[{"data":1,"prerenderedAt":379},["ShallowReactive",2],{"blog-conversational-query-rewrite":3,"blog-prev-conversational-query-rewrite":201,"blog-next-conversational-query-rewrite":318},{"id":4,"title":5,"author":6,"body":7,"canonical":185,"category":186,"cover":185,"date":187,"description":188,"draft":189,"extension":190,"keywords":185,"meta":191,"navigation":192,"noindex":189,"ogImage":185,"path":193,"seo":194,"seoDescription":185,"seoTitle":185,"stem":195,"tags":196,"updated":185,"__hash__":200},"blog\u002Fblog\u002Fconversational-query-rewrite.md","对话式查询改写：12 个关键词扩成 68 个整句问题","元析智能",{"type":8,"value":9,"toc":175},"minimark",[10,14,18,21,24,27,30,33,52,55,58,61,129,132,136,139,142,145,148,151,154,157,160,163,166,169,172],[11,12,13],"p",{},"关键词是搜索框逼出来的产物，AI 搜索收的是整句话。2026 年 2 月我们把一个客户的 12 个核心关键词按四问法改写，得到 68 个整句问题，其中 57 个现有页面答不了。这个缺口就是接下来半年的选题表。",[15,16,17],"h2",{"id":17},"关键词和问题的差别",[11,19,20],{},"\"日志分析工具\"是关键词。它没有主语、没有场景、没有约束。",[11,22,23],{},"\"我们是 30 人的 SaaS 团队，每天 200G 日志，预算有限，用哪家日志分析合适\"是问题。这句里包含了规模、数据量、预算三个筛选条件，AI 要答它，必须找同时覆盖这三点的内容。",[11,25,26],{},"关键词只能覆盖品类，问题才能覆盖决策。",[15,28,29],{"id":29},"四问法",[11,31,32],{},"一个关键词拆四个方向：",[11,34,35,39,40,43,44,47,48,51],{},[36,37,38],"strong",{},"谁在问。"," 团队规模、角色、行业。\n",[36,41,42],{},"什么场景。"," 迁移、扩容、排障、合规审计。\n",[36,45,46],{},"担心什么。"," 成本、锁定、迁移难度、能不能私有部署。\n",[36,49,50],{},"要什么结果。"," 降本、过审、缩短排障时间。",[11,53,54],{},"四个方向各出 1 到 2 个问法，一个词能出 4 到 6 句。12 个词出 68 句，符合预期。",[15,56,57],{"id":57},"一个词的改写示例",[11,59,60],{},"从\"日志分析工具\"出发：",[62,63,64,77],"table",{},[65,66,67],"thead",{},[68,69,70,74],"tr",{},[71,72,73],"th",{},"类型",[71,75,76],{},"改写后的问法",[78,79,80,89,97,105,113,121],"tbody",{},[68,81,82,86],{},[83,84,85],"td",{},"规模",[83,87,88],{},"30 人团队用哪家日志分析工具合适",[68,90,91,94],{},[83,92,93],{},"场景",[83,95,96],{},"从 ELK 迁到别的日志平台要注意什么",[68,98,99,102],{},[83,100,101],{},"担心",[83,103,104],{},"日志分析工具按量计费还是按节点计费",[68,106,107,110],{},[83,108,109],{},"结果",[83,111,112],{},"怎么把排障时间从 2 小时压到 10 分钟",[68,114,115,118],{},[83,116,117],{},"对比",[83,119,120],{},"日志分析和可观测平台是一回事吗",[68,122,123,126],{},[83,124,125],{},"边界",[83,127,128],{},"私有部署的日志分析工具有哪些",[11,130,131],{},"这 6 句都能写成一篇文章，而且互不重复。