[{"data":1,"prerenderedAt":221},["ShallowReactive",2],{"site-settings":3,"$fJ2Ox7HrPn9_qJ8OZcqT5uFs_ZntCL4tmfX0sRsW3Ews":26,"related-posts":48,"$fHRHdHVyockbPnxtXRT04-OL0FAEQJtKl_qaEaFeliH4":220},{"id":4,"site_name":5,"site_subtitle":6,"site_description":7,"author":8,"email":9,"github":10,"icp":10,"page_size":11,"enable_comment":12,"enable_rss":12,"updated_at":13,"site_icon":14,"site_og_image":15,"site_keywords":16,"site_og_image_width":24,"site_og_image_height":25},1,"LiteSite（轻站）","把你想做的网站说出来","LiteSite（轻站）探索自然语言驱动的零代码建站体验，展示响应式网站与前端原型的设计方向。","Admin","support@litesite.com",null,10,true,"2026-10-05T03:56:44.502+00:00","https:\u002F\u002Fnklbhnvyoggxghtaoizu.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-images\u002F2026-10-05-1765bce7-01f1-4206-8d5d-6a717d17f206.png","https:\u002F\u002Fnklbhnvyoggxghtaoizu.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-images\u002F2026-10-05-157c3c1d-2722-474c-860b-d3f38d6a674d.png",[17,18,19,20,21,22,23],"LiteSite","轻站","零代码建站","AI建站","响应式网站","前端原型","网站设计",1200,630,{"id":27,"slug":28,"title":29,"summary":30,"content":31,"status":32,"tags":33,"seo_keywords":37,"published_at":43,"updated_at":44,"category":45},15,"geo-core-metrics-calculation","GEO 的核心指标你怎么理解？怎么算？","用同一组问题和明确的分母，算清 GEO 的品牌提及率、引用率、有效引用率、竞争提及份额与有效线索率。","GEO 核心指标怎么理解？我的判断是：先区分“答案里出现了你”“引用了你的页面”“引用得是否正确”，最后再看有没有带来有效业务行动。把这些数混成一个“AI 曝光量”，很容易得出错误结论。上一篇[从业务理解做 GEO](\u002Fblog\u002Fgeo-business-understanding)讨论了内容该回答什么；这篇只解决怎么量、怎么算。\n\n## 先固定统计口径：一次回答算一个观察值\n\n准备一组贴近业务决策的问题，按发现问题、比较方案、准备采购等阶段分层。品牌词问题与非品牌词问题分开统计，因为“某某品牌怎么样”天然更容易提及品牌。固定所测平台、地区、语言、日期范围和运行方式；同一个问题可以重复运行，以观察答案波动。\n\n下面用一个纯示例说明：在同一平台测 40 个问题，每题运行 3 次，共得到 **120 条可评估回答**。如果有生成失败或无法读取的回答，应另记失败数，不悄悄从报告里消失。这里的 120 是抽样观察值，不代表真实用户看到了 120 次。\n\n![GEO 核心指标算例：120 条回答中 42 条提及品牌、18 条引用来源、16 条引用有效](https:\u002F\u002Fnklbhnvyoggxghtaoizu.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-images\u002F2026-10-05-8569dbee-572b-4cc2-962d-e58f7b4d54c7.png)\n\n*图：同一批回答的逐层统计。每一步的分母不同，不能只比较百分比大小。*\n\n## 1. 品牌提及率：有没有进入答案\n\n**品牌提及率 = 提及品牌的回答数 ÷ 可评估回答数 × 100%**。示例中，120 条回答有 42 条提及品牌，提及率是 **42 ÷ 120 = 35%**。同一条回答重复出现品牌名称，仍只计一次。\n\n这一指标反映在选定问题集中的可见程度，不等于平台真实曝光量。应同时拆出非品牌问题的提及率；否则品牌词占比一提高，指标可能“变好”，但新客发现能力并没有改善。\n\n## 2. 来源引用率：有没有指向你的内容\n\n**来源引用率 = 引用自有域名页面的回答数 ÷ 可评估回答数 × 100%**。假设 18 条回答引用了自有页面，引用率就是 **18 ÷ 120 = 15%**。一条回答引用多个自有 URL，按“回答数”口径仍计一次；另可记录具体 URL 的引用次数，找出哪篇内容更常被用到。\n\n如果这 18 条都属于前述 42 条品牌提及回答，还可以算条件引用率：**18 ÷ 42 ≈ 42.9%**。报告里要写清它的分母是“提及品牌的回答”，与 15% 不是同一个指标。有些回答可能引用页面却不写品牌名，遇到这种情况就单独统计交集，不能直接用 18 除以 42。\n\n## 3. 有效引用率：引用是否真的支持回答\n\n被引用不代表引用正确。先制定人工判定规则：链接是否可打开，页面是否确实支持答案中的关键说法，产品能力、价格或时效是否被误述。**有效引用率 = 符合规则的引用回答数 ÷ 引用自有页面的回答数 × 100%**。若 18 条被引用回答中有 16 条符合规则，结果是 **16 ÷ 18 ≈ 88.9%**。\n\n这项指标适合与“错误陈述清单”一起看。低质量引用比没有引用更值得优先处理：先修正过时页面、模糊表述和相互矛盾的信息，再考虑扩展内容数量。\n\n## 4. 