"id":65,"depth":122,"text":65},{"id":86,"depth":122,"text":86},{"id":95,"depth":122,"text":95},{"id":104,"depth":122,"text":104},{"id":113,"depth":122,"text":113},null,"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。",false,"md",{},true,"\u002Fblog\u002Fcontent-production-pipeline",{"title":5,"description":134},"blog\u002Fcontent-production-pipeline",[143,144,132],"内容流程","内容团队","lOq75rW301Dnk9g-21bHDlwMnf0LYrqihUm274jrrJU",{"id":147,"title":148,"author":6,"body":149,"canonical":131,"category":132,"cover":131,"date":236,"description":237,"draft":135,"extension":136,"keywords":131,"meta":238,"navigation":138,"noindex":135,"ogImage":131,"path":239,"seo":240,"seoDescription":131,"seoTitle":131,"stem":241,"tags":242,"updated":131,"__hash__":246},"blog\u002Fblog\u002Fblog-audit-300-posts.md","一个 300 篇的博客站，审完只留了 78 篇",{"type":8,"value":150,"toc":226},[151,154,157,160,163,166,169,172,175,178,181,184,187,190,193,196,199,202,205,208,211,214,217,220,223],[11,152,153],{},"2026 年 6 月我们审了一个五年的老博客站，300 篇文章。最后建议：留 78 篇，并 96 篇，删 84 篇，改 42 篇。客户第一反应是“删这么多流量不会掉吗”，两个月后流量比审之前还高 6%。",[15,155,156],{"id":156},"这个站的问题在哪",[11,158,159],{},"三个。主题散，12 个不相关的方向都写过；重复多，同一个主题写过三到五篇；没有承接，大部分文章跟产品线对不上。",[11,161,162],{},"五年前写行业观察，三年前写产品更新，去年开始跟风写关键词。每年的方向都不一样，导致整个站没有主题。",[15,164,165],{"id":165},"留的标准",[11,167,168],{},"只有一条：有排名或有外链，且能对上现在的产品线。",[11,170,171],{},"78 篇里，56 篇有排名，22 篇有外链但没排名。这两类都值得留。",[15,173,174],{"id":174},"并的操作",[11,176,177],{},"96 篇合并成 31 篇。合并的原则是同一核心词、同一读者、同一结论。",[11,179,180],{},"先选主文，补齐内容，再 301。这个过程我们花了三周半，比预期慢，因为要给每篇主文补材料。",[15,182,183],{"id":183},"删的顾虑怎么处理",[11,185,186],{},"客户怕删了掉流量。我们给了两组数据：这 84 篇过去 12 个月总曝光占全站 1.7%，总点击占 0.9%。",[11,188,189],{},"删完第一周流量掉了 4%，第三周回升，两个月后比原来高 6%。删掉的是抓取预算里的噪音，不是流量。",[15,191,192],{"id":192},"结果",[11,194,195],{},"三个月后，收录页面数从 210 到 160，但有效收录比例从 52% 到 78%。抓取频次涨了，日志里每天被抓的页面从 300 多降到 180，每页被抓次数上升。",[11,197,198],{},"客户最直观的感受是：写新文章之后，收录速度快了。",[15,200,201],{"id":201},"教训",[11,203,204],{},"内容不是资产，只有被搜到的才是资产。一个站写五年，能留下四分之一就算不错。",[15,206,207],{"id":207},"审计之后怎么写新内容",[11,209,210],{},"老站审计完之后，新内容的选题全部从“留下”的那 78 篇里长出来：给它们补簇页、补对比页、补文档。",[11,212,213],{},"这样写出来的内容自带内链和已有权重，起来得比从零开新题快。",[15,215,216],{"id":216},"一个成本参考",[11,218,219],{},"300 篇的审计，我们两个人做了 11 个工作日。客户自己内部又花了 5 天做决策和沟通。",[11,221,222],{},"如果你们站不到 100 篇，三天能做完。别拖，越拖越贵。",[11,224,225],{},"我们现在给客户做年度复盘，都会顺带看一眼“无效页面占比”。超过 60% 的，先审计再扩写。",{"title":121,"searchDepth":122,"depth":122,"links":227},[228,229,230,231,232,233,234,235],{"id":156,"depth":122,"text":156},{"id":165,"depth":122,"text":165},{"id":174,"depth":122,"text":174},{"id":183,"depth":122,"text":183},{"id":192,"depth":122,"text":192},{"id":201,"depth":122,"text":201},{"id":207,"depth":122,"text":207},{"id":216,"depth":122,"text":216},"2026-10-09","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%。",{},"\u002Fblog\u002Fblog-audit-300-posts",{"title":148,"description":237},"blog\u002Fblog-audit-300-posts",[243,244,245],"内容审计","复盘","博客运营","