原来那个关键词只能写一篇。",[15,133,135],{"id":134},"_57-个缺口的分布","57 个缺口的分布",[11,137,138],{},"我们把 68 句逐条对照现有页面，能答的 11 句，答不了的 57 句。",[11,140,141],{},"缺口集中在三块：计费与成本 14 句、迁移与兼容 12 句、团队规模与选型 11 句。这三块恰好都是销售在电话里反复解释的内容。",[11,143,144],{},"有意思的是，客户自己写过 4 篇\"行业趋势\"，一句缺口都没填上。趋势文带来的是品牌感，不是答案。",[15,146,147],{"id":147},"怎么排优先级",[11,149,150],{},"三个筛子过一遍：",[11,152,153],{},"能不能被回答成一句话？能，优先。需要三千字才能说清的，往后放。",[11,155,156],{},"销售会不会被问到？会，优先。让销售把最近 20 通电话里被问的问题写下来，比任何工具准。",[11,158,159],{},"竞品有没有写？没写，机会大；写了但写得差，机会更大。",[11,161,162],{},"排完，57 句里能立刻动手的 18 句。这 18 句做完，够撑三个月。",[15,164,165],{"id":165},"别把它当关键词工具用",[11,167,168],{},"这套改写的产出不是关键词列表，是问题清单。问题清单要能直接拿去问 AI，测出来它现在怎么答、引了谁。",[11,170,171],{},"拿 68 句去跑 4 个引擎，你会看到一张很有意思的图：哪些问题已经有答案但答案里没你，哪些问题压根没有像样的答案。前者改内容，后者抢首发。",[11,173,174],{},"改写这一步只花两小时，值。",{"title":176,"searchDepth":177,"depth":177,"links":178},"",2,[179,180,181,182,183,184],{"id":17,"depth":177,"text":17},{"id":29,"depth":177,"text":29},{"id":57,"depth":177,"text":57},{"id":134,"depth":177,"text":135},{"id":147,"depth":177,"text":147},{"id":165,"depth":177,"text":165},null,"GEO","2026-02-01","关键词是搜索框的产物，AI 搜索收的是整句。把 12 个核心词按四问法改写，得到 68 个真实问法，其中 57 个现有页面答不了。改写表可以直接抄。",false,"md",{},true,"\u002Fblog\u002Fconversational-query-rewrite",{"title":5,"description":188},"blog\u002Fconversational-query-rewrite",[197,198,186,199],"对话式查询","关键词","内容规划","AmKjOdYlyiHFqI2VNr3NdeEp6N4HTGLXNm8Me-uwiIE",{"id":202,"title":203,"author":6,"body":204,"canonical":185,"category":306,"cover":185,"date":307,"description":308,"draft":189,"extension":190,"keywords":185,"meta":309,"navigation":192,"noindex":189,"ogImage":185,"path":310,"seo":311,"seoDescription":185,"seoTitle":185,"stem":312,"tags":313,"updated":185,"__hash__":317},"blog\u002Fblog\u002Fmerge-similar-posts-301.md","两篇相似文章合并，301 该怎么做才不漏权重",{"type":8,"value":205,"toc":295},[206,209,212,215,218,221,224,227,230,233,236,239,242,246,249,252,256,259,262,265,268,271,274,277,280,283,286,289,292],[11,207,208],{},"两篇内容高度重叠的文章，留着等于自己跟自己抢。合并是对的，但 301 做错一次就能把两个页面的权重一起弄丢。2025 年 12 月我们合并过一对文章，第一次 301 指向了分类页，第一周流量掉了 40%，改回来之后第二周才恢复。",[15,210,211],{"id":211},"先判断该不该并",[11,213,214],{},"两个页面打同一个核心词，或者解决的是同一个待办，就该并。",[11,216,217],{},"但如果两篇都还有排名、外链都不少，合并风险很大。