Skip to content
Back to search

geo

SkillWorks with Claude CodeGitHub

完整的 GEO(生成式引擎优化)服务流水线:给一个产品官网 URL 和介绍材料, 做站点诊断与 AI 答案采样、生成带验收标准的执行工单、产出可直接部署的资产 (llms.txt / JSON-LD / 定义块 / FAQ / 内容大纲与初稿)、自动验收工单是否闭环、 并打包成可直接发给客户的交付物。可按周期复跑,做长期 GEO 运营与月报。 国内和海外双市场并行:国内覆盖 智谱GLM/豆包/DeepSeek/Kimi/MiniMax/纳米AI/百度AI, 海外覆盖 Gemini/ChatGPT/Claude/Grok/Perplexity,问题库与指标按市场分开算。 当用户说「做 GEO」「生成式引擎优化」「让 AI 推荐我的产品」「AI 搜索里搜不到我们」 「GEO 方案/诊断/监测/月报/交付」「给客户做 GEO 服务」,或给一个产品网址要做 AI 可见性优化时使用。GEO 指生成式引擎优化,不是地理信息; 不用于传统 SEO 关键词排名、竞价投放或建站。

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/aigclink--geolook/badge.svg)](/listings/aigclink--geolook)

HTML

<a href="/listings/aigclink--geolook"><img src="/api/listings/aigclink--geolook/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 361 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. ---
  2. name: geo
  3. description: >-
  4. 完整的 GEO(生成式引擎优化)服务流水线:给一个产品官网 URL 和介绍材料,
  5. 做站点诊断与 AI 答案采样、生成带验收标准的执行工单、产出可直接部署的资产
  6. (llms.txt / JSON-LD / 定义块 / FAQ / 内容大纲与初稿)、自动验收工单是否闭环、
  7. 并打包成可直接发给客户的交付物。可按周期复跑,做长期 GEO 运营与月报。
  8. 国内和海外双市场并行:国内覆盖 智谱GLM/豆包/DeepSeek/Kimi/MiniMax/纳米AI/百度AI,
  9. 海外覆盖 Gemini/ChatGPT/Claude/Grok/Perplexity,问题库与指标按市场分开算。
  10. 当用户说「做 GEO」「生成式引擎优化」「让 AI 推荐我的产品」「AI 搜索里搜不到我们」
  11. 「GEO 方案/诊断/监测/月报/交付」「给客户做 GEO 服务」,或给一个产品网址要做
  12. AI 可见性优化时使用。GEO 指生成式引擎优化,不是地理信息;
  13. 不用于传统 SEO 关键词排名、竞价投放或建站。
  14. ---
  15. # GEO 全流程服务
  16. 从**诊断**到**交付**的完整链路,不只是给建议:
  17. ```
  18. 抓取 → 体检 → AI答案采样 → 生成工单 → 产出资产 → 报告 → 自动验收 → 客户交付包
  19. ```
  20. 核心判断只有一句:**GEO 的终点不是排名,是 AI 答案里那句话是不是按你的口径说的。**
  21. 所以最小单位不是页面,是**可被抽取的事实块**。
  22. 项目目录:`/Users/brucejan/geo`,每个客户/产品一个 `work/<slug>/`。
  23. ---
  24. ## 最快的用法:给一个网址
  25. ```bash
  26. python3 scripts/geo.py new --url https://example.com --market both
  27. ```
  28. 九步全自动:建项目 → 抓官网 → 体检 → **自动推导品牌事实/竞品/问题库** →
  29. 重跑体检 → AI 答案采样 → 工单与建设蓝图 → 资产与报告 → 三份交付物。
  30. 产出 `work/<slug>/deliverables/`:
  31. | 文件 | 回答什么 |
  32. |---|---|
  33. | `1-GEO诊断报告.html` | 现在什么样 |
  34. | `2-GEO优化方案.html` | 应该改成什么样、为什么 |
  35. | `3-GEO执行方案.html` | 谁在什么时候做什么、做到什么算完成 |
  36. **自动推导出来的品牌事实必须人工复核**——`bootstrap` 只从官网正文抽取,
  37. 抽不到的一律标「待确认」,绝不用常识填充。标了「待确认」的字段
  38. (成立时间、工商主体、可具名客户等)需要你补齐或明确不对外说。
  39. ## 两种用法
  40. ### 界面(推荐)
  41. ```bash
  42. python3 scripts/geo.py ui # 默认 http://127.0.0.1:8765(前台运行)
  43. ./scripts/service.sh install # 或注册为 macOS 常驻服务:登录自启、崩溃自动拉起,
  44. # 关终端不停;周期复跑因此能按时触发。uninstall 卸载
  45. ```
  46. **全流程都在界面上**,不用记命令:新建项目、编辑配置与问题库、写事实卡、
  47. 一键跑任意步骤或整个周期(带实时日志)、改工单状态、看资产、导入人工采样表、打开交付包。
  48. 界面为暗色四段主线(AtlasGEO):**现状**(总览 / 引擎表现 / 竞品对比 / 问题库)→ **诊断**(差距诊断 / 阵地地图 / 品牌事实库)→ **提升**(行动计划 / 内容工作台)→ **成效**(效果验收 / 报告与交付),外加设置与三步接入引导。统一口径:GEO 健康分(五项加权,未测项权重归一)、提及率、引用份额、阵地、任务、内容。总览标题是由数据自动生成的一句结论;所有数字来自同一份采样,算不出的显示「未测」,不编数。
  49. 任务在后台子进程跑,关掉页面也会继续;同一项目同时只允许一个任务,避免抢同一份 audit.json。
  50. 只用标准库起服务,前端零外部依赖,断网可用。
  51. ### 命令行
  52. ```bash
  53. python3 scripts/geo.py serve --slug <项目>
  54. ```
  55. 抓取 → 体检 → 采样 → 工单 → 资产 → 报告 → 验收上期 → 打包交付,全做完。
  56. 产出在 `work/<slug>/delivery/<日期>/`,可直接发客户。
  57. 首次接一个新客户走下面的步骤 0–2 做好底座,之后每期只跑这一条。
  58. ---
  59. ## 先读什么
  60. - **每次必读**:`references/method.md`(评分口径和所有判据的出处,脚本就是它的代码实现)
  61. - **排外部信源优先级时必读**:`references/cn-source-ranking.md`
  62. (CN-GEO 187,818 条引用实算榜。**官网只占 1.37%**,这个数字会改变资源分配)
