ChatGPT cites a very different web than Google ranks. LoudFace, in a study published November 5, 2026, measured 17,642 cited sources and found that only 4.1% of ChatGPT's cited sources sit on a domain ranking in Google's top 10. Perplexity overlapped more, at 9.6%, and AI Overviews more still, at 12.0%. Yet 34.7% of ChatGPT answers cite at least one top-10 domain somewhere in the answer. The picture that emerges is not that Google rankings are irrelevant to AI search, but that they are far from the whole story — and for ChatGPT in particular, the overlap is strikingly small.
The headline figure is the 4.1%. Of the 17,642 cited sources LoudFace measured, barely one in twenty-five sat on a domain that Google ranks in its top 10 for the query. That means the overwhelming majority of what ChatGPT quotes comes from pages that Google does not consider top-tier results for the same question. The model is drawing on a much wider pool than the first page of search results, and the pages it chooses are often ones that a traditional SEO would never see, because they never appear in the rankings being tracked.
This is a useful corrective to the assumption that AI visibility is simply a new label on old rankings. If the overlap were high, the playbook would be simple: rank in the top 10 and the citations follow. At 4.1%, that playbook breaks. A page can rank first on Google and still not be the one ChatGPT quotes, and a page that never appears in the top 10 can still be cited regularly. The two systems are selecting sources by different criteria, and optimising for one does not automatically optimise for the other.
ChatGPT is the outlier, but the gap exists across the board. Perplexity's cited sources overlap Google's top 10 at 9.6%, more than double ChatGPT's figure, and AI Overviews reach 12.0%. That ordering makes intuitive sense: AI Overviews are built on Google's own results, so their sources should overlap the most, while ChatGPT draws on its own retrieval and reasoning, so its sources overlap the least. Even the best figure, though, is only about one in eight — which means that across every major AI platform, most cited sources are not top-10 Google results.
There is an important nuance in the data. While only 4.1% of individual cited sources sit on a top-10 domain, 34.7% of ChatGPT answers cite at least one top-10 domain somewhere in the answer. In other words, a typical answer may lean mostly on lesser-ranked pages while still including one familiar, high-ranking source. Related context from Ahrefs points the same way: only about 12% of links cited by ChatGPT, Gemini and Copilot appear in Google's top 10 for the same prompt, with Perplexity the outlier at nearly 1 in 3. The pattern is consistent across studies — the top of Google is one ingredient in AI answers, not the recipe.
The practical lesson is to stop treating a top-10 ranking as a guarantee of AI visibility. It helps — more than a third of ChatGPT answers include at least one top-10 domain — but it is not the deciding factor, and a strategy built only on rankings will miss most of the citations on offer. The pages that get cited are the ones the model finds useful, clear and directly relevant to the prompt, whether or not Google has placed them on page one. That shifts the work toward content that answers precisely, earns mentions across the wider web, and is easy for a model to extract and quote.
For link builders, the study is encouraging rather than discouraging. Links and mentions across many domains are exactly what make a page visible to a system that looks beyond Google's top 10, and a broad, genuine footprint is worth more than a single high ranking. The job is no longer just to push a page up one results page; it is to make the page the obvious source for the question, wherever the model looks. LoudFace's numbers say that the web ChatGPT cites is bigger than the web Google ranks — and that is an opportunity for any site willing to be useful beyond the first page.
Sources: LoudFace, November 5, 2026, via GeoAura's AI search comparison, updated November 7, 2026.
Related reading on this site: Facebook Answers' 6 Million AI Pages Are Google's Scaled-Content Test, November Click Spikes in Search Console Are Bot Traffic, Not a Ranking Win, What Gets Cited by ChatGPT: Ahrefs' 1.4 Million-Prompt Study, AI Share of Voice Is Not Organic Rankings — and the Overlap Is Small, Google Can Temporarily Protect Important URLs From Falling Out of the Index, and 3 in 4 Homepages Lost Organic Clicks, Analysis of 110 Sites Finds.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial