What happens in conversations with AI chatbots may not always remain confidential. Recently, it was discovered that some private chats from Anthropic’s Claude were indexed by Google and Bing, becoming accessible through web searches. This issue, initially highlighted by a Reddit user, involved various discussions, including personal advice and even explicit role-play scenarios.
Claude permits users to share specific chat sessions as public URLs. Anthropic utilizes a robots.txt file to instruct web crawlers from search engines not to index these shared conversations. However, preventing pages from showing up in search results is more complex than it seems. While the robots.txt file has been in place since at least September 2025, it does not guarantee non-indexing.
Bing has acknowledged the presence of indexed content related to shared chats, and while it understands the role of robots.txt, it also points out the importance of including a "noindex" tag on individual pages. Google similarly notes that it may disregard robots.txt directives if the pages are linked from anywhere else on the web without the specific noindex tag.
Upon reviewing some of the exposed Claude pages, it was found that they lacked the needed noindex tag, raising concerns about their potential reappearance in search results. Google has clarified that the responsibility for the non-indexing lies with Anthropic; search engines only respect the directives given by site owners.
In previous instances, Anthropic faced criticism for similar indexing problems and had mentioned its reliance on robots.txt as a preventive measure. Although inadequate for this purpose, organizations like Anthropic, Meta, and OpenAI also use robots.txt to restrict competitor access to their sites.
Despite Claude’s messages no longer appearing in Google results, the absence of a noindex tag leaves the door open for future visibility. Users can manage their shared chat settings to safeguard their conversations from being publicly accessible.
