How ChatGPT, Perplexity and Gemini Cite Differently
The engines do not behave the same way. One cites heavily, one barely cites at all, and the difference matters.
Most advice about AI visibility treats the engines as a single thing called AI. They behave differently enough that the distinction changes what is worth doing.
The citation maximalist
Google's AI Mode attaches the most sources by a wide margin, roughly 15 to 22 per answer depending on the subject. That is around three times what the chat assistants return.
More slots means a lower bar for inclusion. It also means each individual citation carries less weight, because it is one chip among twenty.
The citation minimalist
Perplexity sits at the other end. Despite being built around the idea of cited answers, it frequently produces a confident, named answer with no model citations attached at all.
That is worth knowing before you measure anything. A tool reporting your Perplexity citation rate is measuring something that varies enormously by query type.
In the middle
ChatGPT and Gemini return a smaller set of sources than Google AI Mode, which makes each one more consequential. ChatGPT's source mix also moved further than anyone else's in 2025, with forum content collapsing and professional publishers rising sharply.
Scale differs too. ChatGPT is past 800 million weekly actives. Gemini reaches something near 250 million monthly across its surfaces. Perplexity is around 25 million monthly. If you are prioritising, the audience sizes are not close.
What this means practically
Do not optimise for one engine. The overlap in what they trust is larger than the differences: all of them lean on established publishers with real search history.
Do not accept a citation report without knowing which engine, which queries and which week. A number with none of that attached is decoration.
And weight your attention toward the engines with the users. Perplexity is interesting and small; ChatGPT is where the volume is.
The honest caveat
These behaviours are product decisions, not laws of nature, and every one of them has changed at least once. The useful conclusion is not the specific figures. It is that the engines differ enough that a single AI visibility number is almost meaningless.
Getting started
Publisive places clients in the publications every one of these engines already draws from. We write the piece to your brief, you approve every word, and we are honest about what can and cannot be promised.
We write and place features in the publications our clients choose. About us
Frequently asked questions
01Do the AI engines cite sources the same way?
No. Google's AI Mode attaches the most by a wide margin. ChatGPT and Gemini return a smaller set, which makes each one count for more. Perplexity frequently produces a confident named answer with no model citations attached.
02Which engine should I prioritise?
The one with the users, which is ChatGPT by a wide margin at over 800 million weekly actives. Gemini is next at roughly 250 million monthly. Perplexity is interesting and comparatively small at around 25 million monthly.
03Should I optimise differently for each engine?
Mostly no. The overlap in what they trust is larger than the differences: all of them lean on established publishers with real search history. What differs is how many sources they show, not fundamentally which kind.
