Across roughly 76.7M AI Overviews, 957k ChatGPT prompts, and 953.5k Perplexity prompts, only 7 of the top 50 most-cited domains were shared by all three engines. That means 86% of top sources were unique to a single platform. There is no one authority list to win.
Everyone wants the same thing: a list of the websites that ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Overviews trust, so they can go earn a spot on it. The uncomfortable finding from the largest citation studies to date is that no such single list exists. Even for identical queries, the engines cite mostly different sources.
This report breaks down what the data actually shows, why the platforms diverge so sharply, and the one signal that does cut across all of them. Every figure here traces back to a named, published study.
Key Takeaways
- The platforms barely overlap. Of the top 50 most-cited domains, only 7 appear across ChatGPT, Perplexity, and Google AI Overviews, an 86% divergence.
- Each engine chooses differently. Retrieval method, licensing deals, and training bias mean Google AI Overviews, ChatGPT, and Perplexity surface different domains for the same query.
- Freshness is the shared signal. Across 17 million citations, AI assistants cite content that is 25.7% fresher than organic search results.
- ChatGPT skews newest. It cites content published 393 to 458 days more recently than Google's organic results.
- The strategy shifts. You cannot win one universal source list, so you optimize for each engine's mechanism and keep content current everywhere.
The Core Finding: The Platforms Barely Agree on Who to Trust
When Ahrefs compared the 50 most-mentioned websites across Google AI Overviews, ChatGPT, and Perplexity for June 2025, the overlap was strikingly small. Just seven domains made all three top-50 lists. Every other trusted source was specific to one or two engines. You can read the full Ahrefs source study for the complete breakdown.
That is an 86% divergence at the very top of each engine's citation graph. The analysis ran across approximately 76.7 million AI Overviews, 957,000 ChatGPT prompts, and 953,500 Perplexity prompts, so this is not a small-sample quirk. It is how the major engines behave at scale.
The takeaway is not that authority stopped mattering. Good content and strong domains still get cited more often. But each engine reaches its citations through a different mechanism, so the domains they surface diverge sharply, even when the question is identical.
Why Each Engine Cites Different Sources
The divergence comes down to three things: how each engine retrieves information, which publishers it has licensing deals with, and what its training data over-represents. Here is how the major engines behave.
Google AI Overviews
Pulls from the search index and leans on the same authoritative sites you already see in organic results. There is a heavy emphasis on health, finance, and encyclopedic knowledge, plus a clear bias toward Google-owned properties. Media sites, non-Google entertainment, and ecommerce are notably underweight.
ChatGPT
With browsing enabled, it searches, then filters toward trusted and licensed domains before citing. It leans on publishers and non-Google media, especially for news, entertainment, and sports, which reflects its licensing partnerships. It is comparatively thin on health and medicine.
Perplexity
Runs a real-time web search, ranks the results, quotes, and cites the best-matching source. It pulls from a broader international corpus, surfacing local and regional brands the other engines miss, while underweighting social platforms and traditional Western media.
Copilot, Gemini, and Grok
Each layers its own retrieval and partnership biases on top. Copilot draws on Bing's index; Gemini and Grok bring their own source preferences. The practical result is that a source that dominates in one ecosystem can be nearly invisible in another.
The One Signal That Cuts Across Every Engine: Freshness
If the domains differ, what do the engines share? A measurable preference for recency.
Analyzing 17 million cited URLs across seven platforms, Ahrefs found that AI assistants cite content that is 25.7% fresher than what appears in organic Google results. The average cited URL was 1,064 days old for AI assistants, versus 1,432 days for the organic SERP, a full year newer.
Broken out by platform, the pattern is consistent. Google AI Overviews cite the oldest content (averaging around 1,432 days), closely tracking organic search. Perplexity (1,166 days), Gemini (1,118 days), and Copilot (1,056 days) all skew fresher. ChatGPT cites the newest content of all, with an average of 958 to 1,023 days, which is 393 to 458 days more recent than Google's organic results.
In plain terms: every AI assistant except Google AI Overviews leans toward fresher content. So while you cannot win one universal source list, you can make your content eligible almost everywhere by keeping it current. Freshness is the single lever that moves across engines.
One caveat worth stating: the average cited page is still nearly three years old, so recency is a strong preference, not a magic switch. Low-quality content updated daily will not climb. The winning combination is genuinely useful content that is also kept current.
