TL;DR
- Only about 21% of cited domains overlap between ChatGPT and Google AI Overviews — the two engines are pulling from largely different pools.
- URL-level overlap is lower still: roughly 30% on direct comparison queries and about 4% on open-ended discovery queries.
- ChatGPT leans on Wikipedia (cited in ~48% of responses); AI Overviews leans on Reddit (~21% of citations) and vendor pages (~45%).
- Even shared citations like Wikipedia mean different things — AI Overviews groups it with community sources, ChatGPT groups it with institutional references.
- Optimizing "for AI" as one channel is a mistake. Treat ChatGPT and AI Overviews as separate surfaces with separate source diets.
If you assume ChatGPT and Google AI Overviews are re-ranking the same web, you'll build the wrong GEO strategy. The two engines cite meaningfully different sources, weight community versus institutional content in opposite directions, and disagree on what "authoritative" looks like for the same query. Here's what the overlap actually is — and why the delta matters more than the intersection.
The headline number: ~21% domain overlap
Across a broad query set, only about 21.26% of cited domains appear in both ChatGPT answers and Google AI Overviews. That means roughly four out of five domains cited by one engine are absent from the other for the same prompts.
At the URL level, the picture is even more fragmented. On direct comparison queries ("X vs Y") — where you'd expect maximum convergence toward canonical reference pages — URL overlap sits near 30%. On open-ended discovery queries ("best tools for…", "how do I approach…"), URL overlap collapses to around 4%.
Brand overlap is somewhat higher because the same brand can be cited via different pages: ~38% on discovery queries, ~22% on problem-solving queries. So the brands you compete against on ChatGPT may partially resemble those on AI Overviews, but the pages doing the citing work are usually different.
An obvious hypothesis — that both engines pull from the same candidate pool and just rank differently — was tested and rejected. Even pooling all of ChatGPT's citations (primary and supporting) matched only about 25% of AI Overviews' primary sources. This isn't a re-ranking problem. It's a retrieval-and-selection problem, and the two systems are solving it with different priors.
Wikipedia vs Reddit: the source-diet inversion
The clearest structural difference is how the two engines treat community content versus curated reference content.
ChatGPT's diet skews toward reference and institutional sources:
- Wikipedia appears in roughly 47.9% of responses.
- Reddit appears in about 11.3% of responses.
- Community forums collectively account for around 25% of citations.
Google AI Overviews inverts that ratio:
- Reddit accounts for about 21% of citations.
- Wikipedia accounts for only about 5.7%.
- Vendor and competitor pages dominate at roughly 45% of citations.
- Community sources overall sit near 4% — a much narrower community footprint than the Reddit share alone suggests, because AI Overviews concentrates its community pull on Reddit specifically rather than spreading across forums.
This is consistent with Google's post-2024 partnership behavior around Reddit and its long-standing preference for surfacing merchant/vendor pages in commercial SERPs. ChatGPT, by contrast, appears to lean on sources that were heavily represented in its pretraining corpus and in retrieval-augmentation contexts where reference material is favored.
For GEO planning: if your only Reddit strategy is aimed at Perplexity, you're missing that AI Overviews is also heavily Reddit-dependent — while ChatGPT is not.
Citation volume and self-referencing differ too
Google AI Overviews cites more sources per answer than ChatGPT — roughly 16 sources on average vs. about 10 for ChatGPT. That alone changes the game: on AI Overviews, being one of 16 is more achievable but also less prominent per citation; on ChatGPT, each of the ~10 slots carries more weight.
AI Overviews also self-references (pointing back to a tool's or brand's own site) in about 20% of citations, versus about 15% for ChatGPT. Practically, that means owned domain pages carry more direct citation power in AI Overviews. In ChatGPT, third-party validation — reference sites, aggregators, and community endorsements — does proportionally more of the lifting.
Even shared citations signal different things
Both engines cite Wikipedia. That looks like common ground until you inspect the neighborhood of citations around it.
In Google AI Overviews, Wikipedia typically appears alongside Reddit threads, YouTube, and vendor pages — a mixed authority stack where community signals sit next to encyclopedic ones. In ChatGPT, Wikipedia usually shares space with government sites, academic references, and established publishers — a curated authority stack.
The same URL is being used as two different kinds of evidence. On AI Overviews, Wikipedia is one voice in a crowd. On ChatGPT, it's a spine.
The implication for content strategy: getting mentioned on a page that also gets cited by both engines does not guarantee equal treatment. The surrounding citation context tells you which authority signal you're actually inheriting.
What this means for a GEO program
A few pragmatic takeaways:
- Split your reporting. Track ChatGPT citations and AI Overviews citations as separate KPIs. Domain overlap of ~21% means one dashboard won't tell you both stories.
- Match tactics to the engine. Reddit presence (subreddit participation, AMA-style threads, well-answered questions) pays off disproportionately on AI Overviews. Reference-site placements, Wikipedia-adjacent citations, and institutional coverage pay off disproportionately on ChatGPT.
- Own the vendor page for AI Overviews. With ~20% self-referencing and ~45% vendor/competitor weighting, your own product and comparison pages are direct citation targets — treat them as answer surfaces, not brochures.
- Diversify by query type. Overlap is highest on comparison queries and lowest on open-ended discovery. Prioritize by which query class your buyers actually run.
- Don't assume "AI SEO" is one channel. It's at least two, with different source economies.
FAQ
Is citation overlap increasing over time as both engines mature?
Not obviously. As Google deepens its Reddit and vendor-page reliance and OpenAI keeps ChatGPT's retrieval biased toward reference sources, the structural gap in source preference appears to be persisting rather than closing. Overlap may drift, but the underlying priors are different.
If I only have budget for one, which should I optimize for?
Depends on your query mix. B2B software with heavy "vs" and "best" queries benefits more from AI Overviews optimization (vendor pages, Reddit, comparison content). Research-heavy or high-consideration categories — health, finance, education — often see more upside on ChatGPT via reference and institutional placement.
Does ranking in traditional Google search predict citation in AI Overviews?
Partially. Traditional ranking correlates with AI Overviews citation more than it does with ChatGPT citation, but it's not deterministic. AI Overviews frequently pulls Reddit threads and vendor pages that don't hold the top blue-link position, so top-10 ranking is helpful but not sufficient.
