Field note
Some AI engines cite 14 sources per answer. Others, none.
I counted the sources six AI engines show. The spread ran from ~14 an answer down to none at all.
Everyone talks about "getting cited by AI" as if the engines behave alike. They don't. In an audit of one category, I asked six assistants the same set of questions and counted how many sources each actually surfaced. The spread was wide:
- ChatGPT — about 14 sources per answer, always showed some.
- Perplexity — about 13, and the widest variety (the most distinct domains overall).
- Claude — about 11.
- Copilot — about 8.
- Google AI Mode — about 7, but showed no visible sources at all in 15 of 25 answers.
- Gemini — about 5, the leanest.
Two things jump out. First, the citation surface is uneven by an order of magnitude: an engine that cites a dozen sources per answer is a completely different opportunity from one that cites a handful — or none. Second, Perplexity spread its citations across the most distinct sources, while Google AI Mode often showed none — so "get cited" means very different things depending on which engine you're trying to appear in.
The honest caveats, because they're the point of doing this properly: this is one company, one category, on one day, with web search on. The counts are the sources captured in the audit — a close proxy for what each engine showed, not a perfect one. Engines drift between sessions, so treat this as a snapshot, not a law. And an engine showing few sources may be presenting them differently, not ignoring them.
But the headline holds: don't assume the engines cite alike. They don't — and the ones that show their sources are where being a genuinely citeable source pays off first.
This is the kind of thing an audit surfaces — see also who AI cites and where AI looks.
New here? Start with the overview → · By Denes Csaszar · written in the open