ChatGPT Chats Are Being Indexed by Google
TL;DR: ChatGPT's shared conversations briefly showed up in Google search results, then vanished within 24 hours after OpenAI deployed a four-step deindexing fix (noindex, canonical, URL Removal Tool, robots.txt). The privacy story made the headlines; the SEO mechanics stay relevant. This piece covers how Google found those URLs without inbound links, how to check your own share links, how to remove them, and how other AI assistants handle public conversations today.
The first time I saw a chatgpt.com/share URL ranking #14 against a client's blog post sitting at #11, I assumed Google would take weeks to react. I was wrong. Less than 24 hours later, OpenAI had patched the leak and the indexed chats were gone from Google. Bing took longer. Some non-English locales took longer still. The episode left a clean technical record of how fast a major brand can pull URLs out of the index.
Searchengineland's writeup of the incident framed it as a privacy story. TechCrunch's coverage went the same way. The angle this piece takes is the one a working SEO has to plan around: what the discovery mechanism reveals, what the cleanup playbook teaches, and what every other AI assistant is doing with its public share links right now.
What happened with ChatGPT's /share URLs
The factual sequence, reconstructed from monitoring logs and contemporaneous coverage:
- ChatGPT shipped the public "Share" feature with a
robots.txtdirective that explicitly allowed crawling of/share/underUser-agent: *. That made every shared conversation a normal indexable HTML page. - Google began indexing those pages. By the time SEO Twitter noticed, a meaningful slice were already ranking for long-tail queries. The first viral screenshot showed a shared conversation about CRM software ranking in the top 20.
- Within roughly 36 hours of the first public reports, OpenAI deployed a
<meta name="robots" content="noindex">tag site-wide on/share/, then a canonical pointing to the homepage, then (almost certainly) a bulk submission to Google's URL Removal Tool. - Google deindexed every
/shareURL. Thesite:chatgpt.com/sharequery returns zero results today. Bing still showed roughly a million/sharepages a week later, because OpenAI's cleanup never reached Bing Webmaster Tools.
That's the structure. The interesting parts are downstream from it.
Why this matters for personal privacy
Most people who clicked "Share" on a ChatGPT conversation were sending a single link to a single person: a colleague, a friend, sometimes a therapist. The shared URL was treated as a deep-link, not a publication. When those URLs started ranking, the conversations attached to them became open-web content available to anyone running the right query.
The visible content included the kind of material you'd expect from an LLM used as a thinking aid: medical questions, legal questions, salary negotiations, draft messages to estranged family, business plans tied to identifiable companies. Enough of it was sensitive that the privacy framing dominated the response. OpenAI's noindex rollout was almost certainly accelerated by that pressure.
If you have ever clicked Share on a ChatGPT conversation, assume someone could have indexed it. Open your shared links list and delete the ones you don't actively want public. The noindex tag covers existing shares going forward but does not scrub cached copies that may still exist outside Google.
Why this matters for SEO and AI search visibility
Three things happened during the 36 hours these URLs were indexed:
- Shared chats competed in real SERPs. I tracked three client domains where a
chatgpt.com/shareURL appeared in the same SERP as a long-form post the client had spent weeks producing. In one case the share URL ranked above the client. - Citations got created and then destroyed. A handful of AI assistants and news outlets had already cited specific
/shareURLs by the time the deindex landed. Every one of those citations now points at a noindexed page that carries zero ranking signal. - The freshness loophole got demonstrated and closed in public. If the loophole had stayed open for a week instead of a day, the long-tail SERPs would have been flooded with AI-generated content from one source, regardless of whether the content was good.
If your competitor's chat is ranking for a query you spent six months earning, that's the business problem. The privacy question is real, but the SEO question is what the board will ask.
Christopher Penn, quoted in Searchengineland's coverage
The privacy story explains why OpenAI moved quickly. The SEO story explains why you still need to monitor competing AI-platform URLs in your SERPs, because the next AI company will probably not move as quickly.
How /share pages got indexed in the first place
This part shows how Google discovers content with zero inbound link signals. Three mechanisms stacked:
robots.txtleft the door open. The original ChatGPTrobots.txtallowed crawling of/share/underUser-agent: *. That's a direct invitation for Googlebot to fetch, render, and treat each shared conversation as a normal HTML page. Almost certainly an oversight; OpenAI was focused on the sharing feature, not the SEO surface it was creating.- Google's hidden-URL discovery is broader than people think. Even without inbound links, Google can surface URLs through passive channels. Chrome's Safe Browsing design docs confirm visited URLs can be sampled for the crawl queue. The SEO community has long speculated about Gmail link previews, Workspace previews, Android intent tracking, and DNS lookups feeding the index. Treat any Google-owned channel that touches a URL as a plausible discovery vector.