  63. - 打国内平台:`references/cn-platforms.md` | 打海外:`references/global-platforms.md`
  64. - 写内容/改页面:`references/content-patterns.md`
  65. - 更深方法或原始数据:`references/sources.md`(含 GEORank / GEOFlow 选型说明)
  66. 不要凭经验给 GEO 建议。**每条建议都要能追到 `method.md` 里的某条实测数字**,
  67. 说不出依据的别写进方案。
  68. ---
  69. ## 步骤 0 · 建项目
  70. ```bash
  71. python3 scripts/geo.py init --url <产品官网> --name <品牌名> --market both
  72. ```
  73. `--market` 取 `cn` / `global` / `both`。
  74. **双市场不是"顺带也做海外"**,是两套并行的战场——问题库、内容、竞品清单、指标全部分开。
  75. 纪律见 `global-platforms.md` 第 5 节。
  76. ## 步骤 1 · 吃透产品,建事实底座
  77. 读官网 + 用户给的材料(PDF/PPT/文档/公众号文章都行),补全 `geo.json`:
  78. - `brand`:规范名、**别名和常见错写**(品牌消歧的地基)、产品线、行业、目标用户、业务目标
  79. - `competitors`:3–6 个真实竞品,带别名。**首期可以留空**——先跑一轮"推荐类"问题的采样,
  80. 从 AI 答案里反推真实竞争集,比主观拍脑袋准得多。双市场时国内外竞品通常不是同一批,都要列全
  81. (排名指标是相对这个清单算的,漏掉真实对手会**高估**名次)
  82. 同时写 `work/<slug>/content/facts.md` 品牌事实卡,模板见 `content-patterns.md` 第 1 节。
  83. **每条事实标证据等级 A–E,没来源的标"待确认",不许编。**
  84. 这份文件是后面所有资产生成的输入——`llms.txt`、JSON-LD、定义块都从它来。
  85. 材料不够就直接问用户要,别猜产品能力和价格。
  86. ## 步骤 2 · 建问题库
  87. 问题库决定采样什么、写什么内容、怎么算指标。按七组各出 3–6 题:
  88. 推荐 / 比较 / 替代 / 价格 / 风险 / 品牌验证 / 场景。
  89. **每题必须标 `market`**,脚本据此路由——中文题不会打到 Perplexity,英文题不会打到豆包:
  90. ```json
  91. {"id": "q001", "group": "推荐", "market": "cn", "text": "国内做私域运营的 SaaS 有哪些好用的?"}
  92. {"id": "q101", "group": "推荐", "market": "global", "text": "What's the best CRM for small B2B teams in 2026?"}
  93. {"id": "q900", "group": "品牌验证", "market": "both", "text": "<品牌> 是家什么公司?"}
  94. ```
  95. - `cn`:中文口语问法,不要翻译腔 | `global`:**英文原生问法**,不是机翻
  96. - `both` 只留给品牌验证类。注意这类问题**点名了品牌**,答案必然复述品牌名,
  97. 脚本会把它们单独归入「品牌认知」,不混进可见性指标
  98. ---
  99. ## 无自有网站的项目(电商商品、线下品牌、小程序…)
  100. 没有官网也能做 GEO——AI 可见性本来就主要靠外部阵地,官网只占国内引用的 1.37%:
  101. ```bash
  102. python3 scripts/geo.py init --no-site --name "商品名" --materials 介绍材料.md --market cn
  103. ```
  104. 材料文本取代官网正文成为推导底座(不给 `--materials` 会生成模板让你填)。
  105. 差异只有三处:抓取/体检自动跳过(技术层不适用)、llms.txt 与 JSON-LD 不产出
  106. (需挂自有域名)、**引用官网率显示「不适用」而不是 0**。采样、竞品、阵地地图、
  107. 问题库、内容工作台、发布、验收、交付全部照常。
  108. ## 每期循环(`serve` 自动做完,这里说明每步在干什么)
  109. ### 抓取 + 体检
  110. 按六维打分:可抓取性 / 长度 / 结构 / **可抽取块** / 权威信号 / 对题性。
  111. `audit.json` 的 `layers` 字段按「访问 → 定向 → 理解 → 可引用」四层给出修复顺序:
  112. 每层依赖上一层,**先修失败的最上游层**(详见 method.md 四层依赖模型)。
  113. 读结果时先看五件事:
  114. 1. **`layers` 里最上游的 fail 层**——访问层失败时,下游的一切优化在引擎侧不可见
  115. 2. **SPA 空壳页**(`word_count` 接近 0)——国内官网最常见致命伤,AI 抓取器看到的是空白
  116. 3. **robots / WAF 有没有拦 AI 抓取器**——`ai_bots_blocked` 是 robots 封禁(含通配符组),
  117. `ai_ua_blocked` 是换 AI 爬虫 UA 实测被 WAF/CDN 拒绝(浏览器里看不出来的那种)
  118. 4. **`language_coverage` 中英是否对等**——做海外却没英文原生页直接 P0
  119. 5. **`block_gap` 缺得最多的块**——内容工程第一优先级
  120. ### AI 答案采样
  121. | 市场 | 平台 | 变量 | 说明 |
  122. |---|---|---|---|
  123. | 国内 | 智谱GLM | `ZHIPUAI_API_KEY` | OpenAI 兼容端点,不联网 |
  124. | 国内 | 豆包 | `ARK_API_KEY` | 火山方舟;**联网要在控制台单独开通内容插件**,没开通自动降级 |
  125. | 国内 | DeepSeek | `DEEPSEEK_API_KEY` | 不联网,测模型参数化知识里的品牌认知 |
  126. | 国内 | Kimi | `MOONSHOT_API_KEY` | 默认不联网 |
  127. | 国内 | MiniMax | `MINIMAX_API_KEY` | OpenAI 兼容端点,不联网 |
  128. | 海外 | Gemini | `GEMINI_API_KEY` | OpenAI 兼容端点,不带 grounding |
  129. | 海外 | OpenAI(ChatGPT) | `OPENAI_API_KEY` | Chat Completions 默认不联网 |
  130. | 海外 | Claude | `ANTHROPIC_API_KEY` | Anthropic 原生协议,不联网 |
  131. | 海外 | Grok | `XAI_API_KEY` | xAI API,不联网 |
