Most-cited domains in AI search

Headline pattern

All three figures above are Partial — reported from a vendor corpus (a stated 680 million citations) that is not independently public. Reproduced here as claims, not verified findings.

Source skew by engine

Why domain type, not domain authority, predicts citation

The pattern above is easiest to explain by retrieval architecture, not any property of the cited domains themselves. ChatGPT Search has historically leaned on a Bing-derived index, with a strong bias toward broad, structured reference content. Perplexity's own crawler and citation-first UI design appear to reward community-vetted, frequently-updated discussion. Google's AI Overviews can draw on YouTube and the Knowledge Graph, inputs the other two engines don't have native access to. None of this is primarily about which domains have the highest backlink profiles.

Domain type vs. entity type — a distinction worth making

It's worth separating two things that get blurred in casual discussion of this data. One is the type of domain being cited: encyclopedia, forum, video platform, brand site. The other is the type of entity the citation is about: a product, a person, a place, a concept.

The reported skew above describes domain type. It says relatively little about how citation patterns might differ by entity type. A product-comparison query and a historical-fact query might show very different domain-type distributions, even on the same engine. That breakdown isn't available in the vendor-reported figures used throughout this page. It's a candidate dimension for the Citation Index's own query-set stratification.

How little engines agree

An ~11% overlap means optimizing for one engine's citation pattern gives you very little assurance of showing up in another's. This is the single strongest argument for measuring AI-search visibility per-platform, rather than as one blended "AI visibility" number. That's consistent with the principles laid out in the measurement standard.

What this means for a brand site

A brand or business site is not Wikipedia or Reddit, and can't become either. The transferable implication isn't "publish on those platforms." It's that each engine rewards a different kind of source structure. ChatGPT rewards broad, neutral, well-organized reference content. Perplexity rewards specific, experience-based, frequently updated discussion. Google AI Overviews rewards content with a strong multimodal or video companion.

Matching your content's structure to the engine you most care about is a more tractable strategy than trying to out-rank an encyclopedia.

A worked mapping exercise for your own content

Here's a practical way to apply this pattern. Take a single piece of existing content. For each engine, ask what structural change would make it more consistent with that engine's apparent preference.

Take a hypothetical product-comparison article. Making it read more neutrally and comprehensively, less promotional, covering competitors fairly, moves it toward ChatGPT's encyclopedic lean. Adding a genuine discussion or Q&A section with specific user experiences moves it toward Perplexity's community lean. Pairing it with an embedded explainer video moves it toward AI Overviews' multimodal lean.

None of these changes are guaranteed to produce a citation. The underlying data is Partial throughout. But they represent the most defensible, evidence-consistent direction to experiment in, given what's currently known.

A quick audit for your own pages

Pick your most important page and ask three quick questions. Does it read like a neutral reference entry, or like a sales pitch? A sales-pitch tone works against you on ChatGPT specifically.

Does it include any real, specific, first-hand detail, a tested result, a genuine edge case, the kind of thing a Reddit thread would naturally contain? That's the property Perplexity leans toward. Does it have a companion video or visual explainer nearby? That's the property Google's AI Overviews lean toward. A page that's weak on all three isn't competing well on any engine.

Verification status

Every percentage on this page traces to one vendor-reported corpus that isn't independently public. All are consistently graded Partial throughout, following the standard set in the provenance audit. Once the Citation Index's own query set is live, this page gets rebuilt against a corpus we collected ourselves, one we can publish in full.