Your ‘Shared’ AI Chats Can Show Up in Google. How to Check.

What actually happened with shared AI chats?

Shared AI chat links turned out to be publicly searchable, not just privately shareable. Over the weekend of July 26, 2026, Reddit and X users reported that a site-specific Google search surfaced a list of shared Claude conversations. It later surfaced shared Artifacts too, the interactive documents and apps built inside Claude.

Some of the surfaced material reportedly included sensitive content, ranging from work notes and programming discussions to personal details, according to Fortune. By the start of the following week, the affected conversation pages had largely stopped appearing in Google results, though Axios reported shared Artifacts were still findable for a period after that.

It is worth being precise about scope, because the headlines were alarming. This affected only chats where a user had actively created a public share link. Conversations that were never shared were not exposed. The issue is about what “shared” turns out to mean, not about private chats leaking.

Is this a data breach?

No, and the distinction matters for understanding your actual risk. A data breach is an unauthorised intrusion that steals data you never chose to expose. This is different: the affected content was placed on a public URL by the user, through a feature designed to share it.

That is exactly where reasonable people disagree, and the debate is worth presenting fairly rather than settling. One view is that this is user error made likely by design: many people who click “share” expect to send a link to one person, not to publish a page the whole web can crawl. The other view is that the behaviour is the documented consequence of creating a public link, and that the platforms warn users the link is viewable by anyone who has it. Both positions have merit, and which one you hold shapes whether you see a product-design failure or a misunderstanding of a working feature.

What is not in dispute is the practical lesson. A public link is a public web page. Google’s position, stated in response to earlier versions of this issue, is that search engines do not decide which pages are public, and that site owners control crawling and indexing through standard directives. Whoever bears responsibility, the outcome for you is the same, so the useful response is to manage your own share links rather than to assign blame.

Why does “anyone with the link” mean anyone?

A share link is an ordinary web address, and web addresses can be found. When an AI platform generates a public share URL, it creates a page that lives on the open internet. The only thing protecting it is that the address is long and hard to guess.

That protection is weaker than it sounds. Once a link is posted anywhere a crawler can reach it, a search engine can discover the page and add it to its index. That includes a forum, a social post, a public document, or a message that later gets scraped. From there it is searchable by anyone, not only by the person you sent it to. The gap between “unlisted” and “private” is the whole problem, and it is easy to miss when a single click produces the link.

This is the same mechanism behind similar incidents at other AI services, which is why the fix is universal rather than specific. Treat any AI share link the way you would treat posting the content on a public web page, because functionally that is what it is.

Which AI platforms has this affected?

Shared-chat indexing is an industry pattern, not a single company’s problem, and that framing matters because the safe habit applies everywhere. The same underlying “share link becomes searchable” issue has surfaced across several major AI chatbots.

The consistency across platforms is the point. Each uses a share feature that mints a public URL, and each has seen those URLs become searchable. A similar Claude incident in 2025 saw an estimated 600 conversations indexed before the pages were removed, and a separate case exposed roughly 100,000 publicly shared ChatGPT conversations, so the pattern is neither new nor small. If you use more than one AI tool, the checking routine below applies to all of them. For a broader comparison of the major assistants, our Claude versus ChatGPT versus Gemini overview covers how they differ elsewhere.

How do you check and revoke your shared AI chats?

Do this now if you have ever used a share feature, even once. Revoking a link you no longer need costs nothing and closes the exposure. The exact path differs slightly by platform, but the location is the same idea in each: a privacy or sharing area in settings.

1. Open your privacy settings. In Claude, go to Settings, then Privacy, then Shared Chats. In ChatGPT, look under Settings for Data Controls and shared links. Each platform keeps a list of every conversation you have shared.
2. Review the full list. Read through every shared item, not just recent ones. People forget links they created months ago, and those are the ones most likely to have been indexed.
3. Unshare anything you do not need public. Revoking the link disables the public page. Keep only links you have a current, deliberate reason to keep live.
4. Prioritise anything sensitive. Deal first with any shared chat that contains personal details, work information, credentials, or anything about another person.
5. Check each AI tool you use. Repeat the process on every chatbot you have shared from, since this is not limited to one service.

If you are unsure whether you ever shared a particular conversation, unsharing it anyway does no harm. There is no downside to revoking a link you are not actively using.

What if a sensitive chat was already exposed?

Unsharing stops future discovery, but it does not instantly erase the past, and knowing that lets you act on the parts that still matter. A revoked link disables the page, yet a cached copy in a search engine or a web archive can linger until it refreshes, and anyone who already saved the URL may retain it.

So treat anything genuinely sensitive that was in an exposed chat as potentially seen, and respond to the content rather than the link. Change any password, key, or credential that appeared in a shared conversation. Rotate an API key that was visible. If financial details or another person’s information were involved, act as you would for any exposure of that data. Our guide to checking whether your email has been in a data breach is a useful companion for gauging what else about you is already public.

Stop here and take extra care if a shared chat contained another person’s private information, because the exposure is not only yours to manage. Let them know, so they can take their own protective steps. For anything involving payment credentials, contact the provider directly using the number on your card rather than any link.

How do you share AI chats safely from now on?

The feature is genuinely useful, so the goal is safer use rather than avoidance. A few habits keep sharing convenient without turning private thinking into a public page.

• Before sharing, assume the link could become public, and decide whether the content survives that test
• Strip sensitive details, such as names, credentials, and account numbers, out of a conversation before you generate a share link
• Revoke a share link as soon as its purpose is served, rather than leaving it live indefinitely
• Prefer a screenshot or a copied excerpt when you only need to show one part of a chat
• Keep a periodic habit of reviewing your shared-links list, the way you would review app permissions

The wider lesson sits alongside a familiar one: people increasingly treat AI chatbots as a private space and put real personal and business information into them. That makes the sharing controls more consequential than they look. Our overview of cybersecurity and data privacy covers the broader habits worth having, and if a leaked detail later fuels a scam email, our guide to the ShinyHunters sextortion scam explains how exposed data gets reused.

FAQ

Are my private AI chats visible in Google search?

No, only conversations you actively shared with a public link were affected, not your private chat history. If you never used a share feature, there is nothing exposed.

How do I stop my shared Claude chats from being public?

Open Settings, then Privacy, then Shared Chats, and unshare any conversation you do not want public, which disables its link. Review the whole list rather than just recent items, since older shared links are the ones most likely to have been indexed.

Is this a Claude or Anthropic security breach?

No, and the same issue has affected other AI platforms including ChatGPT and Grok. It is not an intrusion that stole data, but a case of user-created public share links becoming searchable rather than data taken without permission.

Does unsharing a chat remove it from Google immediately?

Not always instantly, because a cached version in a search engine or web archive can persist until it refreshes, and anyone who already saved the link may still have it. Treat anything sensitive that was exposed as potentially seen, and change any credentials that appeared in it.

Is it safe to share AI chats at all?

Yes, if you treat a share link as a public web page: remove sensitive details before sharing, and revoke the link once its purpose is served. For showing just one part of a conversation, a screenshot is often safer than a full public link.