  132. | 海外 | Perplexity | `PERPLEXITY_API_KEY` | 原生联网并返回 citations,海外证据质量最好 |
  133. Key 放项目根目录 `.env`(已 gitignore,权限 600),脚本自动加载。
  134. 没有公开联网 API 的——国内纳米AI搜索/百度AI/豆包 App,海外 ChatGPT 网页版/
  135. Claude 网页版/Google AI Overviews——走人工或浏览器采样:
  136. ```bash
  137. python3 scripts/geo.py sample-sheet --slug <项目> # 全量采样表
  138. python3 scripts/geo.py sample-sheet --slug <项目> --intent buyer --limit 20 # 每周买家题轻量核查
  139. python3 scripts/geo.py sample-import --slug <项目> --file <采样表>
  140. ```
  141. 更快的路径是配套的 **Chrome 采样助手插件**(`extension/`,README 有安装说明):
  142. 侧栏载入周检队列 → 填入问题 → 一键提取答案与引用 → 回传看板自动入库。
  143. 手动模式回车由人按(零合规风险,交付给客户用这个);「自动跑队列」需显式开启,
  144. 会代你发送,仅限自己账号小批量、人在场、撞验证码即停——README 里写明了风险边界。
  145. 数据只进本机 127.0.0.1,无后台采集、无遥测。
  146. **意图分组**贯穿三处:问题库页每组一张卡(该组提及率、失守题数、无内容题数),点卡片筛选;
  147. 插件按分组挑队列并分节排列;周检表 `--intent buyer` 取的也是同一批组。
  148. 分三类看——**买家意图**(推荐/比较/替代/价格,离成交最近,优先补)、
  149. **需求教育**(场景/风险)、**探测**(品牌验证,点名题不计入提及率)。
  150. 采样进来后到看板 **样本库**(现状 → 样本库)复核:机器判读是正则,品牌名撞词、
  151. 否定语境、竞品别名都可能误判。逐条纠正提及/位次/竞品,改完立刻重算当日指标,
  152. 纠正过的样本打 `manual_override`,重跑不会覆盖人工结论。
  153. **口径纪律**(详见 `method.md` 第 4 节):
  154. - **国内 ≠ 海外**,跨市场平均提及率没有解释力,报告里各占一张表
  155. - API ≠ 网页端,Web ≠ App(**千问 web 与 App 信源只有 24.5% 重合**),单独记单独算
  156. - 没有答案原文/截图/采样环境记录的,**不许标成"真实采样"**
  157. - 归因默认写"观察相关",除非有基线窗口 + 对照 Prompt + 竞品对照
  158. - 采样频率克制,遵守平台服务条款,不批量滥采、不模拟登录
  159. ### 生成工单 `plan`
  160. 诊断结果 → `tasks.json`,这是执行状态的**单一真相源**。每条工单必须有:
  161. 依据(追到 `method.md` 哪一条)、负责角色、工作量、时间窗口、**验收标准**、市场。
  162. 七个工作包:实体消歧 / 页面技术 / 内容矩阵 / 标题体系 / 知识库 / 外部证据 / 监测闭环。
  163. 优先级顺序(`method.md` 第 6 节):
  164. 门票问题 → 事实错误 → 抽取块缺口 → 高价值问题无承接 → **外部信源(P1,不是 P2)** → 长尾扩量。
  165. > 外部信源是 P1 不是 P2,因为**官网只占全库引用 1.37%**。
  166. > 官网是事实源不是引用源,把它从 60 分做到 90 分的边际收益,远低于拿下一个榜单站词条。
  167. ### 产出资产 `generate`
  168. 到 `work/<slug>/assets/`,中英分开:
  169. | 产物 | 说明 |
  170. |---|---|
  171. | `llms.txt` / `llms.en.txt` | 官方事实索引,直接传网站根目录 |
  172. | `jsonld/*.json` | Organization / SoftwareApplication / FAQPage / Article / BreadcrumbList,贴进 `<head>` |
  173. | `snippets/definition.*.html` | 定义块,放首屏口号下方(口号保留,不影响转化) |
  174. | `snippets/faq.*.html` | FAQ 块,**答案必须在静态 HTML 里可见**,不要纯 JS 折叠 |
  175. | `outlines/*.md` | 每个目标问题一份内容大纲(含标题候选、章节骨架、字数与抽取块要求) |
  176. | `drafts/*.md` | 加 `--draft` 时调用已配 LLM 出全文初稿 |
  177. **分工**:结构性资产由代码确定性生成(不会漏 schema 字段);**文章正文由你按 outline 写**——
  178. 代码写不出好文案。`--draft` 的初稿**必须人工核实事实后才能发布**。
  179. ### 自动验收 `verify`
  180. 重抓站点 → 跑 checker → 回写工单状态。**这是"服务"和"建议"的分界线**:
  181. 能自动验收的,就不靠人口头说做完了。做完会翻成 `done`,回归了会翻回 `todo`。
  182. 无法程序判定的(如百科词条是否过审)标「待人工」,确认后手动标记:
  183. ```bash
  184. python3 scripts/geo.py task --slug <项目> --id T-003 --status done --note "词条已上线"
  185. python3 scripts/geo.py status --slug <项目> # 进度看板
  186. ```
  187. ### 客户交付 `deliver`
  188. 打包到 `work/<slug>/delivery/<日期>/`:总览 index、诊断报告、执行方案、
  189. 工单表(HTML + CSV 可导进项目管理工具)、验收表、assets 目录、交付说明。
  190. 跑完之后你要做的事(脚本做不了的部分):
  191. 1. 读 `reports/latest.md` 看 delta:均分涨跌、提及率变化、新出现的 P0
  192. 2. **看「AI 实际引用的信源域名 Top」和「与全国大盘对照」**——这是当期真实的分发地图,
  193. 用它修正外部信源计划
  194. 3. 检查采样答案里有没有**事实错误**(品牌说错、张冠李戴、价格错、**归错行业**)→ 立刻 P0
  195. 4. 按 `outlines/` 写内容,按 `plan.md` 更新进度
  196. 建议节奏:**页面体检每周,答案采样每两周或每月**(采样有成本,指标本身有噪声,跑太密看不出信号)。
  197. ---
  198. ## 全部命令
  199. | 命令 | 作用 |
  200. |---|---|
  201. | `new` | **★ 只给一个网址,全自动出三份交付物** |
  202. | `autopilot` | 对已建好的项目跑完整引导流程 |
