Which sources do AI engines actually cite?

When ChatGPT, Gemini or Claude answer a buyer's question, they pull from a surprisingly concentrated set of sources. A few platforms show up again and again — and which ones differ meaningfully by engine. If you know where the answers come from, you know where you need to be. That gap — between the sources that win the answer and the ones you are absent from — is what a source-gap analysis exists to map.

The platforms that dominate

Across multiple independent vendor studies, the same handful of domains keep topping the citation charts: Reddit, YouTube, Wikipedia, and LinkedIn, with Forbes, G2 and review sites close behind for B2B and product queries (Peec AI, 30M sources, via Search Engine Land). These are correlational vendor datasets — credible and large, but snapshots — so read them as shape, not gospel.

YouTube being the most-cited domain in Google AI Overviews — roughly 20.9% of citations (Ahrefs, via Quattr) — surprises most marketers, who think of AI search as a text game. And on the U.S. consumer ChatGPT side, Wikipedia (~13%) and Reddit (~12%) together account for more than a quarter of all citations, while major outlets like the WSJ, NYT and Bloomberg did not appear in the top 20 at all (5W Research).

The mix differs by engine

This is the part that breaks one-size-fits-all GEO advice: the source mix is not the same across engines. You cannot optimize for 'AI search' as a single thing.

Format matters as much as domain

It is not only which sites get cited but which kind of page. 'Best X' listicles are the single most-cited page format for ChatGPT, at roughly 43.8% of cited page types (Ahrefs). That is why getting onto — and ranking well within — third-party 'Best [category] tools' roundups is often higher-leverage than another post on your own blog. Engines reach for the comparison page, not the brochure.

From 'who gets cited' to your source-gap

Knowing the universe of high-citation sources is step one. Step two is specific to you: for the exact prompts your buyers ask, which sources win the answer — and which of those are you absent from? That is the Source-Gap Analyzer. It collects the real citations for your prompts across all three engines, identifies the sources that get quoted again and again, and turns them into a prioritized 'get into these sources' playbook — so you start with the ones most likely to matter for your prompts, not an alphabetical list.

Because the mix differs by engine, the playbook is per-engine too. The Reddit thread that wins you a ChatGPT citation is not the YouTube video that wins the Google AI Overview. Start by seeing where you stand: run a free audit, or read how our engines and data work.