这种情况先观察，把其中一篇改成更细的角度，别急着动。",[15,219,220],{"id":220},"主文选谁",[11,222,223],{},"看外链数和现有排名，不看发布新旧。老文外链多、排名稳，就该当主文，把新内容并进去。",[11,225,226],{},"2025 年 12 月那次，主文是一篇 2023 年的老文，外链 14 个，另一篇是 2024 年新写的，外链 1 个。选了老的，虽然内容旧，但权重在。",[15,228,229],{"id":229},"操作顺序",[11,231,232],{},"第一步，把主文内容补齐，覆盖次文所有有价值的信息点。这一步没做完，不要动 301。",[11,234,235],{},"第二步，次文正文顶部加一行提示，链到主文，挂两周。让搜索引擎和用户都先看到关系。",[11,237,238],{},"第三步，做 301，指向主文的完整 URL，不带参数。",[11,240,241],{},"第四步，把站内指向次文的内链全部换成主文。这步最容易被漏，漏了之后次文还会被反复抓取。",[15,243,245],{"id":244},"_301-指向哪","301 指向哪",[11,247,248],{},"指向主文，别指首页，别指分类页，也别指 404。",[11,250,251],{},"指向首页是懒办法，等于告诉搜索引擎这个页面的权重给全站，不给具体内容。我们那次就是图省事指了分类页，结果那个分类页跟内容不相关，权重等于白给。",[15,253,255],{"id":254},"三篇并一篇别新写","三篇并一篇，别新写",[11,257,258],{},"如果两个页面内容都一般，有人会想“新写一篇更好的，把两篇都 301 过去”。这个做法听起来干净，实际上三篇的权重都归零。",[11,260,261],{},"正确做法是从两篇里选一篇当主文，改到足够好。",[15,263,264],{"id":264},"怎么验证并对了",[11,266,267],{},"合并后第 28 天，看主文的关键词覆盖数，应该接近两篇原来覆盖的合并值，能到 70% 就算成功。",[15,269,270],{"id":270},"内链要一起改",[11,272,273],{},"站内指向次文的链接，在 301 之后会跳到主文，但锚文本还写着次文的主题。这些锚文本本身指向性很强，不改等于浪费。",[11,275,276],{},"我们的做法是合并后一周内，把次文的内链锚文本统一改成主文的核心词，改不完的先记进清单。",[15,278,279],{"id":279},"合并之后要不要重写主文",[11,281,282],{},"要。合并的本质是把两个页面的信息合到一个页面，主文必须比原来任何一篇都完整。",[11,284,285],{},"我们一般会给主文加两到三个小节，把次文里独有的例子、数据、问答搬进去。搬完之后字数通常比原来多 40%。",[15,287,288],{"id":288},"别做的三件事",[11,290,291],{},"别把 301 指向首页。别在两篇之外新写第三篇当主文。别在合并当天同时改 URL 结构，两个变量一起动，出问题查不出来。",[11,293,294],{},"排名先掉后回是正常的，通常两到四周。四周后还持续下滑，检查内链和 301 有没有漏。",{"title":176,"searchDepth":177,"depth":177,"links":296},[297,298,299,300,301,302,303,304,305],{"id":211,"depth":177,"text":211},{"id":220,"depth":177,"text":220},{"id":229,"depth":177,"text":229},{"id":244,"depth":177,"text":245},{"id":254,"depth":177,"text":255},{"id":264,"depth":177,"text":264},{"id":270,"depth":177,"text":270},{"id":279,"depth":177,"text":279},{"id":288,"depth":177,"text":288},"内容运营","2026-01-31","合并文章最贵的错误是把 301 指向首页。我们合并过一次，第一周流量掉了 40%，第二周才回来。