  203. | `bootstrap` | 从官网正文推导品牌事实、竞品、问题库 |
  204. | `deliverables` | 出三份正式交付物(诊断/优化/执行) |
  205. | `init` | 只建项目骨架,不跑流程 |
  206. | `crawl` / `audit` | 抓站 / 六维体检 |
  207. | `sample` / `sample-sheet` / `sample-import` | API 采样 / 导出人工采样表 / 回灌 |
  208. | `expand` | 拓词:百度下拉/Google 补全扩出真实需求候选题(`--no-llm` 用模板转写) |
  209. | `plan` | 诊断结果 → 带验收标准的工单 |
  210. | `generate` | 产出可部署资产(`--draft` 加 LLM 初稿) |
  211. | `report` | Markdown + 自包含 HTML 报告,含 delta 与大盘对照 |
  212. | `verify` | 重抓并自动验收工单(`--no-recrawl` 用现有结果) |
  213. | `task` / `status` | 单条工单状态 / 项目看板 |
  214. | `deliver` | 打包客户交付物 |
  215. | `ui` | **全流程界面**:新建项目、配置、问题库、事实卡、一键运行、工单、资产、交付 |
  216. | `serve` | **全流程一条命令** |
  217. | `cycle` | 轻量循环(抓取→体检→采样→报告,不含工单与交付) |
  218. | `list` | 所有项目 |
  219. ## 排期自动执行
  220. 用 `schedule` skill 建定时任务:
  221. > 每周一早上跑 `python3 /Users/brucejan/geo/scripts/geo.py serve --slug <项目>`,
  222. > 然后读 `work/<项目>/reports/latest.md`,如果出现新的 P0、提及率下降超过 10 个百分点,
  223. > 或有工单从 done 回归成 todo,就告诉我。
  224. ---
  225. ## 边界
  226. **做**:GEO 诊断、方案、工单、内容工程、资产生成、AI 答案监测、验收闭环、客户交付包。
  227. **不做**:
  228. - 传统 SEO 关键词排名、外链买卖、竞价投放
  229. - 建站/改站的实际代码实施(给方案、给可直接用的片段,落地由客户开发做)
  230. - 绕过平台限制的批量采集、模拟登录、刷量
  231. - 编造品牌事实、客户案例、价格或资质——**查不到就标"待确认",问用户**
  232. **不承诺**任何平台一定会引用某个页面。GEO 提高的是概率,不是保证。
  233. 给客户写方案时这句话必须原样写进去。
  234. ---
  235. ## 目录
  236. ```text
  237. /Users/brucejan/geo/
  238. ├── SKILL.md
  239. ├── references/
  240. │ ├── method.md 评分口径与全部判据出处
  241. │ ├── cn-source-ranking.md 国内信源实测榜(CN-GEO 实算,含复算脚本)
  242. │ ├── cn-platforms.md 国内平台适配
  243. │ ├── global-platforms.md 海外平台适配
  244. │ ├── content-patterns.md 可抽取内容模板
  245. │ └── sources.md 资料索引(含 GEORank / GEOFlow 选型)
  246. ├── scripts/
  247. │ ├── geo.py CLI 总入口
  248. │ ├── crawl.py 抓站 audit.py 六维体检
  249. │ ├── sample.py 多平台采样 benchmark.py 与全国大盘对照
  250. │ ├── tasks.py 工单系统 generate.py 资产生成
  251. │ ├── verify.py 自动验收 deliver.py 客户交付包
  252. │ ├── bootstrap.py 自动推导底座 deliverables.py 三份交付物
  253. │ ├── blueprint.py 建设地图
  254. │ ├── dashboard.py 界面后端 ui.html 前端工作台
  255. │ ├── jobs.py 后台任务(子进程 + 实时日志)
  256. │ └── report.py 报告渲染 geolib.py 共用工具
  257. └── work/<slug>/
  258. ├── geo.json 品牌、竞品、问题库、平台、目标
  259. ├── content/facts.md 品牌事实卡(所有资产的输入)
  260. ├── evidence/ 抓取快照 audit.json 体检结果
  261. ├── samples/ 采样原始答案 metrics/ 每期指标
  262. ├── tasks.json 工单(执行状态单一真相源)
  263. ├── assets/ 可部署资产
  264. ├── reports/ 每期报告 + latest.md
  265. ├── verify/ 每期验收结果
  266. ├── delivery/<日期>/ 客户交付包
  267. ├── history/ 历史基线(算 delta)
  268. └── plan.md 30/60/90 方案
  269. ```
README.md · README · 247 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. # Geo**Look**
  3. **Open-source, self-hosted platform for end-to-end GEO implementation**
  4. For a specific project: status analysis → diagnosis → strategy → implementation tickets → execution → verification
  5. English · [简体中文](README.zh-CN.md) · [日本語](README.ja.md)
  6. ![License](https://img.shields.io/badge/license-MIT-9184d9) ![Python](https://img.shields.io/badge/python-3.9%2B-9184d9) ![Platform](https://img.shields.io/badge/platform-macOS%20%7C%20Linux-9184d9) ![Deps](https://img.shields.io/badge/deps-requests%20·%20bs4%20·%20lxml-9184d9)