How to Earn Citations When Every Engine Is Different
If there is no universal authority list, the strategy changes from chasing one target to covering the mechanisms each engine rewards. Five steps.
1. Track citations per engine, not in aggregate
A single blended "AI visibility score" hides the divergence that this entire report is about. Monitor ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Overviews separately so you can see exactly where you are cited and where you are absent. (For a full walkthrough, see how to track AI search visibility in 2026.)
2. Match content to each engine's bias
Authority signals and structured data help with AI Overviews. Publisher-grade depth and originality help with ChatGPT. Clear, directly quotable passages help with Perplexity. The same page can be optimized to qualify across all three at once.
3. Keep content fresh on a rolling cadence
Freshness is the one signal that cuts across every engine, and it decays continuously rather than resetting after a single update. A steady republishing schedule beats a one-time refresh every time.
4. Build authority off-site, in the right places
Engines cross-reference sources before citing them. Consistent mentions on the publications and communities each engine trusts raise your eligibility where it counts, not just on your own domain.
5. Write answer-first, self-contained passages
Retrieval-based engines lift the passage that best answers the query. Lead each section with a direct, quotable answer that stands on its own, then support it, so any engine can cite it cleanly.
How Per-Engine Optimization Works in Practice
Covering each engine's mechanism, rather than chasing one list, is exactly how Lureon runs client programs. A few results from that approach:
- Lumanu increased AI-search conversions by 155% in 30 days.
- Ilunafriq grew AI-search impressions by 22% in three weeks.
- A B2B SaaS client saw AI citations jump 400% in 60 days after struggling with DIY methods for six months.
The method behind those numbers: track citations across ChatGPT, Claude, Perplexity, Gemini, and Grok separately, produce content built to each engine's citation mechanism, and keep it fresh on a rolling cadence of 8 articles and 4 DA40+ backlinks a month. Clients typically see their first AI citations within 30 days and 5x visibility by month three. You can see the full approach on our B2B SaaS optimization page.
Conclusion
The instinct to find "the list" of domains AI engines trust is understandable, but the data says the list fragments the moment you look across platforms. Seven shared domains out of fifty is not a list; it is a rounding error.
What works instead is unglamorous and durable: optimize for how each engine actually retrieves and cites, write passages built to be quoted, and keep everything current, because freshness is the one preference every engine shares. Do that consistently and you stop betting on a single list that does not exist, and start earning citations engine by engine.
Read Next:
- How to Track AI Search Visibility Across ChatGPT, Perplexity, Gemini, and Claude in 2026
- AI Search Optimization Tools for ChatGPT: Strategies & Top Picks 2026
- llms.txt Explained: How to Configure Your Site for ChatGPT, Claude, and Perplexity in 2026
FAQs:
1. Do ChatGPT, Perplexity, and Google AI Overviews cite the same websites?
Mostly no. In Ahrefs' analysis of the top 50 most-cited domains across roughly 76.7M AI Overviews, 957k ChatGPT prompts, and 953.5k Perplexity prompts, only 7 domains appeared in all three top-50 lists. That means 86% of top sources were specific to one or two engines, so there is no single authority list that wins across every AI search platform.
2. Why do different AI engines cite different sources?
Three reasons: retrieval method, licensing partnerships, and training-data bias. Google AI Overviews pull from the search index and favor established authority and Google-owned properties. ChatGPT filters toward trusted and licensed publishers. Perplexity runs real-time web search across a broader international corpus. Even for identical queries, these different mechanisms surface different domains.
3. Is there any factor that improves citations across all engines?
Yes: content freshness. Analyzing 17 million citations, Ahrefs found AI assistants cite content 25.7% fresher than organic search results on average, with ChatGPT citing content 393 to 458 days more recently than Google's organic results. Keeping content current is the one lever that raises eligibility across nearly every engine.
4. How should I optimize if the target keeps changing by platform?
Optimize for each engine's mechanism rather than a fixed list of sites. Track citations per platform, match content to each engine's bias (authority and schema for AI Overviews, publisher-grade depth for ChatGPT, quotable passages for Perplexity), maintain a rolling freshness cadence, build off-site authority, and write answer-first passages that any engine can lift cleanly.
5. Does domain authority still matter for AI citations?
Yes, but it is not sufficient on its own. Strong domains and good content still get cited more often, and engines cross-reference sources before citing them. The nuance is that a high-authority domain in one ecosystem can still be underrepresented in another, so authority has to be built in the specific places each target engine trusts.