Shared Claude chats briefly became searchable through search engines, according to reports described by THE DECODER. The issue centered on conversations shared from Anthropic's AI chatbot and exposed through ordinary search results.
The reported discovery matters because shared AI conversations can include material users may not expect to appear beyond a direct link. In this case, some publicly accessible chats reportedly included crypto keys and legal queries.
What users found
Users on Reddit found thousands of shared chats with Anthropic's AI chatbot by using a simple Google search: site:claude.ai/share. That search pattern focused on Claude's shared-chat pages and made indexed conversations easier to surface.
The affected pages were tied to the Share with link feature. A link-sharing tool can be useful when someone wants to send an AI exchange to another person, but the report shows how quickly that convenience can become a visibility problem when search engines include those pages in results.
The source says the conversations were visible to anyone while indexed. That does not require a viewer to be part of the original chat; it means the shared pages could be found through search results during the period when they were exposed.
The missing signal
The apparent technical issue was the absence of a noindex tag. A noindex tag tells search engines not to include a page in their results.
Without that instruction, shared pages can be treated like other web pages available to search engines. The source does not describe the internal cause beyond the apparent missing tag, so the safest conclusion is narrow: the shared Claude conversations were indexed, and the pages apparently lacked the signal that would have told search engines to leave them out.
The report also notes that OpenAI made the same mistake last year and later removed the feature. That comparison shows this is not only about one product. Any AI service that lets users publish or share conversations needs to decide whether those pages should be discoverable by search engines or limited to people who already have the link.
Why the exposure mattered
The sensitivity comes from the content people may place into AI chatbot conversations. The source specifically mentions crypto keys and legal queries among the publicly accessible chats.
Those examples point to two different kinds of risk. Crypto keys can be operationally sensitive. Legal queries can reveal personal or professional concerns that a user may have intended to keep within the context of the chatbot session or a private share.
The larger issue is expectation. A feature labeled around sharing by link can feel limited to the people who receive that link. Search engine indexing changes the practical audience, because a conversation can become discoverable by people who were never sent the link directly.
How search engines responded
Anthropic appears to have responded quickly, and the Google results disappeared soon after. That suggests the most visible part of the issue changed fast on Google.
The source says the chats stayed visible longer on Bing and Brave Search. That difference matters because removal from one search engine does not automatically mean removal everywhere. Users checking whether shared pages remain discoverable may need to consider more than Google.
The article does not state how many chats remained visible on Bing or Brave Search, or exactly how long they stayed available there. It only reports that the visibility lasted longer than it did on Google.
What Claude users can check
Users can review and manage their shared conversations in Claude under Settings -> Privacy -> Shared Chats. That is the key action named in the source.
For anyone who has used shared Claude chats, that settings area is the place to inspect what has been shared. The report does not say that every shared chat contained sensitive material, but it does show why reviewing shared conversations is a practical step.
The episode is a reminder that AI chat sharing is still web publishing when the result is a public page. If a conversation contains information that should not be broadly visible, the safest assumption from this report is that users should treat shared links with care and manage them through Claude's privacy controls.