  7. <a href="https://www.producthunt.com/products/geolook?embed=true&utm_source=badge-featured&utm_medium=badge&utm_campaign=badge-geolook" target="_blank"><img src="https://api.producthunt.com/widgets/embed-image/v1/featured.svg?post_id=1211264&theme=dark&t=1786200566986" alt="GeoLook - Open-source, self-hosted platform for end-to-end GEO | Product Hunt" width="250" height="54" /></a>
  8. ![Product demo](docs/demo.en.gif)
  9. 🌐 [Website geolook.cc](https://geolook.cc) · 🔍 [Live demo (read-only)](https://geolook.cc/demo/) · 📹 [HD demo video (mp4)](docs/demo.en.mp4) · 🖼 [All screenshots](docs/screenshots-en/)
  10. <sub>Mirror while DNS propagates: [geolook.cc](https://geolook.cc) · [demo](https://geolook.cc/demo/)</sub>
  11. </div>
  12. > GEO = Generative Engine Optimization: getting AI engines (ChatGPT, Perplexity, Gemini, DeepSeek, Doubao…) to **proactively mention and cite your brand** when answering user questions. Not geographic info, not classic SEO.
  13. ## 1. Problems it solves
  14. More and more users ask AI directly — "best tools for X", "X vs Y, which one". If your brand:
  15. | Problem | What GeoLook gives you |
  16. |---|---|
  17. | **AI never mentions you** — you're not in the candidate set for category questions | Samples real answers engine by engine; quantifies mention rate / rank / citation share; diagnoses "absent" vs "competitor-dominated" |
  18. | **You don't know why** — AI is a black box | 6-dimension site audit + gap diagnosis: uncrawlable pages? missing extraction blocks? absent from the channels AI actually cites? inconsistent messaging? |
  19. | **Advice never lands** — recommendations pile up, nobody executes or verifies | Generates implementation tickets with acceptance criteria; 86% auto-verifiable in the sample project (18/21) — "done" is measured, not claimed |
  20. | **Did the work even help?** | Per-question before/after across sampling rounds + task-level before/after |
  21. | **You deliver GEO as a service and packaging is painful** | One click produces diagnosis report, strategy, execution plan, ticket CSV, and acceptance sheet for clients |
  22. ## 2. Feature map
  23. Four stages plus operations, all in one self-hosted dashboard:
  24. **Status** — Engine performance across 17 engines (10 automated via API + 7 manual, incl. Google AI Overviews and Metaso): mention rate, rank, citation share, what each engine actually cites, **sample replay** of raw answers, suspected-negative flags; **brand mention distribution** (you vs. competitors, per engine and aggregated); competitor tables with each rival's strongest engine one click away — plus each rival's **citation-source mix** and a "channels they have that you don't" hit-list; a 7-category question bank with **intent-group cards** (buyer / educate / probe — see at a glance which intent class you're absent from) where every question gets a **diagnosis type** (suspected-negative > competitor-dominated > absent > low-ranked); a **sample library** where every AI answer's metadata is browsable and human-correctable (regex parsing misreads — name collisions, negation — get fixed here; corrections recompute metrics instantly and survive re-sampling), with per-sample **citation-source breakdown** (domain × count · share) and sampling-environment provenance (sandbox / incognito / dedicated profile / personal — personal auto-downgrades to "needs review").
