Four reasons the answers differ
They source differently
Some engines lean on a live web search at the moment of the question. Others lean more on what the model already knows. An engine that searches will favour pages it can fetch and quote right now; one that does not will favour whatever was well represented in its training data.
They weight authority differently
One engine may lean heavily on a small number of high-authority publishers. Another may spread across many smaller specialist sources. The same page can be plainly good enough for one and invisible to the other.
Recency matters more to some than others
A question about pricing or a fast-moving category will pull recent sources on a search-backed engine, and potentially much older ones elsewhere.
They answer different surfaces
Google alone has two: AI Overviews above traditional results, and AI Mode as a dedicated conversational experience. They answer differently, and being present in one does not mean being present in the other.
What the variation tells you
This is the practical part. The pattern of where you appear is diagnostic.
Cited on search-backed engines, absent elsewhere usually means your page is fetchable and answerable but you have little established presence. The page is working; the reputation is not there yet.
Named everywhere but cited nowhere is a citability problem rather than an awareness one. The engines know you exist and are not using your page as a source. That is a page-structure question.
Absent on one engine and cited on the rest is worth investigating rather than fixing blindly. Look at which sources that engine used instead.
In a run we did on our own category in August 2026, one domain was cited sixteen times by OpenAI and by no other engine at all. That kind of single-engine dominance is not rare, and you would never see it measuring one engine.
Why measuring one engine is measuring almost nothing
If you check only ChatGPT, you are seeing one sourcing model, one authority weighting, one recency behaviour. A result there tells you very little about the other five.
It also makes it impossible to tell a page problem from a presence problem. Absent everywhere points at your page. Absent in one place points at that engine’s sourcing. You cannot distinguish those without the comparison.
A worked example
In a run across our own category in August 2026 — thirty questions, six engines, three passes — the same question set produced markedly different source pools per engine.
One domain was cited sixteen times by OpenAI and by no other engine at all. Another was strong on Google AI Mode and Google AI Overviews but barely present elsewhere. A third appeared across Gemini, AI Mode and Perplexity but not once on Claude.
None of those patterns would be visible from a single-engine check. And each implies something different: a source that only one engine trusts, a source that Google surfaces but others do not, and a source with broad but not universal acceptance.
Figures are from our own measured run. Your category will have its own pattern — the point is that the pattern exists and is worth reading, not that these particular numbers apply to you.
What to do with it
Do not write six versions of a page. Structure helps everywhere — a clear, self-contained passage that answers the question is easier for any engine to lift. What differs is emphasis, not the fundamentals.
Fix the page once, properly. Then re-measure across all six and see which engines moved. That tells you more about each engine’s behaviour than any amount of speculation.
One caution: a single pass proves nothing. Answers vary between runs, so a difference you see once may not survive a second look. Multiple passes are what separate a real pattern from a coincidence.
Common questions
Which engine should I care about most?
The ones your buyers use, which is usually more than one. If you must pick, look at where you are absent rather than where you are strong - the gap is where the work is.
Does the same page need different versions per engine?
No. Structure and answerability help across all of them. Per-engine differences are a matter of emphasis and authority, not of writing six pages.
Why does Google appear twice in engine lists?
AI Overviews and AI Mode are different surfaces that answer differently. Being cited in one does not mean being cited in the other, so they are measured separately.
Is the variation just randomness?
Partly - answers do vary between runs, which is why multiple passes matter. But consistent differences between engines across several passes reflect genuine differences in how they source answers.