ChatGPT has overtaken Perplexity on one important measure of how AI answers show their work, even as Perplexity keeps a wide lead on another. New panel data separates the two measures clearly, and understanding the difference matters for anyone trying to get cited by AI engines.
Citation inclusion and citation density sound similar, but they measure different things. Inclusion asks whether an answer carries at least one source at all, while density asks how many citations an answer carries on average. A platform can lead on one and trail on the other, and in this panel that is exactly what happened.
Across 61,680 answers and 372,111 citations, ChatGPT attached at least one source to 93.9% of answers against Perplexity's 90.2%. Perplexity kept the density lead at 9.68 citations per answer against ChatGPT's 3.29, AI Overviews' 7.51, Gemini's 4.97, and Claude's 2.41. The same panel measured 28.94 query fan-outs per tracked question and a 77.9% overall citation-inclusion rate.
Inclusion is the gatekeeper metric. If an answer carries no source, no publisher gets cited, no matter how good their content is. ChatGPT's lead on this measure means a source-backed answer from ChatGPT is now slightly more likely to point somewhere than one from Perplexity, even though Perplexity answers typically point to far more places at once.
For publishers, that distinction changes how to think about the odds. Inclusion decides whether the citation game is even being played in a given answer, while density decides how crowded the field is once it is. Both matter, but they reward different kinds of content work, and chasing one without watching the other can give a misleading picture of progress.
Density is where Perplexity's approach stands out. At 9.68 citations per answer, a typical Perplexity answer spreads its sourcing across many publishers, which means more pages get a mention but each mention competes with more neighbours. ChatGPT's 3.29 citations per answer, by contrast, concentrates attention on fewer sources, so each citation may carry more weight even though fewer pages receive one.
AI Overviews at 7.51, Gemini at 4.97, and Claude at 2.41 sit between those poles, which underlines that there is no single house style for AI sourcing. Each engine makes its own trade-off between breadth and focus, and a publisher's chances depend on which trade-off a given engine is making for a given question.
The panel's 28.94 query fan-outs per tracked question is a reminder of how much work happens behind a single AI answer. One user question can trigger many underlying queries, each pulling from different sources, before the engine settles on what to say. That means the page that gets cited is not necessarily the page that best matches the user's wording, but the page that best matches one of the many expanded queries behind it.
For SEOs, the practical response is breadth of coverage rather than a single perfectly optimised page. Content that answers the question, the follow-up questions, and the variations a curious user might ask stands a better chance of surfacing somewhere in that fan-out, even if no single page can cover all of it.
Targeting ChatGPT means first clearing the inclusion bar: content that is factual, clearly sourced, and easy to quote gives the engine something it can attach to an answer. Targeting Perplexity means accepting a crowded field and standing out within it, with distinctive data, original analysis, or a clearer explanation than the many other pages in the same answer.
Above all, the panel argues against treating AI visibility as one number. Track inclusion and density separately, watch them by engine, and judge a citation strategy by whether it earns mentions in the engines your audience actually uses. A strategy that only counts total citations will miss the story this panel tells.
It also helps to treat any single panel as a snapshot rather than a permanent ranking of the engines. Platforms change how they source answers, publishers change what they publish, and the balance between inclusion and density can shift from one measurement to the next. The value of this panel is less in crowning a winner than in giving SEOs two separate lenses to watch over time, so that a change in either measure can be noticed early and investigated rather than discovered months later when traffic has already moved.
Sources: GeoAura, citing Foglift Research, October 30, 2026, with panel data dated October 24, 2026.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial