Private Claude Chats Revealed in Google and Bing Search Results

What occurs between you and an AI chatbot isn’t always private. Just ask Google.

Recently, many were taken aback to learn that some conversations with Anthropic Claude were easily discoverable through web searches. This situation, initially highlighted by a Reddit user, revealed chats where individuals sought advice on political party affiliation, ethical self-reporting requirements for attorneys in Kansas, and engaged in erotic role play.

Claude allows users to distribute “snapshots” of their chats by generating a public URL linked to a specific chatbot dialogue. The reason some of these URLs were indexed by prominent search engines relates to the fundamental operations of websites and search engines, particularly when generative AI enters the equation.

Essentially, Anthropic directs web crawlers, like those employed by search engines such as Google and Bing, not to index chats that users opt to share publicly. They achieve this via a robots.txt file, which has traditionally served as a guideline for web scrapers regarding accessible parts of a site. Anthropic’s robots.txt has prohibited access to “shared” chats by web scrapers since at least September 2025, based on a snapshot from the Wayback Machine.

However, preventing pages from appearing in search engine results can be more challenging than expected.

Bing, which still displays “about 612 results” for the search term “site:claude.ai/share” at the time of this writing, notes in its technical documents that developers can use robots.txt to block its web crawlers—but also recommend adding a “noindex” tag to individual pages.

Some chats mentioned in the Reddit post had been removed by the time WIRED checked them, and results no longer appear on Google when searching with the “share” query that still functions on Bing. According to a developer guide from Google, it disregards robots.txt instructions if a page is linked from elsewhere on the internet and the page owner has not included a specific “noindex” HTML tag or a “x-robots-tag” in the page’s response header.

WIRED examined a selection of the exposed Claude chat pages and found that they lacked the “noindex” tag that both Bing and Google consider when determining whether to index a page.

Microsoft, the parent company of Bing, did not provide comments prior to publication. Anthropic has not responded to multiple requests for comment.

Google spokesperson Ned Adriance informed WIRED that it is Anthropic’s responsibility to manage the indexing of shared Claude chats. “Neither Google nor any other search engine can control which pages become public on the web, and these pages were indexed by numerous search engines,”
Adriance stated. “We provide site owners with clear options to decide whether their pages can be crawled or indexed, and we always honor those directives.”

Anthropic did not address inquiries about why the “noindex” tag was absent on the shared chat pages.

Last September, Anthropic faced criticism for a similar problem and informed Forbes that it utilizes robots.txt to instruct crawlers not to access shared chats. However, the Forbes article emphasized that there is no assurance this will prevent search engines from indexing certain web pages.

Even if robots.txt isn’t always effective in stopping pages from being indexed by search engines, AI labs continue to utilize it for various purposes. Many labs assure creators that their sites won’t be used for AI training as long as developers ensure they “disallow” specific crawlers in their robots.txt files, and the labs themselves adhere to this guidance as well.