Your chats with ChatGPT may end up on Google — what you should know.

Users who share chats with ChatGPT through the 'Share' button may unexpectedly make them visible to everyone in Google search results. Some of these pages may contain sensitive details — from stories about mental struggles to information that can identify a person.

This is reported by Fast Company.

How conversations end up in search and why it raises concerns

Google indexes any content available on the open web, so shared links to ChatGPT chats that users intentionally make public appear in search results. A simple search query with a fragment of the unique link generated when clicking 'Share' reveals dialogues where people share personal details — including about addictions, experiences of physical violence, or serious mental health issues. Although the ChatGPT interface does not display names, users sometimes reveal their identity through specific circumstances and facts of their lives.

By clicking 'Share', people often forward the conversation to close ones or simply save the URL for themselves — hardly expecting that the page would end up indexed by Google and become accessible to anyone. Currently, searches return almost 4,500 such conversations, and this is not necessarily a complete list, as Google does not index all pages. Given the sensitivity of the topics, examples of such chats are not provided nor linked.

The context makes the finding particularly alarming: according to surveys, nearly half of Americans have turned to large language models for psychological support over the past year. Three-quarters sought help with anxiety, two-thirds for advice on personal issues, and nearly six in ten regarding depression. Unlike the confidentiality in conversations with a therapist, transcripts of chats with AI may appear in regular search results.

Website owners can remove pages from retrieval, and shared ChatGPT links 'by default' are not indexed — they must be consciously made accessible for searching. Users are also warned not to post sensitive data and are allowed to remove shared links.

OpenAI CEO Sam Altman recently warned users not to share their most intimate details in ChatGPT: the company may be legally obliged to provide this data at the request of a court.

It should be noted that AI has already become part of everyday life, and ChatGPT entered the top five most popular websites in June. Despite the convenience, experts warn: some queries to chatbots may lead to safety or even financial issues.

We have also written that AI assistants help to instantly search for information, but at the same time pose a potential threat to privacy. Experts highlight seven types of data that should never be trusted to ChatGPT and similar systems to avoid becoming a victim of leaks or fraud.

The post emphasizes that sharing chats via ChatGPT may pose a risk of confidential information becoming available online through Google indexing. Such data may be particularly sensitive and contain private details, and thus users are advised to be cautious and avoid sharing their most intimate details through this platform. Furthermore, the public availability of such information may become a problem for users' privacy and safety.