The six-engine divergence index

There is no such thing as "the AI answer." Ask the same buying question across the major engines and you do not get one verdict with minor wording changes — you get genuinely different answers, built from different sources, naming a different set of brands. In an audit of 118,000 AI responses across ChatGPT, Perplexity, Google AI Mode and Claude, the engines shared only about 11% of their cited domains. Nearly nine in ten sources were unique to a single engine.

We call the gap between those answers the six-engine divergence index: a measure of how far apart ChatGPT, Perplexity, Claude, Gemini, Google AI and Reddit land on the same prompt. The higher the divergence, the more a brand's fate depends on which engine a buyer happens to ask — and the less any single "AI visibility" number can mean.

What is the divergence index?

The divergence index is how far apart the major AI engines land on the same prompt, measured by the overlap in the sources they cite and the brands they name. Low overlap means high divergence. It matters because high divergence means a brand can be the confident default answer in one engine and completely absent in another — for the exact same question.

Traditional search had one index and one rank to check. AI search has at least six engines, each reading its own slice of the web and applying its own trust rules. There is no shared leaderboard. "Where do we rank?" has quietly become "which engines name us, for which questions, this week?" — covered in how AI engines choose what to cite.

How far apart do the engines actually land?

Far enough that there is no reliable substitute for measuring each one. BrightEdge's analysis found pairwise top-100 overlap in the sources engines cite ranges from just 16% to 59% — a 43-point spread. Even Google's own two AI surfaces, AI Mode and AI Overviews, only share about 59% of their top sources. Citation volume for a single brand can vary by hundreds of times between platforms.

The practical translation: a company that dominates Perplexity's citations can be nearly invisible in ChatGPT, and content that earns a citation on one engine has no guaranteed standing on the next. A single optimization strategy does not carry across — which is the entire reason multi-engine tracking exists rather than one universal score.

Why do the engines diverge?

Because each engine reads a different library and trusts it differently. The clearest split is how much weight each places on authoritative institutional sources versus user-generated content. Gemini leans hardest on government, academic and institutional domains; Google's AI Overviews lean hardest on forums and UGC — almost the inverse mix for the same web.

That single difference cascades. An engine that trusts institutions will quote a standards body or a university; an engine that trusts experience will quote a Reddit thread — the dynamic we cover in why Reddit runs the AI answer. Perplexity adds a third axis by citing roughly three times more sources per answer than ChatGPT, widening its net. None of these are bugs; they are editorial choices baked into each model.

Six engines are not six copies of the same answer with different fonts. They are six editors, each with a different idea of who to trust.

Do they at least recommend the same brands?

More than they agree on sources — but not nearly enough to ignore the gap. This is the most important nuance in the data: engines pull from wildly different sources yet converge somewhat on which brands they name. Pairwise overlap in recommended brands runs 36-55%, versus 16-59% for sources. The shortlist is steadier than the citations behind it — but a third to a half of it still changes from engine to engine.

And that brand agreement is highly uneven by category. Where the market has obvious leaders, engines mostly agree; where trust is contested, they scatter.

The takeaway is double-edged. If you lead an established category, the engines may already agree on you — and your job is to defend that. If you compete in finance, healthcare, or any category without a settled leader, divergence is your opening: there is no consensus answer yet, and each engine is up for grabs.

Which engine is the real outlier?

On sourcing, Gemini stands apart for its institutional bias. But the sharpest divergence is between ChatGPT and Google. Profound's analysis of 650-plus ChatGPT queries against Google results found that for product questions, ChatGPT's preferred sources correlated about -0.98 with Google's rankings — a near-perfect inverse. The pages ChatGPT favored for commercial queries were essentially the ones Google buried.

This is why "we rank #1 on Google" tells you almost nothing about your standing in ChatGPT, and vice versa. The surfaces are not just different — for some query types they are opposed.

What does divergence mean for your brand?

It means you cannot optimize once and check one number. With sources overlapping as little as 16% and brand lists as little as 36%, every engine is a separate front. The shared foundations — clear structure, citable facts, third-party mentions — earn you the right to be considered everywhere; after that, each engine is won or lost on its own terms.

1. Stop chasing one score. There is no universal AI rank. Track mention and citation rate per engine, for your real buying questions.
2. Find your widest gaps. Divergence shows where you are strong on one engine and absent on another — those gaps are the fastest wins.
3. Match the engine's trust rule. Institutional engines want authoritative corroboration; experience-led engines want real user discussion. Feed each what it trusts.
4. Re-measure on a clock. Engines update independently and without notice, so the index moves. A snapshot ages fast.

Measuring all of this at once — the same prompts, scored across every engine — is exactly what AI brand monitoring is for, and the optimization side lives in answer engine optimization.

Key takeaways

• The same prompt gets different answers — engines share only ~11% of cited sources across 118,000 responses.
• Source overlap runs 16–59% and brand overlap 36–55%; the shortlist is steadier than the citations, but both diverge.
• Engines diverge because they trust different sources — Gemini institutional, AI Overviews UGC, ChatGPT nearly the inverse of Google.
• Agreement is high in settled categories and low in finance and healthcare — so divergence is both a risk to defend and an opening to win.

The six-engine divergence index is really one idea: AI visibility is plural. There is no single answer to be the answer to — there are six, they disagree, and they change. The only way to know where you stand is to measure each of them, on your questions, over time.