BeXglAleqx5QaAGL5FMO3tEWCjC1SBagwwp3Y_vHgqE",{"id":248,"title":249,"author":6,"body":250,"canonical":131,"category":357,"cover":131,"date":358,"description":359,"draft":135,"extension":136,"keywords":131,"meta":360,"navigation":138,"noindex":135,"ogImage":131,"path":361,"seo":362,"seoDescription":131,"seoTitle":131,"stem":363,"tags":364,"updated":131,"__hash__":368},"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",{"type":8,"value":251,"toc":348},[252,255,258,261,264,267,270,273,276,279,282,285,288,291,294,297,300,303,306,309,312,315,318,321,324,327,330,333,336,339,342,345],[11,253,254],{},"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。",[15,256,257],{"id":257},"先看频次怎么分布",[11,259,260],{},"按 UA 分组，把主流 AI 爬虫分开：OpenAI、Anthropic、Google 的 AI 抓取、国内的几家。",[11,262,263],{},"按路径分组，看它抓的是产品页、文档页还是博客。",[11,265,266],{},"按日期分组，看是持续抓还是集中在某几天。",[11,268,269],{},"这三组数据拉出来，才能谈频次高低。",[15,271,272],{"id":272},"频次高说明什么",[11,274,275],{},"第一，你的内容被反复检查。AI 需要新数据时会回来。",[11,277,278],{},"第二，站点在训练或索引的候选池里。不是所有站都会进入。",[11,280,281],{},"第三，某些页面被高频抓取，通常是被引用的候选。",[11,283,284],{},"频次高不代表被引用多，但它是前置信号。",[15,286,287],{"id":287},"频次低说明什么",[11,289,290],{},"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。",[11,292,293],{},"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。",[11,295,296],{},"也可能是被 robots 挡住了。检查 robots.txt 里有没有误屏蔽 AI 爬虫。",[15,298,299],{"id":299},"怎么用这个数据",[11,301,302],{},"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。",[11,304,305],{},"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。",[11,307,308],{},"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。",[15,310,311],{"id":311},"一个真实的对比",[11,313,314],{},"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。",[11,316,317],{},"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。",[11,319,320],{},"频次变化比引用变化来得早，这就是它的价值。",[15,322,323],{"id":323},"两个注意",[11,325,326],{},"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。",[11,328,329],{},"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。",[11,331,332],{},"这两点不处理，频次数据就没法用。",[11,334,335],{},"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。",[15,337,338],{"id":338},"一个态度",[11,340,341],{},"AI 爬虫的抓取频次是目前最便宜的可观测信号。",[11,343,344],{},"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。",[11,346,347],{},"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":121,"searchDepth":122,"depth":122,"links":349},[350,351,352,353,354,355,356],{"id":257,"depth":122,"text":257},{"id":272,"depth":122,"text":272},{"id":287,"depth":122,"text":287},{"id":299,"depth":122,"text":299},{"id":311,"depth":122,"text":311},{"id":323,"depth":122,"text":323},{"id":338,"depth":122,"text":338},"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。",{},"\u002Fblog\u002Fai-crawler-frequency-signal",{"title":249,"description":359},"blog\u002Fai-crawler-frequency-signal",[365,366,367],"AI爬虫","日志分析","GEO","WZmCv1YmECps7Y7ZtmMIX7qHwvYk3rt__56rLii2Xo8",1794088100931]