  25. **Keyword mining** — expand the question bank from real search demand: Baidu suggest (CN) + Google autocomplete (Global) terms from brand/competitor/category roots (free public endpoints, no keys). Each round is snapshot-diffed to flag **rising demand** (affects topic ordering, never metrics); alternative/vs phrasings mined from competitor roots feed the Competitors page. Candidates only — adding to the bank is always a manual check.
  26. ![Engines](docs/screenshots-en/engines.png)
  27. ![Sample library](docs/screenshots-en/samples.png)
  28. **Diagnosis** — Site audit organized as a **four-layer dependency chain** (Access → Orientation → Understanding → Quotability: each layer depends on the one above, and a failing Access layer makes everything downstream invisible to engines — the fix order is computed for you). The Access layer goes well beyond robots checks: RFC 9309-compliant robots parsing (catches wildcard-group blocks, shared UA groups and specificity overrides that line-by-line regexes miss), **WAF/CDN differential probing with real AI-crawler UAs** (robots may allow GPTBot while your CDN 403s it — invisible from a browser), X-Robots-Tag header noindex, llms.txt link validation, hreflang coverage, sitemap index-pollution and duplicate title/content detection. **Passage-level quotability**: retrieval picks passages, not pages — pages with sections but zero independently quotable passages get flagged with a concrete fix. Plus gap diagnosis (content → channels → facts); a **channel map of 19 channels** weighted by real citation-corpus data; and a **brand facts library** as the single source of truth that llms.txt, JSON-LD and content drafts are generated from.
  29. ![Site audit · four-layer chain](docs/screenshots-en/siteaudit.png)
  30. ![Channel map](docs/screenshots-en/channels.png)
  31. **Action** — Structured tickets (rationale / owner / effort / window / acceptance criteria) carrying an independent **risk grade** (low-risk quick wins / observe with 7-14-28-day recheck / high-risk technical changes with backup & rollback discipline — priority says how important, risk says how careful), with "first-measured → current → target" progress bars and automatic reopening on regressions; tickets deep-link to the exact question they most need written; a **content workbench** (topic pool sorted by "not mentioned + no content", required extraction blocks and brand facts at hand, live citability pre-check, fabrication-risk lint for AI drafts, and a **distribution checklist** matching each piece to its target channels); **deploy assets** (llms.txt, JSON-LD, HTML snippets, plus an **AI-traffic attribution pack**: GA4 "AI engines" channel-group regex, server-log counting script, source-snapshot guidance — closing the loop from "cited" to "converts"); **publishing** with channels grouped General / China / Global — GitHub, WordPress drafts, WeChat OA drafts, webhook, plus real **X** (teaser tweet with auto-backlink to your latest long-form publish) and **Reddit** (full-markdown self-post) integrations, each with an in-dialog step-by-step credential guide; multi-channel checkbox publishing per article, and publish status synced back to the action plan, pending list and question bank. Platforms without a personally usable official API (Weibo, Xiaohongshu, Toutiao, Bilibili, LinkedIn, Facebook, Instagram) are deliberately **not** integrated — better absent than fake; the page says why and offers the webhook bridge instead. Publishing is always manually confirmed, article by article.
  32. ![Action plan](docs/screenshots-en/plan.png)
  33. ![Workbench](docs/screenshots-en/workbench.png)
  34. ![Publishing channels](docs/screenshots-en/publishing.png)
  35. **Results** — Per-question before/after (all / CN / global tabs), task-level before/after, verification history; boss-ready one-pager, execution plan, and a complete client delivery package (HTML + CSV).
  36. **Operations** — Scheduled full-cycle re-runs (every 7/14/30 days; register the dashboard as a **macOS standing service** with `scripts/service.sh install` — starts at login, restarts on crash, keeps running with every terminal closed, so scheduled re-runs actually fire), multi-brand with one-click switching, and a manual-sampling loop that feeds the same metrics: export a full sheet or a **weekly buyer-intent sheet** (`sample-sheet --intent buyer --limit 20`), or use the **Chrome sampling assistant** below.
  37. **Sampling assistant (Chrome extension, `extension/`)** — for the engines without APIs, which is where the real users are. Side panel loads your buyer-intent queue (grouped by intent), fills the question into the chat box (**you** press Enter in the default manual mode), extracts the finished answer with all citation links in one click, and posts samples back to your local dashboard as A-grade evidence — a weekly check drops from ~30 minutes to ~10. An explicit opt-in auto-run mode exists with hard guardrails (you stay present, rate-limited, capped, halts on any CAPTCHA/anti-bot signal — ToS risks documented honestly in its README). Ships with `sandbox.sh`: one command launches a disposable clean-browser sandbox (no history, no cookies, extension auto-loaded) or a persistent logged-in profile for engines that require accounts, so sampling hygiene ("what a stranger sees, not what AI thinks of you") is one command instead of a discipline.
