Real case · Our own site

From 9 stale sitemap URLs to 338 live ones, in one day.

oldksports.com is a sports-industry community site operated by our own team. We ran it through the exact GeoSeen workflow we offer clients. Every fact below is publicly verifiable today - and we are equally open about what is not fixed yet.

Run a public scan on your site

What the scan and manual review found

A sitemap that never worked

The public /sitemap.xml was a stale static file with 9 URLs. The dynamic generator behind it had returned 500 since the day it was written - the static file silently masked the failure, so no article ever reached the sitemap.

Canonical hosts at war

index.html and the sitemap declared www, every page-level canonical declared the apex domain, and both hosts answered 200 with no redirect - splitting indexing signals in half.

A CDN blocking all AI crawlers

A CDN-managed robots.txt block was denying GPTBot, ClaudeBot, CCBot and others site-wide. The owner did not know: the rules lived in the CDN dashboard, not in the codebase.

What was done, and how it was verified

Fixes

Repaired the dynamic sitemap generator and removed the masking static file. Unified every signal to the apex host with a site-wide 301. Added Organization/WebSite JSON-LD and a real llms.txt. Turned off the CDN's managed AI-crawl block.

Acceptance criteria, all publicly checkable

www returns 301 to the apex host. /sitemap.xml serves 338 URLs including every article. /llms.txt returns text/plain. robots.txt no longer carries the CDN block. Anyone can curl these right now.

The honest boundary

  1. The public health score moved only from 50 to 52 - the remaining issues (JavaScript-only content, soft-404 behaviour) need an architecture-level prerendering decision, not a quick fix.
  2. Indexing impact takes weeks and must be verified in Search Console, not claimed in advance.
  3. We promise verifiable repairs, never rankings or traffic.
Why this case matters

The most damaging issues - a silently broken sitemap generator and a CDN blocking AI crawlers - were invisible to a score-only audit. Evidence-led review found them, fixed them, and left acceptance criteria anyone can re-check.

真实案例 · 我们自己的网站

网站页面清单从 9 条旧 URL 到 338 条,一天完成。

oldksports.com 是我们团队自己运营的体育行业社区站。我们用交付给客户的同一套 GeoSeen 流程修复了它。下面每一个事实今天都可以公开验证——没修好的部分我们也同样明说。

为你的网站运行公开扫描

扫描与人工复核发现了什么

一个从未工作过的页面清单

公开的 /sitemap.xml 是只有 9 条 URL 的静态旧文件。背后的动态生成器从写出来那天起就一直返回 500——静态文件悄悄掩盖了故障,导致所有文章从未进入页面清单。

互相打架的正式网址

index.html 和 sitemap 声明 www,所有页面级 canonical 却声明裸域,且两个主机都返回 200、互不跳转——收录信号被分裂成两半。

CDN 替网站屏蔽了全部 AI 爬虫

CDN 托管的 robots.txt 区块把 GPTBot、ClaudeBot、CCBot 等全站拒之门外。站长并不知情:规则在 CDN 后台,不在代码仓库里。

做了什么,以及怎么验收

修复动作

修好动态 sitemap 生成器并删除遮蔽它的静态文件;用整站 301 把所有信号统一到唯一主机;补上 Organization/WebSite 结构化数据和真实的 llms.txt;关闭 CDN 的 AI 爬虫托管封锁。

验收标准,全部公开可查

www 任意路径 301 到裸域;/sitemap.xml 返回 338 条 URL 且包含全部文章;/llms.txt 返回纯文本;robots.txt 不再有 CDN 注入区块。任何人现在就可以自行验证。

诚实的边界

  1. 公开健康度分数只从 50 提高到 52——剩余问题(内容只在 JS 里、Soft 404)需要预渲染这类架构级决策,不是快修能解决的。
  2. 收录变化需要数周时间,必须在 Search Console 里验证,不能提前宣称。
  3. 我们承诺可验证的修复,从不承诺排名或流量。
这个案例说明什么

伤害最大的两个问题——静默坏掉的 sitemap 生成器和 CDN 层的 AI 封锁——只看分数的审计工具根本发现不了。证据驱动的复核找到了它们、修好了它们,并留下了任何人都能复查的验收标准。