竞争提及份额：在同一问题集里占多少\n\n对同一批非品牌问题，记录自家及指定竞品各被提及的回答数。**竞争提及份额 = 自家提及次数 ÷（自家与所选竞品提及次数之和）× 100%**。另以这组非品牌问题为例，假设自家 24 次、竞品 A 30 次、竞品 B 18 次，份额为 **24 ÷ (24 + 30 + 18) ≈ 33.3%**。\n\n一条回答可能同时提及多家，所以这里算的是“所选品牌的提及次数份额”，不是回答覆盖率；结果也取决于纳入了哪些竞品。不要把它写成整个市场份额。\n\n## 5. 有效线索率：内容有没有帮助业务\n\n在能够识别来源的访问中，可以计算 **有效线索率 = 符合事先定义条件的咨询数 ÷ 对应来源的访问会话数 × 100%**。例如某段时间记录到 80 次可识别的 AI 引荐会话，其中 4 次形成有效咨询，则为 **4 ÷ 80 = 5%**。有效咨询的条件应提前和业务团队约定，例如明确需求、联系方式真实且符合服务范围。\n\nAI 产品可能不传递来源信息，用户也可能先看回答、过几天直接访问网站。这使得“可识别引荐”只是部分路径，不能把 5% 当成全部 GEO 流量的转化率，更不能把所有新增收入归功于 GEO。结合咨询时提到的问题、销售反馈和页面路径做辅助判断。\n\n## 一张周报只需要把口径写全\n\n最小周报可以列：平台与日期、问题集版本、品牌词\u002F非品牌词比例、可评估回答数、提及率、来源引用率、有效引用率、竞争提及份额，以及可识别引荐带来的有效线索数。保留每次回答和判定依据，便于复核。\n\n比较前后两周时，尽量保持问题、平台和重复次数一致。模型回答会波动，连续几周同向变化比单次截图更有参考价值；问题集或竞品范围变了，就另起一组基线。核心不是找一个漂亮的百分比，而是知道用户在哪类问题上找不到你、哪里引用错了，以及这些改动是否最终帮助了业务。","published",[34],{"name":35,"slug":36},"GEO","geo",[38,39,40,41,42],"GEO核心指标","GEO指标怎么算","品牌提及率","AI引用率","生成式引擎优化效果","2026-10-05T03:40:09.175+00:00","2026-10-05T03:40:12.638839+00:00",{"id":46,"name":47,"slug":47},3,"技术笔记",{"items":49,"total":218,"page":4,"pageSize":219},[50,68,84,100,118,136,152,168,173,186,201],{"id":51,"slug":52,"title":53,"summary":54,"status":32,"tags":55,"seo_keywords":60,"published_at":66,"category":67},22,"llm-output-instability-hallucination","模型返回不稳定 \u002F 幻觉怎么办？","区分随机波动与事实幻觉，沿着样本评测、证据检索、结构校验、业务规则和人工回退建立可验证的大模型输出链路。",[56,58],{"name":57,"slug":57},"工程",{"name":59,"slug":59},"大模型",[61,62,63,64,65],"模型幻觉怎么办","大模型输出不稳定","LLM幻觉","RAG事实校验","模型评测","2026-10-05T04:17:13.315+00:00",{"id":46,"name":47,"slug":47},{"id":69,"slug":70,"title":71,"summary":72,"status":32,"tags":73,"seo_keywords":76,"published_at":82,"category":83},21,"llm-cost-budget-guardrails","大模型调用怎么控制成本？把预算控制放进请求链路","从请求准入、原子预留、真实用量结算到超支降级，设计一套能在并发与重试下运行的大模型调用预算控制链路。",[74,75],{"name":57,"slug":57},{"name":59,"slug":59},[77,78,79,80,81],"大模型调用成本控制","LLM预算管理","Token费用","模型调用限额","API成本监控","2026-10-05T04:14:07.621+00:00",{"id":46,"name":47,"slug":47},{"id":85,"slug":86,"title":87,"summary":88,"status":32,"tags":89,"seo_keywords":92,"published_at":98,"category":99},20,"llm-structured-output-reliability","结构化输出怎么保证模型按格式返回？","模型返回 JSON 不等于系统拿到可用数据。本文区分提示词、JSON 模式与 Schema 约束，说明如何用服务端校验、有限重试和业务规则构建可靠的结构化输出链路。",[90,91],{"name":57,"slug":57},{"name":59,"slug":59},[93,94,95,96,97],"大模型结构化输出","JSON Schema","模型按格式返回","LLM输出校验","约束解码","2026-10-05T04:10:02.025+00:00",{"id":46,"name":47,"slug":47},{"id":101,"slug":102,"title":103,"summary":104,"status":32,"tags":105,"seo_keywords":110,"published_at":116,"category":117},19,"geo-vs-seo-explained","理解 GEO 是什么？和 SEO 有什么区别？","GEO 是什么，和 SEO 有什么区别？