How ChatGPT, Perplexity, and AI Overviews Cite Sources Differently

“Optimize for AI search” sounds like one job. It isn’t. ChatGPT, Perplexity, and Google’s AI Overviews each build answers from a different set of sources, and the gap is enormous: when Ahrefs looked at the top 50 most-cited domains across the three, only seven appeared on all of them. That is about a 14% overlap. Treat them as one channel and you’re invisible on two of them.

I track this per engine at Geonimo, and the differences are consistent enough to plan around. Here is what each engine actually trusts, and how to earn citations on each, the practical layer beneath the broader AEO playbook.

Why you can’t treat AI search as one channel

Each engine has its own index, its own retrieval logic, and its own editorial biases about which sources are trustworthy. That is why the citation overlap is so low. A domain that dominates ChatGPT answers for your category can be nearly absent from AI Overviews, and vice versa. So the first strategic decision isn’t “how do I rank in AI,” it’s “which engines matter for my buyers, and what does each one reward.”

ChatGPT: publisher authority and Reddit

ChatGPT leans heavily on established publishers and high-authority media, Reddit, Wikipedia, Forbes, Business Insider, Wired, and similar, partly because of content licensing deals. The median domain rating of its top cited pages sits around 90, and its citations overlap Google’s top results only loosely (studies put it somewhere between 10% and 35%). The takeaway: your own page ranking #1 in Google does little here. What moves ChatGPT visibility is being mentioned on the publishers and communities it trusts. Earn coverage on authoritative sites and show up helpfully in the right Reddit threads.

Google AI Overviews: established authority, YouTube, Reddit

AI Overviews favor established, authoritative sources, health, finance, encyclopedic content, plus Google’s own properties (YouTube is roughly 5.6% of AIO citations) and Reddit. There used to be a tight link to classic rankings: around 76% of AIO citations came from pages in Google’s top 10. That has fallen to roughly 38%, as Overviews increasingly pull from YouTube and Reddit. So strong SEO still helps, but it’s no longer sufficient; video and community presence matter more every quarter.

Perplexity: basically Google with citations

Perplexity is the most aligned with traditional search. A large share of its citations come from pages already ranking in Google’s top results (estimates range from ~29% to ~70% overlap, far higher than ChatGPT). The practical implication is the happy one: if you already rank well in Google, you have a natural head start in Perplexity, and classic SEO is the most direct lever. It’s the easiest engine to win if your search foundations are solid.

Google AI Mode: not the same as AI Overviews

Here’s the trap people miss: AI Mode and AI Overviews are both Google, but they don’t cite the same sources. Their citation overlap is only about 13.7%, even though their answers are ~86% semantically similar, they say similar things from different sources. AI Mode’s top cited domain is YouTube by a wide margin, followed by Google and Wikipedia, and it leans on Quora (about 3.5x more than AIO) and social platforms more heavily. If AI Mode matters for your audience, video and Q&A presence deserve real weight.

How to prioritize without building four strategies

You don’t need a separate playbook per engine, the fundamentals, genuinely useful content, earned mentions, and topical authority, help everywhere. But two filters tell you where to spend first. One is market share: Google’s AI features and ChatGPT command the vast majority of AI-search traffic today, so they earn priority. The other is overlap with what you already do well: if you rank in Google, lean into Perplexity and AI Overviews; if you don’t, ChatGPT’s publisher-and-Reddit game may be the faster path. Knowing where each engine pulls from turns a vague “do AI SEO” into a concrete list of citations to earn.

Knowing how each engine cites is step one; the next two are measuring your share of voice on each and then earning the citations that move it. Per-engine tracking is exactly what I built Geonimo to do, see the rest of my work or get in touch.