Do AI search engines cite the same websites as Google?

How much do AI search and Google sources overlap?

Far less than most marketers assume: usually well under half of sources are shared. Zhang and colleagues (opens in a new tab) compared six AI search engines, including ChatGPT, Gemini, Perplexity and Google’s AI Mode, with Google and Bing. Only 38% of domains appeared on both sides, while 37% appeared only in the AI engines.

Google’s own products do not line up either. Grossman and colleagues (opens in a new tab) compared Google’s regular results, its AI Overviews (the AI summary at the top of Google’s results) and Gemini. On average, only 18% of the sources returned by either AI Overviews or the regular results were returned by both. We look at what that means for rankings in whether Google rankings lead to AI Overviews.

Huang and colleagues (opens in a new tab) found the same pattern for an OpenAI model with web search. Its 100 most-cited domains overlapped only 24% with AI Overviews and 25% with Google’s regular results. AI Overviews and the regular results, by contrast, shared 68% of their top domains.

Each figure counts something different, from whole domains to single pages, so they are not directly comparable. The direction is the same in all three. Zhang and colleagues suggest one reason: AI engines break a question into smaller searches instead of searching for it as typed.

Do pages that rank on Google get cited by AI assistants?

Some do, but they are a minority, and the share varies widely by assistant. In our study of AI citations and Google rankings, we compared 2,492 citations from four assistants with Google’s results for 80 US buyer questions. Of the pages ChatGPT cited, 8.3% ranked in Google’s top 10 for the question; for Claude it was 25.7%.

Most of the gap is not explained by ranking at all. 68.5% of ChatGPT’s citations were not in Google’s top 100 for the question, or for searches of the kind ChatGPT runs. Being a trusted site helps more than ranking one page: ChatGPT’s site-level overlap with Google, 22.9%, was nearly three times its page-level overlap. For one engine, see whether backlinks still matter for Perplexity.

Where AI engines and Google do share a domain, it often sits at the top of Google’s list. Zhang and colleagues found that the most common position for these shared domains was the first result: 14.53% of cases on Google and 23.27% on Bing.

What kinds of sites do AI engines prefer instead?

They favor third-party editorial and reference sources, and fewer social or brand-owned pages. Chen and colleagues (opens in a new tab) sorted sources into brand-owned, earned (independent publications and reviews) and social. For US queries, Google returned 45.4% earned sources, while AI search returned 72.7% earned and negligible social content.

Social platforms drop out almost entirely for some engines. Huang and colleagues found the OpenAI model with search drew 0.1% of its citations from social platforms, against 13.4% for Google’s regular results. AI engines also cite far fewer pages: a mean of 4.3 web addresses per answer in the Zhang study, against 10.3 for traditional search.

Age matters too. In our freshness study, assistants cited pages first published about half as long ago as Google’s top 10 for the same questions, a ratio of 0.50.

Do AI engines at least agree with each other?

No. In some tests they overlap with each other even less than with Google. In a small test on 6 June 2026, Tannenbaum (opens in a new tab) sent 15 commercial prompts to ChatGPT, Microsoft Copilot, Google and Perplexity. Of all cited web addresses, 96.4% appeared in only one engine. The author founded a company that sells AI visibility software, and 15 prompts is a narrow sample. We compare the engines in more depth in whether AI engines cite the same sources.

Our own comparison of AI Mode and AI Overviews found the same within Google. On searches where both cited sources, the two surfaces shared a mean of 13.9% of cited web addresses. Of the pages in Google’s organic top 10, the AI Overview cited 29.7% and AI Mode 16.3%.

What should you do about it?

Treat Google rankings and AI visibility as related but separate results, and measure each one. Concretely:

1. Track citations engine by engine. ChatGPT, Gemini, Perplexity, Claude, AI Overviews and AI Mode each draw on different sources, so one combined score hides real differences.
2. Map the third-party sites that AI engines cite in your category. Industry publications, review sites and reference pages carry more weight in AI answers than in Google’s results.
3. Strengthen your whole site, not just single ranking pages. Assistants often cite a different page from a site that ranks.
4. Keep core pages current. AI assistants lean toward recently published pages more than Google’s top 10 does.
5. Keep investing in search rankings. When AI engines and Google share a source, the first result is the most common match. Our explainer on how GEO differs from SEO sets out where the two overlap.

If you want a measured starting point, our generative engine optimization service begins with an engine-by-engine audit.

What does the research not tell us yet?

The studies agree on the direction, but several questions remain open.

• They measure overlap in different units (domains, pages, top-100 lists), so no single overlap figure applies everywhere.
• Most collections are one-off snapshots from US locations, and AI sources change from day to day.
• Several studies used developer versions of AI models rather than the consumer apps people use.
• None shows why an AI engine picks one source over another, or whether changing a page would change its citations.
• The evidence on agreement between AI engines comes partly from a 15-prompt vendor test and needs larger, independent replication.

Frequently asked questions

If I rank first on Google, will ChatGPT cite me?

Not reliably. In our test of 80 buyer questions, only 8.3% of the pages ChatGPT cited ranked in Google’s top 10, and 68.5% were outside Google’s top 100 for the question.

Do Google’s AI Overviews use the same sources as Google search?

Only partly. Grossman and colleagues found AI Overviews and Google’s regular results shared 18% of sources per query on average in December 2025.

Which AI assistant is closest to Google’s rankings?

Claude, among the four we tested: 25.7% of its citations ranked in Google’s top 10, against 8.3% for ChatGPT.

Can a Google ranking report measure AI visibility?

No. Across six AI engines, only 38% of cited domains also appeared in Google or Bing results, so most AI sources never show up in a ranking report.

Sources

• Zhang et al. (2025), Source Coverage and Citation Bias in LLM-based vs. Traditional Search Engines (opens in a new tab), arXiv:2512.09483.
• Grossman, Liu, Chen, Smith, Borcea and Chen (2026), How Generative AI Disrupts Search: An Empirical Study of Google Search, Gemini, and AI Overviews (opens in a new tab), arXiv:2604.27790.
• Huang, Goyal, Saha and Chandrasekharan (2026), Answer Bubbles: Information Exposure in AI-Mediated Search (opens in a new tab), arXiv:2603.16138.
• Chen, Wang, Chen and Koudas (2025), Generative Engine Optimization: How to Dominate AI Search (opens in a new tab), arXiv:2509.08919.
• Tannenbaum (2026), Scoring With the Engine: Retrieval Exposure, Cross-Engine Divergence, and the Limits of Engine-Agnostic GEO Scores (opens in a new tab), arXiv:2609.22655.
• Underneath (2026), Do ChatGPT, Gemini, Perplexity and Claude cite pages that rank?
• Underneath (2026), AI Mode vs AI Overviews: how different are the sources?
• Underneath (2026), How fresh are the pages AI engines cite?