Google AI Mode and AI Overviews do not cite the same sources. Even though both are Google. Two 2026 studies put numbers on the gap: the two surfaces share only about 59% of their top-100 cited sources (BrightEdge, April 2026), and AI Mode pulls far less from classic search rankings than AI Overviews do. Roughly 19% of AI Mode citations came from the organic top-20 versus about 56% for AI Overviews (seoClarity). Yet the brands recommended stay remarkably steady across surfaces. If you assume a citation in one Google AI surface means a citation in the other, the data says you're wrong.
This piece lays out what the two studies actually measured, why the surfaces diverge, and what it changes about how you earn and track AI citations. Both are single-vendor analyses using different methods and dates, so we read the numbers as directional, not precise. The direction is well corroborated across them.
How different are the two Google surfaces, in numbers?
Here are the load-bearing figures, each with its source and date:
| Finding | Figure | Source (date) |
|---|---|---|
| AI Mode ↔ AI Overviews top-100 source overlap | ~59% (highest of any engine pair measured) | BrightEdge, AI Catalyst, 10 industries (24 Apr 2026) |
| Gemini ↔ AI Mode / Gemini ↔ AI Overviews overlap | 27% / 34% | BrightEdge (24 Apr 2026) |
| Range of source overlap across all engine pairs | 16-59% (a 43-point spread) | BrightEdge (24 Apr 2026) |
| Range of brand-recommendation overlap across engines | 36-55% (a 19-point spread) | BrightEdge (24 Apr 2026) |
| Share of AI Mode citations from the organic top-20 | ~19% | seoClarity, 1,000 transactional US queries, 12,011 citations (Sep 2025) |
| Queries where AI Mode cited ≥1 organic top-20 page | 81% (only 28% for top-10) | seoClarity (Sep 2025) |
| Share of AI Overview citations from the organic top-20 | ~56% | seoClarity, 5.1M citations across 362,000 AI Overviews (Oct 2025) |
The pattern is consistent across both vendors: the two Google surfaces are more alike than any other engine pair, yet still meaningfully different: and AI Mode is markedly less tethered to classic rankings than AI Overviews are.
Why do two Google products cite different pages?
Because they are different systems solving the question differently, even sharing an index. AI Overviews generate a summary block on the standard results page and still lean substantially on the pages that rank well for the query, which is why more than half of their citations trace to the organic top-20. AI Mode is a Gemini-powered conversational experience that decomposes a question into many sub-queries. Query fan-out, and retrieves sources for each branch. That fan-out reaches well beyond the top-10 blue links, so most of what it cites sits outside the classic rankings.
In seoClarity's data the individual-position pull is also weaker than for AI Overviews: the organic #1 result was cited about 25% of the time in AI Mode, #2 about 21%, #3 about 16%. A real but modest advantage, far from a guarantee. Ranking first buys you a strong chance, not a slot.
If sources differ, why are the same brands recommended?
This is the most important finding, and it resolves an apparent paradox. BrightEdge measured both layers and found they behave differently: source overlap between engines swung widely (16-59%), but brand-recommendation overlap stayed tight (36-55%). In its words, "source agreement between engines varies widely and inconsistently; brand agreement is consistently steady."
The engines disagree on which page to quote but agree on which brand to name, because a well-established entity is recognised across all of them, even when each surface reaches for a different URL as its citation. This maps exactly onto the two-lens split we describe in the brand-mention gap vs source gap audit: your mention strength travels across surfaces, your cited pages are decided per surface. It also underlines why entity strength is the durable lever. It is the thing that stays constant when the citation plumbing does not.
The exact page an engine cites is volatile and surface-specific. The brand it recommends is stable and entity-driven. Build for the second and you become resilient to the first.
What does this change about your strategy?
The fundamentals don't change. Reachable, extractable, corroborated, fresh pages win everywhere. What changes is how you measure and where you place your bets:
- Don't treat one Google citation as two. A page cited in AI Overviews may be absent from AI Mode's answer to the same question, and vice versa. Verify each surface.
- Classic SEO still carries into AI Overviews (56% of citations from the top-20), so ranking work is not wasted there. It carries less into AI Mode (19%). You can be cited by AI Mode without ranking on page one, and ranked without being cited.
- Win AI Mode by answering the fan-out, not just the head query. Self-contained, answer-first passages for the predictable sub-questions. That's the subject of our companion guide, how to get cited in Google AI Mode.
- Invest in the entity layer. It's what keeps you recommended even as the cited URLs shift between surfaces and over time.
- Measure each surface separately, over time. A single check on one surface is noise; the signal is the trend on each, tracked across repeated snapshots.
Tracking whether you're cited in AI Overviews, AI Mode, and every other engine, as separate lines that each move on their own. Is exactly what Buffy Intel is built to measure, so you can see which surface you're winning and which one is quietly leaving you out.