Thousands of shared Claude chats turned up in Google and Bing search

A Reddit user's discovery over the weekend revealed that thousands of Claude conversations shared via Anthropic's "share chat" feature had become searchable on Google and Bing, exposing private exchanges to anyone running a simple site:claude.ai/share query.

The feature lets users generate a public link to a Claude conversation or Artifact for anyone who has the URL. The problem was that many of those links lacked the noindex tags needed to keep search engines from crawling and surfacing them.

404 Media first reported the issue Monday morning, and by that afternoon Google searches using the method described in the original Reddit post no longer returned results, according to TechCrunch. Bing was slower to catch up: Wired found roughly 612 results still visible for the same search query well after Google's cache had cleared.

What was exposed

The scope of leaked material was substantial. Futurism reported finding a detailed medical report, clinical trial results listing patient names, internal company documents, employee performance reviews, and files containing children's names and phone numbers.

Wired separately identified chats covering political strategy advice, a Kansas attorney's questions about ethics obligations, and erotic role-play sessions. TechCrunch cited a Fortune report describing a chat labeled "shared by Anthropic" in which Claude generated erotica, output that runs against the company's own usage policy barring sexually explicit content. Exposed Artifacts, meanwhile, included source code and personal work notes, and some chats reportedly contained cryptocurrency keys.

Anthropic's response and the recurring pattern

Anthropic spokeswoman Amie Rotherham told TechCrunch the company does not submit chat directories or sitemaps to search engines, and that share links only get indexed if someone posts the URL somewhere crawlers can already see, such as a forum or social media post. Google spokesperson Ned Adriance added that indexing shared Claude pages is Anthropic's responsibility, since Google follows whatever directives a site owner sets.

That defense is complicated by history. Anthropic told Forbes back in September 2025, after a nearly identical exposure was first reported, that it relies on robots.txt to block crawlers, and its robots.txt has carried that instruction since at least that month. Both Google and Bing publish guidance stating that robots.txt alone is not sufficient and that noindex tags are the more reliable mechanism, since robots.txt only discourages crawling rather than blocking indexing outright.

The company has now faced this same disclosure twice in less than a year, echoing a nearly identical episode OpenAI went through in 2025, when shared ChatGPT links were found indexed on Google before the company pulled the sharing feature entirely.

For now, Anthropic has not announced plans to retire share links. Users worried about existing exposure can review and revoke shared conversations under Settings, then Privacy, then Shared Chats.