Private Claude Chats Showed Up In Search Engine Results. A Missing Web Setting Is Drawing Scrutiny

Some conversations shared through Anthropic’s Claude chatbot briefly appeared in search engine results over the weekend, drawing attention to how AI platforms manage publicly accessible links and protect user privacy.

The issue involved Claude’s optional “shared chat” feature, which allows users to generate a public web link to a specific conversation so it can be viewed by others. A number of those shared pages became discoverable through search engines, raising concerns among users after examples circulated online.

The issue was first highlighted by users on Reddit and later examined by WIRED, which reported that some indexed conversations included requests for legal guidance, political advice and fictional role-playing. Several of the pages identified in early searches were no longer accessible by the time WIRED reviewed them, either because they had been deleted or removed from search results.

The incident did not involve private Claude accounts being hacked or internal systems being breached. Instead, it centered on publicly accessible URLs created through Claude’s sharing feature and how search engines handled those pages.

Anthropic has long instructed web crawlers not to index shared chat pages through its robots.txt file, a standard mechanism website operators use to tell automated crawlers which sections of a site should not be crawled.

However, search engine documentation shows that robots.txt alone does not always prevent a page from appearing in search results.

Both Google and Microsoft Bing advise website owners to add a separate “noindex” directive on individual pages if they do not want those pages indexed. The outlet reported that the Claude shared-chat pages it reviewed did not contain that additional noindex tag, even though both search engines describe it as an important signal for preventing indexing.

Google spokesperson Ned Adriance told the outlet that search engines index pages made publicly available by website owners and that the responsibility for implementing indexing controls rests with the site operator.

The incident is not the first time Anthropic’s shared links have drawn attention.

Forbes reported in 2025 that publicly shared Claude conversations could also appear in search results under certain circumstances. At the time, Anthropic said it relied on robots.txt instructions to discourage search engines from indexing shared chats. Security researchers noted, however, that robots.txt is not designed to guarantee that publicly accessible pages will never appear in search indexes.

The episode also highlights broader questions about how AI companies manage publicly shared content.

Many AI developers use robots.txt files not only to manage search engine crawling but also to signal whether AI training crawlers may access parts of their websites. WIRED noted that Anthropic, Meta and OpenAI all restrict competitors’ AI crawlers from accessing chatbot pages through robots.txt directives.

At the same time, companies including Google recommend combining robots.txt with page-level indexing controls when publishers want content excluded from search results.

Users who have created shared Claude links can review or remove them through Claude’s privacy settings, where shared conversations can be managed or deleted.