  38. ## 3. How it differs from other GEO tools
  39. Most GEO products are **monitoring SaaS**: they show mention rates and rankings, charge monthly, and keep your data in their cloud. GeoLook is an **implementation platform**:
  40. | | Typical GEO monitoring SaaS | GeoLook |
  41. |---|---|---|
  42. | **Loop depth** | Monitor + advise | Monitor → diagnose → **tickets → assets → auto-verify → deliver** |
  43. | **Verification** | None (or manual check-off) | Programmatic: re-crawl + next sampling round decide; regressions reopen automatically |
  44. | **Metrics** | Black-box scores | Fully reproducible; a "where do these numbers come from" panel in the UI; unmeasured shows "unmeasured", never faked |
  45. | **Chinese market** | Mostly Western engines | First-class CN engine matrix (GLM / Doubao / DeepSeek / Kimi / MiniMax / Nano / Baidu AI) + CN channels calibrated on citation-corpus data (Baike, ranking sites, WeChat, Toutiao…); CN and Global measured separately |
  46. | **Scoring basis** | Heuristics | Anchored in public empirical data: 602 prompts / 21,143 citations / 187,818 deduplicated CN citations ([references/](references/)) |
  47. | **Data ownership** | Vendor cloud | **Everything on your machine** under `work/` (JSON/Markdown); `git init` is your backup |
  48. | **Cost** | Subscription | Free and open source; you only pay your own engine API sampling costs (can be zero — manual sampling works) |
  49. | **Deliverables** | Dashboard screenshots | Client-ready diagnosis report / strategy / execution plan / ticket CSV — built for agencies and consultants |
  50. Honest limits: single-machine tool, no accounts or team collaboration; sampling frequency and volume depend on your own API budget; "suspected negative" flags are leads for human review, not verdicts. These are deliberate design choices.
  51. ## 4. Deployment
  52. ### Requirements
  53. - macOS or Linux (Windows via WSL — the code uses `fcntl` file locks)
  54. - Python **3.9+**
  55. - Exactly three third-party packages: `requests`, `beautifulsoup4`, `lxml`
  56. ### Three steps
  57. ```bash
  58. # 1. Clone and install
  59. git clone https://github.com/aigclink/geolook.git
  60. cd geolook
  61. pip3 install requests beautifulsoup4 lxml
  62. # 2. Start the dashboard (opens your browser)
  63. python3 scripts/geo.py ui # → http://127.0.0.1:8765
  64. # macOS: ./scripts/service.sh install registers it as a standing service
  65. # (starts at login, restarts on crash, survives closing every terminal)
  66. # 3. (Optional) Configure engine API keys
  67. # A: in the dashboard — Settings → Engines & Keys → "Configure" (writes local .env)
  68. # B: cp .env.example .env and edit
  69. ```
  70. **Zero keys works too**: automated sampling is skipped; use the manual sampling sheet loop instead. Crawling, auditing, tickets and assets need no keys. One CN-capable key (e.g. DeepSeek/GLM) unlocks auto-derivation of the question bank / brand facts and AI first drafts.
  71. ### Remote / server deployment
  72. The server binds to `127.0.0.1` by default. Two ways to access it remotely:
  73. ```bash
  74. # Option A (recommended): SSH tunnel, no port exposed
  75. ssh -N -L 8765:127.0.0.1:8765 user@your-server # then open http://127.0.0.1:8765 locally
  76. # Option B: public bind + access token (both required — refuses to start without a token)
  77. export GEOLOOK_TOKEN=$(openssl rand -hex 16)
  78. export GEOLOOK_HOST=0.0.0.0
  79. python3 scripts/geo.py ui
  80. # Enter the token on first visit (or open http://server:8765/?token=TOKEN);
  81. # afterwards access is via HttpOnly cookie. API calls: X-Geolook-Token header.
  82. ```
  83. For public deployments put an HTTPS reverse proxy (nginx/caddy) in front — a token over plain HTTP can be intercepted. `.env` and `work/` contain secrets and project data — mind file permissions.
  84. ### Upgrading
  85. ```bash
  86. git pull # your data lives in work/ and .env, both gitignored
  87. ```
  88. ## 5. Usage
  89. ### Route A: fully automated (10–30 min)
  90. ```bash
  91. python3 scripts/geo.py new --url https://example.com --market both
  92. ```
  93. `--market` is `cn` / `global` / `both`. Nine steps run automatically: crawl → audit → derive facts/competitors/questions → sample every engine → tickets → assets → report → auto-verify → delivery package, landing in `work/<slug>/delivery/<date>/`.
  94. ### Route B: dashboard walkthrough (recommended first time)
  95. 1. **Onboard** — `python3 scripts/geo.py ui`, follow the 3-step wizard (URL + market), hit "create & auto-bootstrap". The first cycle runs in the background.
  96. 2. **Review the foundation** (important, ~10 min) — facts are extracted only from your site copy; unknowns are marked "unconfirmed". Check **Brand Facts** (fix wording, add aliases — missing aliases understate mention rate) and the **Question Bank** (do questions read like real user queries?).
  97. 3. **Status** — Overview for the one-line verdict and health score; **Engines** to drill into each engine and replay raw answers (log factual errors on the spot); **Competitors** to see each rival's strongest engine — that's the channel to go build.
  98. 4. **Diagnosis** — Site audit (click a missing block → its fixing ticket) → Gap diagnosis → Channel map (open any channel for its build plan).
  99. 5. **Execute** — Take tickets P0-first in **Action Plan** (click a title for why/how/acceptance); write in the **Workbench** (pre-check ≥ B, publish as final, then work through the distribution checklist); deploy **Assets** (llms.txt to site root, JSON-LD into `<head>`, snippets into templates — see DEPLOY.md).
  100. 6. **Verify** — Settings → "Auto-verify" re-crawls and judges tickets; after the next sampling round, check per-question before/after in **Verification**.
  101. 7. **Operate** — enable scheduled re-runs (7/14/30 days); generate monthly reports and client packages in **Reports & Delivery**.
