The number that matters here is, unusually, one nobody has: how many private Claude conversations are sitting in Google's index right now. 404 Media first reported the exposure on the morning of July 27, after Reddit users spent the weekend running site:claude.ai/share searches and posting what surfaced. Futurism, which reviewed the results and chose not to republish links, lists the categories: a detailed medical report on a real patient, clinical trial results that included patient names, documents naming primary school children with their phone numbers, company documents marked internal-only, and employee reviews containing personal information. Outlets that reviewed the results describe the true scale as unknown.
The mechanism is a gap between what 'share' means to a user and what it means to a crawler. Creating a Claude share link produces a public URL with the warning that 'anyone with the link can view' the chat; it does not say the link can be indexed by search engines, and most users reasonably read 'anyone with the link' as 'only people I send it to.' The Artifacts publishing flow does carry a search-indexing warning, which makes the chat flow the weaker of the two. On the technical side, Anthropic's robots.txt blocks the share path and the pages send an X-Robots-Tag: none header, but as Search Engine Journal explains, that combination backfires: because Googlebot is disallowed from crawling the page, it never sees the noindex, and a blocked URL that is linked anywhere public can still enter the index as a URL-only listing.
Anthropic's response, via spokeswoman Amie Rotherham: 'We give people control over sharing their Claude conversations publicly, and in keeping with our privacy principles, we do not share chat directories or sitemaps with search engines like Google. These shareable links are not guessable or discoverable unless people choose to share them themselves. When someone shares a conversation, they are making that content publicly accessible, and like other public web content, it may be archived by third-party services.' Which is true, and beside the point: users did choose to share the links, with colleagues and in bug reports and on forums, and each of those public mentions was enough for Google. Google results for the share path were cleared by the afternoon of July 27, per TechCrunch and The Next Web; the links themselves still work wherever they were posted.
None of this is new, and that is the story. Forbes found roughly 600 Claude conversations indexed last year, and Anthropic scrubbed them; a researcher scraped about 100,000 shared ChatGPT links in 2025; Grok chats turned up in search the same way. The lesson the industry keeps re-teaching is that a share feature is a publish feature, whether or not the dialog says so, and the only numbering that would end the pattern, an actual count and a takedown date per incident, is the one nobody provides. If you have ever clicked 'share' on a Claude chat, the practical advice is short: assume it is public, assume it is indexed, and treat the delete button as the only real privacy control.