AI answer engines barely cite the same sources — only about 1 in 9 domains is shared between ChatGPT and Perplexity, so “AI visibility” is a per-engine claim, not one number

What the two datasets measured

The primary sources are 5W's “The State of AI Citations 2026”, published May 2026, which synthesizes 12 separate citation studies and more than 680 million tracked citations spanning August 2024 through early 2026, and Profound's “AI Platform Citation Patterns”, an analysis of a 680-million-citation dataset from August 2024 to June 2025. They are two independent reads of the same question — when an AI engine answers, whose page does it point to? — and they land on the same shape: the engines are not drawing from one shared pool of trusted sources. They are drawing from mostly different ones.

How little the engines overlap

Read the first two rows together and the practical point falls out: roughly nine in ten domains that ChatGPT cites are not the domains Perplexity cites, and only about one AI citation in eight comes from a page currently ranking in Google's top-10. The last two rows sharpen it — AI Overviews almost always lean on the organic index (93.67% touch a top-10 result), yet the specific page they cite matches a page-1 organic URL only 4.5% of the time. Being in the neighborhood of the top of Google is common; being the exact page the answer cites is rare.

Each engine prefers a different kind of source

The divergence isn't only which URLs — it's which kind of source each engine trusts. In Profound's data, Wikipedia accounts for 47.9% of ChatGPT's top-10 source share, while Reddit accounts for 46.7% of Perplexity's. 5W puts Wikipedia at 18.4% of all AI citations across verticals — a reference-first tilt on some engines, a forum-first tilt on others. The content that earns a citation on one engine is often the wrong shape for another, which is why a single content play rarely wins them all.

Why this matters for AI visibility

• “We got you seen in AI search” is meaningless without naming the engine. At ~11% domain overlap between ChatGPT and Perplexity, a client cited by one is, more likely than not, absent from the other. A visibility claim that doesn't say which engine is measuring almost nothing.
• A page-1 ranking barely predicts an AI citation. Only ~12% of AI-cited URLs are in Google's top-10, and only 4.5% of AI Overview URLs match a page-1 organic result exactly. An SEO retainer that lifts a client to page one has not, by that fact, earned them a citation inside the answer — those are different outcomes measured on different surfaces.
• The content that wins a citation is engine-specific. Wikipedia-heavy on ChatGPT, Reddit-heavy on Perplexity: the same brief cannot be optimal for both. Whatever you publish for one engine's citation logic is a partial bet on the others.

What an agency should do with this

• Report AI visibility per engine, never as one number. Give a client a mention rate for ChatGPT, for Perplexity, for Gemini, for Google's AI surfaces — separately. A single blended “AI visibility score” hides exactly the fragmentation these datasets measured, and it will overstate coverage every time.
• Don't let an SEO deliverable quietly stand in for AEO. Ranking and being cited overlap in only about one case in eight. Price and report them as separate work, or you'll credit a page-1 win for an AI citation the client never earned.
• Baseline each client across engines before you price the retainer. The honest first deliverable is a per-engine measurement of where the client actually stands as a cited source today — because the answer is almost certainly “visible on some, invisible on others,” and that split is the whole reason to hire you.

The takeaway

The durable finding across both datasets is not any single percentage — it's that the AI answer layer is not one place. ChatGPT, Perplexity, Gemini and Google's AI surfaces cite mostly different domains, reward mostly different source types, and only loosely track the Google ranking an SEO team can move. That makes “AI visibility” a set of per-engine facts, and it makes the honest first step a measurement of where a business stands on each engine — before anyone promises to fix a number they haven't yet split apart.

You can run the free 60-second check on any business and see its mention rate across ChatGPT and Perplexity as separate numbers — the earned-citation surface, per engine (the $79 audit adds Gemini and Claude, 25 questions sampled three times each, because one run is a sample, not a score). If you run audits for clients, the $249 agency 5-pack runs the full repeated-sample audit across five businesses, white-labeled — the per-engine baseline you need before you price a fix on top of it.