  102. ### Products and brands without a website
  103. No site is fine — brand-owned sites account for only 1.37% of CN citations; AI visibility
  104. lives on external channels anyway:
  105. ```bash
  106. python3 scripts/geo.py init --no-site --name "Product" --materials brief.md --market cn
  107. ```
  108. Your brief replaces site copy as the basis for brand facts, competitors and the question
  109. bank (omit `--materials` and a template is generated for you). Only three things differ:
  110. crawl/audit are skipped, llms.txt and JSON-LD are not generated (they need your own domain),
  111. and **own-site citation rate shows "n/a" instead of 0**. Everything else runs as usual.
  112. ### Manual sampling for engines without APIs
  113. ```bash
  114. python3 scripts/geo.py sample-sheet --slug <project> # full sheet
  115. python3 scripts/geo.py sample-sheet --slug <project> --intent buyer --limit 20 # weekly buyer-intent check
  116. python3 scripts/geo.py sample-import --slug <project> --file <sheet>
  117. ```
  118. Faster: install the Chrome sampling assistant (`extension/README.md`) and launch a clean sandbox with `extension/sandbox.sh` — queue, one-click extraction and upload, ~10 minutes per weekly round. Review imported samples in the dashboard's **Samples** page (machine parsing is correctable there; corrections recompute metrics instantly).
  119. ### CLI cheat sheet
  120. | Command | Purpose |
  121. |---|---|
  122. | `new` / `serve` / `cycle` | Automated new project / full cycle / light loop |
  123. | `ui` | Full-workflow dashboard |
  124. | `bootstrap` / `crawl` / `audit` | Derive foundation / crawl / 6-dimension scoring |
  125. | `sample` / `sample-sheet` / `sample-import` | API sampling / manual sheets |
  126. | `plan` / `generate` / `lint` | Tickets / assets (`--draft` adds drafts) / fabrication-risk check |
  127. | `verify` / `report` / `deliverables` / `deliver` | Auto-verify / reports / formal deliverables / client package |
  128. | `publish` / `task` / `status` / `list` | Publish content / ticket status / project board / projects |
  129. Every command has `--help`.
  130. ## FAQ
  131. **Q: AI answers differ every time — how can sampling results be stable?**
  132. A single AI answer is inherently stochastic, so GeoLook **never reads a single answer** as a metric. Stability comes from four layers:
  133. 1. **Aggregation** — mention rate and friends are ratios over dozens of questions × multiple engines; per-question jitter averages out.
  134. 2. **Fixed variables** — each engine's sampling model is pinned (visible and changeable in Settings), the question bank is fixed, and the same set is reused across rounds; the only thing that changes is time.
  135. 3. **Repeat rounds** — `sample --repeat N` samples each question multiple times when your budget allows.
  136. 4. **Attribution discipline** — the UI explicitly labels single-round movement as observational ("a drop isn't necessarily a regression — sampling is noisy"); only two consecutive rounds moving the same way count as a trend, and ticket verification relies on deterministic signals (re-crawling the site) rather than sampling alone.
  137. Honestly put: it measures a **distribution**, not a fixed value — which is exactly what real users experience, since every real query to an AI engine draws a random sample too.
  138. ## Evidence-based scoring
  139. All six audit dimensions are anchored in public empirical data; `scripts/audit.py` implements [references/method.md](references/method.md):
  140. - High-impact pages average **1,943 words**; low scorers just 170 (11.4×)
  141. - Numbers **+61.6%**, definitions **+57.3%**, comparisons **+55.3%**, how-to **+41.2%** citation-probability lift
  142. - Pure Q&A formatting is **−5.7%** — looking like an FAQ doesn't help
  143. - Topical relevance is the strongest predictor (r = 0.432), above authority
  144. - Brand-owned sites get only **1.37%** of CN citations — your site is the fact source; external channels are the citation sources
  145. ## Design principles & security boundaries
  146. - **Single-machine, self-hosted**: stdlib `http.server` on 127.0.0.1; no DB, no accounts; data is plain files
  147. - **Never fabricate**: facts only from site copy; inventing competitor names is forbidden; AI drafts must pass lint + human review
  148. - **Verification is the product**: anything auto-verifiable never relies on someone saying "done"
  149. - **Publishing is always manual**: channel credentials in local `.env` (mode 600); every publish is an explicit click; WeChat/WordPress go to drafts only
  150. ## Claude Code integration (optional)
  151. This repo doubles as a Claude Code skill ([SKILL.md](SKILL.md)): drop it into your skills directory and tell Claude "do GEO for example.com". Claude is optional — every script is a plain CLI.
  152. ## Layout
  153. ```
  154. scripts/ All logic (geo.py CLI · dashboard.py server · ui.html single-page UI · service.sh)
  155. extension/ Chrome sampling assistant + sandbox.sh + its own e2e tests
  156. references/ Methodology: sampling discipline, content patterns, attribution, citation structures
  157. tests/ Unit tests
  158. work/<slug>/ Per-project data (gitignored, never leaves your machine)
  159. docs/ Screenshots and the 40-second demo video
  160. ```
  161. ## Acknowledgements
  162. - [@yaojingang](https://github.com/yaojingang)
  163. ## Contact
  164. Questions, ideas or collaboration — email [bingqiang2008@gmail.com](mailto:bingqiang2008@gmail.com), or open an [issue](https://github.com/aigclink/geolook/issues).
  165. ## License
  166. [MIT](LICENSE)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.