GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531 —
CL-OP46 497 —
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531 —
CL-OP46 497 —
CL-OP48 490 -0.2%
← Back to feed

A Missing noindex Tag Let Google Index Claude Shared Conversations, Exposing Medical and Company Data

A Reddit user’s search query, site:claude.ai/share, returned hundreds of other people’s Claude conversations. The exposure was real: medical information, school materials, employee documents, and internal company content were accessible via Google search, without any login required.

The cause is a single missing line. Anthropic’s share feature generates public URLs for Claude conversations and Artifacts: interactive apps, dashboards, or documents built inside Claude. Free, Pro, and Max users can create these links and share them externally. When they did, those pages became discoverable by Google.

Why robots.txt Wasn’t Enough

Anthropic had blocked crawlers through robots.txt, which instructs well-behaved bots not to crawl claude.ai/share. That’s standard practice. The problem is that robots.txt is a crawl directive, not an index directive.

Google can still add a page to its index when it encounters a link to that page from outside the restricted domain: a tweet, a blog post, a forum. The page appears in results because Google learned of its existence from the external link, not by crawling claude.ai. To prevent indexing of externally-linked pages, each page needs a noindex meta tag in its HTML, which tells search engines to exclude it from results regardless of how they found it.

Anthropic’s share pages lacked that tag.

Scope of the Exposure

Indexed pages included medical records and internal company materials. The volume of exposed conversations is unquantified. Shared links are created voluntarily, and the exposure only affects conversations or Artifacts for which users explicitly generated a public URL and then shared that URL somewhere indexable.

Unshared conversations are private. The issue does not affect conversations that were never turned into share links. Anthropic says it does not submit share page directories or sitemaps to search engines, and that links only become discoverable after users distribute them externally.

Remediation

Anthropic has now added the noindex directive. The fix is one HTML tag. Users who created share links can review and revoke them under Settings > Privacy > Shared chats.

The structural issue is that a feature designed for “anyone with the link can view” conflated URL-based access control with search invisibility. The two are not equivalent. A URL that works only if you have it is not the same as a URL that won’t appear in search results. Fixing that gap requires page-level controls that were absent at launch.