- AI-generated content looks fresh to the freshness classifier. Every
/sharepage held text that was nowhere else on the open web. Combined with the allowed crawl and the high-authority domain, freshness ranking gave those pages immediate visibility.
robots.txt, Chrome telemetry, and a freshness-weighted classifier was enough.No internal links does not equal no discovery. If your site has any user-generated content surface (chat exports, shared dashboards, public profiles, generated reports, anything with a token in the URL), Google is probably finding more of it than you realise. Audit your robots and noindex headers on those surfaces before someone else notices.
How to check if your own shared links got indexed
The cleanest check is a site: query in Google scoped to the platform domain plus an identifying token. ChatGPT's format started the story; the pattern generalises:
- ChatGPT:
site:chatgpt.com/share <keyword>. Should return zero results today. If anything still appears, it is cached or being served by a non-Google engine; compare in Bing and DuckDuckGo. - Claude:
site:claude.ai/chat <keyword>. Anthropic's share format keeps URLs gated by default. - Perplexity:
site:perplexity.ai <keyword>. Perplexity intentionally indexes its public threads. - Gemini:
site:gemini.google.com/share <keyword>or the equivalentg.co/geminipath. Non-indexable by default.
For a broader audit, pull a Search Console export filtered to any AI-platform host you've pasted into a piece of content. If you see traffic landing on those pages from your own properties, you have a citation to update.
How to remove your own shared links from Google
Two paths, depending on whether you own the platform or are just a user.
If you are a user with a shared link still hanging around in cache:
- Open the conversation in ChatGPT, click the share menu, and delete or unshare the link. This removes it from your own list.
- For any cached version still appearing in Google, submit the URL to Google's Remove Outdated Content tool. This is the public-facing version of the URL Removal Tool and works for URLs you don't control.
- Check Bing, DuckDuckGo, and the Wayback Machine separately. The Remove Outdated Content tool is Google-only. Bing has its own equivalent inside Bing Webmaster Tools.
If you are running a site with publicly-accessible URLs that need to disappear quickly, the four-step playbook OpenAI used is the one to copy:
| # | Step | What it does | Why the order matters |
|---|---|---|---|
| 1 | Add <meta name="robots" content="noindex"> |
Tells Googlebot to keep crawling but drop the page from the index. | Fastest signal. Propagates on the next crawl, often within 12 hours on a high-authority domain. |
| 2 | Set <link rel="canonical" href="https://example.com/"> |
Consolidates residual ranking signals to a canonical destination. | Prevents canonicalised duplicates from re-appearing once the noindex has propagated. |
| 3 | Bulk-submit to Google's URL Removal Tool | Hides URLs from results within minutes for ~6 months while the permanent deindex propagates. | Buys the "make it disappear today" window. Run after the on-page signals are live, not before. |
| 4 | Update robots.txt to Disallow the path |
Stops crawl requests entirely. Reduces bandwidth and prevents new URLs entering the queue. | Last. Disallow before the noindex has propagated and you block Google from recrawling the page to see the noindex in the first place. |
robots.txt blocks the crawl that would otherwise let Google see the noindex.I've run a variant for client emergencies, most recently for a SaaS site that accidentally indexed 4,000 internal support tickets through a misconfigured help desk integration. The bulk of URLs were gone in 28 hours.
What this teaches us about AI-generated content discoverability
Three takeaways for anyone managing an SEO-sensitive content surface.
Discoverability is broader than your link graph. If a URL exists, is HTML, and is reachable without auth, assume Google can find it. Robots and noindex headers are now the primary defence, not link suppression.
Freshness flattens authority. A brand-new shared chat with zero domain history ranked above six-month-old long-form posts because the freshness classifier valued the novelty. The weighting is fine for legitimate news but it creates exploitable edges every time a new content surface comes online. Plan for the next AI platform to make the same mistake; plan for your own UGC surfaces to do it inadvertently.
Citations are a liability, not just an asset. AI Overviews, ChatGPT citations, and outbound references in your own posts can all point at content that disappears in 24 hours. An evergreen blog post can carry six dead citations within a quarter without you noticing. The AI Overview citation playbook covers the inbound side; the outbound side is the same problem in reverse. The AI Visibility Checker flags external URLs pointing at noindexed or removed AI-platform paths.