从用户提问、内容呈现、优化动作与衡量指标四个角度讲清两者的关系，并给出一套适合建站团队的入门做法。",[106,109],{"name":107,"slug":108},"建站","jianzhan",{"name":35,"slug":36},[111,112,113,114,115],"GEO是什么","GEO和SEO的区别","生成式引擎优化","SEO","AI搜索优化","2026-10-05T04:07:05.673+00:00",{"id":46,"name":47,"slug":47},{"id":119,"slug":120,"title":121,"summary":122,"status":32,"tags":123,"seo_keywords":128,"published_at":134,"category":135},18,"how-nuxt-ssr-works","Nuxt 的 SSR 是怎么做的？从请求、数据获取到 Hydration","沿着一次文章页请求，看 Nuxt 如何用 Nitro 接收请求、在服务端执行 useFetch 并生成 HTML，再由浏览器复用 payload 完成 Hydration。",[124,125],{"name":57,"slug":57},{"name":126,"slug":127},"Nuxt","nuxt",[129,130,131,132,133],"Nuxt SSR","Nuxt服务端渲染","Nuxt useFetch","Hydration","Nitro渲染流程","2026-10-05T03:55:36.035+00:00",{"id":46,"name":47,"slug":47},{"id":137,"slug":138,"title":139,"summary":140,"status":32,"tags":141,"seo_keywords":144,"published_at":150,"category":151},17,"llm-api-cost-control","大模型调用怎么控制成本？一笔账算清缓存、输出与重试","先按输入、输出和额外调用算清单次成本，再用缓存、模型路由、上下文裁剪和预算限额降本，并用可用答案质量守住效果。",[142,143],{"name":57,"slug":57},{"name":59,"slug":59},[145,146,147,148,149],"大模型调用成本","LLM API成本控制","Token费用计算","提示词缓存","模型路由","2026-10-05T03:49:26.666+00:00",{"id":46,"name":47,"slug":47},{"id":153,"slug":154,"title":155,"summary":156,"status":32,"tags":157,"seo_keywords":160,"published_at":166,"category":167},16,"geo-technical-pipeline-design","GEO 的技术链路你会怎么设计？","从问题池、证据库、可读页面到答案采样与质量回写，设计一条可维护的 GEO 技术链路，并明确外部 AI 系统的边界。",[158,159],{"name":57,"slug":57},{"name":35,"slug":36},[161,162,163,164,165],"GEO技术链路","GEO技术方案","生成式引擎优化架构","AI搜索内容工程","GEO监测","2026-10-05T03:44:34.947+00:00",{"id":46,"name":47,"slug":47},{"id":27,"slug":28,"title":29,"summary":30,"status":32,"tags":169,"seo_keywords":171,"published_at":43,"category":172},[170],{"name":35,"slug":36},[38,39,40,41,42],{"id":46,"name":47,"slug":47},{"id":174,"slug":175,"title":176,"summary":177,"status":32,"tags":178,"seo_keywords":181,"published_at":184,"category":185},14,"geo-business-understanding","GEO 从业务理解开始：把真实问题写成可引用的内容","做 GEO 之前，先弄清业务服务谁、用户在什么场景下提问、哪些证据能支持回答，再用真实问题和业务结果验证内容。",[179,180],{"name":107,"slug":108},{"name":35,"slug":36},[35,113,182,183,115],"业务理解","GEO内容策略","2026-10-05T03:33:52.358+00:00",{"id":46,"name":47,"slug":47},{"id":187,"slug":188,"title":189,"summary":190,"status":32,"tags":191,"seo_keywords":193,"published_at":199,"category":200},13,"website-launch-checklist","网站上线前后检查清单：从可访问到可维护","一份实用的网站上线清单，逐项检查域名、HTTPS、手机体验、页面速度、SEO、表单、备份与上线后的监测。",[192],{"name":57,"slug":57},[194,195,196,197,198],"网站上线检查清单","网站上线","网站性能优化","网站SEO","网站维护","2026-10-05T03:24:44.969+00:00",{"id":46,"name":47,"slug":47},{"id":202,"slug":203,"title":204,"summary":205,"status":32,"tags":206,"seo_keywords":208,"published_at":214,"category":215},12,"website-planning-checklist","建站之前：用一张页面清单理清目标、结构与内容","从网站目标、访问者任务到页面清单和内容准备，用一套可执行的方法完成建站规划，避免边做边改。",[207],{"name":57,"slug":57},[209,210,211,212,213],"建站规划","网站信息架构","网站内容规划","页面清单","个人网站建设","2026-10-05T03:23:27.683+00:00",{"id":216,"name":217,"slug":217},7,"项目实践",11,100,[],1791174018341]