这篇写清判断标准、主文选择、301 顺序和验证方法。",{},"\u002Fblog\u002Fmerge-similar-posts-301",{"title":203,"description":308},"blog\u002Fmerge-similar-posts-301",[314,315,316],"内容合并","301重定向","内容审计","MmSL8CjCnZRWIojDyBSiRCwgm1pI_f8aYENgguGRAks",{"id":319,"title":320,"author":6,"body":321,"canonical":185,"category":306,"cover":185,"date":368,"description":369,"draft":189,"extension":190,"keywords":185,"meta":370,"navigation":192,"noindex":189,"ogImage":185,"path":371,"seo":372,"seoDescription":185,"seoTitle":185,"stem":373,"tags":374,"updated":185,"__hash__":378},"blog\u002Fblog\u002Fai-generated-byline-risk.md","AI 批量生成署名文章，我们为什么不接这种单子",{"type":8,"value":322,"toc":363},[323,326,330,333,336,339,342,345,348,351,354,357,360],[11,324,325],{},"2026 年 1 月，一个客户拿来 60 篇已经用 AI 生成好的稿子，要求我们挂上团队作者的署名发布，一篇 200 块。我们拒了。不是因为 AI，是因为\"署名\"这两个字有重量。",[15,327,329],{"id":328},"ai-稿在作者信号上留下的三个痕迹","AI 稿在作者信号上留下的三个痕迹",[11,331,332],{},"第一是没有第一手细节。AI 能写出\"优化后效果明显提升\"，但写不出\"改完第 11 天，核心词从第 9 到第 5，询盘多了 3 个\"。读者和搜索引擎都能感觉到这种空。",[11,334,335],{},"第二是句式跨文章重复。单篇看不出来，把 10 篇放一起读，开头、转折、结尾的结构几乎一样。这种一致性在批量内容里非常显眼。",[11,337,338],{},"第三是作者页和内容对不上。一个作者今天写 Kubernetes 调优，明天写财务共享中心，后天写医疗合规，作者页上却没有对应的履历。这个矛盾是致命的。我们见过一个站，作者页上写着\"十年安全运维\"，文章却在教人如何\"关闭防火墙解决连接问题\"，两者放一起，谁看都觉得不对劲。",[15,340,341],{"id":341},"署名意味着责任",[11,343,344],{},"署名的意思是：这篇文章里的判断，这个人愿意负责。技术文章里的一个配置参数写错了，读者照着做会出故障；客户案例里的一个数字写错了，客户会找上门。",[11,346,347],{},"搜索引擎这几年在做的事，是把内容和实体关联起来。作者是谁、在哪个公司、写过什么、别人怎么评价他，这些信号越来越重要。批量署名等于主动告诉算法：这些文章没有负责人。反过来，一个真实作者持续在一个领域写，即使更新慢，积累的信任也是实的。",[15,349,350],{"id":350},"我们的底线",[11,352,353],{},"可以用 AI 做的事：查资料、列大纲、写初稿、改语句、生成对比表。不能让 AI 做的事：编数字、编客户、批量署名、不审校直接发。",[11,355,356],{},"2025 年我们发了 40 多篇文章，其中 18 篇的初稿有 AI 参与，但每一篇的事实、案例、数字都经过人工核对，客户案例都找当事人确认过。没有一篇是\"生成即发布\"。",[11,358,359],{},"那 60 篇稿子里的问题很典型：同一个\"行业痛点\"段落出现在 17 篇里，只是换了主语。发出去的前 5 篇可能没事，第 20 篇开始，同一批关键词下全是相似表述，谁都能看出来。",[11,361,362],{},"判断标准很简单：作者页上那个人，能不能回答这篇文章里的问题。如果答不上来，这个署名就是负资产，越早撤越好。",{"title":176,"searchDepth":177,"depth":177,"links":364},[365,366,367],{"id":328,"depth":177,"text":329},{"id":341,"depth":177,"text":341},{"id":350,"depth":177,"text":350},"2026-02-02","2026 年 1 月有客户拿来 60 篇 AI 稿，要求挂我们团队的名字。我们拒了。这篇讲 AI 署名在作者信号上留下的三个痕迹，以及什么情况下用 AI 辅助写作没问题。",{},"\u002Fblog\u002Fai-generated-byline-risk",{"title":320,"description":369},"blog\u002Fai-generated-byline-risk",[375,376,377],"AI写作","署名","E-E-A-T","4XZBx71zk0Fsz3l1HUau9mjXA_pehltAKWo8TYYaAVQ",1794088108731]