How four major AI assistants handle public conversations today
The cleanest defence against another ChatGPT-style incident is understanding what each major AI assistant does with share links by default. The picture as of May 2026:
| Platform | Default share visibility | Robots / noindex posture | Discoverable? | SEO surface to watch |
|---|---|---|---|---|
| ChatGPT (OpenAI) | Link-only; not listed publicly. | noindex site-wide on /share/ post-incident; robots.txt Disallows it. |
No (since the 2026 patch). | Cached citations still pointing at dead URLs. |
| Claude (Anthropic) | Link-only; explicit "publish to web" step required. | Published conversations carry no noindex but are gated behind the publish step. |
Only if the user opts in. | Published Claude conversations are intentional content. Treat as guest posts. |
| Perplexity | Public by design; every thread is a candidate for the public site. | Threads explicitly indexable; built as a search surface. | Yes, by default. | Active competitor in SERPs. Cites your content and can outrank it. |
| Gemini (Google) | Shared links via g.co/gemini or gemini.google.com/share; gated. |
noindex by default on share paths. |
No. | Watch if Google starts surfacing Gemini conversations as featured content. |
| Copilot (Microsoft) | Sharing gated through Microsoft accounts. | Disallowed in robots.txt for the share path. |
No. | Copilot citations feed Bing more than Google. |
| Grok (xAI) | Public share links default in some flows; user opt-out. | Mixed. Some share paths carry noindex, some don't. Behaviour has changed twice since launch. |
Sometimes. | The most volatile surface. Re-check every quarter. |
The pattern: Perplexity is the only major platform where public discoverability is part of the value proposition. Everyone else is closer to a private chat surface than a publication. Perplexity threads are now an active competing surface, in the same way Reddit threads became a SERP-feature target after the 2024 partnership. The other platforms are mostly a citations concern, not a direct ranking competitor.
What AI Overviews get wrong
The meta-mistake worth tracking happens upstream of all of this: Google's own AI Overviews kept citing chatgpt.com/share URLs for weeks after those URLs had been deindexed. I saw three different Overview cards in our monitoring set still listing share URLs as primary sources well into mid-2026.
The mechanism is straightforward. AI Overviews are generated from a grounding corpus that updates on a longer cycle than the live web index. A URL that gets deindexed today may continue to appear as an Overview citation for weeks because the model was grounded on a snapshot pre-dating the deindex. Google AI Overviews cause measurable drops in click-through partly for this reason: the citation graph drifts out of sync with the actual index.
Your AI Overview citations are a snapshot of the index from weeks or months ago, not the index of today. If a competitor's Overview citation has been deindexed, the Overview will not reflect that for a meaningful window. The same logic that makes your good content slow to appear makes it slow to disappear when the underlying URL changes.
References to the original /share URLs in AI Overviews, cached LLM training data, and third-party blog posts will keep surfacing for the rest of the year. Making your brand show up in ChatGPT and optimising for AI-search citations are both downstream: the goal is to be the source the AI grounds on, not a citation of a source that itself moves.
FAQ
Are ChatGPT shared conversations still being indexed by Google?
No. The site:chatgpt.com/share query returns zero results after OpenAI's 2026 patch. Bing showed a residual ~1 million /share URLs for several weeks after the Google deindex; the engines remained out of sync.
Can I make my own ChatGPT share link private?
Yes. Open the conversation in ChatGPT, click the share menu, and delete or unshare the link. For cached versions still appearing in search results, submit the URL via Google's Remove Outdated Content tool.
Why did Google deindex shared ChatGPT conversations so quickly?
The chatgpt.com domain has high crawl priority, OpenAI deployed all four deindex signals close to simultaneously, and Google has automated systems that prioritise quick action on thin or user-generated content surfaces at scale.
Does deindexing ChatGPT chats hurt SEO for sites that linked to them?
Marginally. Outbound links to a now-deindexed URL lose the equity they were passing, and the destination loses any citation value. If you cited a /share URL in a blog post, swap it for a screenshot or a properly-credited quote.
What should I do if a competitor's ChatGPT share page is ranking above my content?
Run the query first; don't trust last week's snapshot. If you find a competing share URL, file it via Google's Spam Report tool as thin or user-generated content. Then focus on your own page; Google re-evaluates all candidates when a page drops from the SERP.
Are Perplexity threads going to face the same problem?
Probably not in the same form. Perplexity is built as a public search surface, so its threads are intentionally indexable. Treat them as competing publications, not accidentally-exposed UGC.
How long do AI Overview citations of deindexed URLs persist?
Weeks to months. The AI Overview grounding corpus updates on a longer cycle than the live web index. There is no public tool to flush a specific URL from Overview citations; the workaround is to wait for the next grounding refresh.
The real takeaway
The ChatGPT /share story is one of the cleanest emergency-deindex case studies the open web has produced. The four-step playbook is the default to copy when your own site has a URL surface that needs to disappear quickly. Google may have cleaned up overnight, but Bing's residual /share URLs and AI Overviews' lagged citations are reminders that "fixed in one place" is rarely "fixed everywhere."
AI-platform content surfaces will keep getting indexed by accident as new sharing flows ship. Monitoring competing URLs in your SERPs now has to extend across every AI assistant your audience uses. SEOJuice monitors competing URLs across your most important keywords and flags new entrants. Free to try, no credit card.
Related reading:
- How to Make Your Brand Come Up in ChatGPT
- Optimising for Perplexity, ChatGPT Search and Google AI Mode in 2026
- Optimising for AI Overview Citations
- Google AI Overviews Cause Massive Drop in Search Clicks
- How LLMs Change SMO Meaning in Digital Strategy
- How to Get Your Lovable Site Indexed
- AI Visibility Checker
- SEOJuice Data & Research
Read More
no credit card required