AI Summary
We ran one query through Google’s two AI answer surfaces on the same day, from the same United States session: the AI Overview and AI Mode. The query was “best AI search optimization tools.” Between them the two surfaces cited 22 distinct domains. Only 5 appeared in both. That is about 23 percent overlap. Ranking inside one of Google’s AI answers does not mean you appear in the other.
The test, in one screen
The AI Overview is the AI summary that sits at the top of a normal Google results page. AI Mode is Google’s dedicated conversational tab, a separate experience that returns a longer, multi-part answer. We searched the identical query in both, in one US English session, then counted the sources each surface cited in its own answer. This is the same head to head logic we used for featured snippets versus AI Overviews, applied to the two surfaces buyers now actually see.
AI Overview: 12 domains, almost all vendor round-ups
The AI Overview returned a tidy, categorized list and cited 12 domains: Semrush, AthenaHQ, SE Ranking, Whatagraph, Rankability, OneLittleWeb, Surfer SEO, Freddie Chatt, Veza Digital, 20North Marketing, Behind Rankings, and Apiary Digital. Every cited page was a “best AI SEO tools” listicle or a vendor blog. No forum. No video. The named winners were Semrush One, AthenaHQ, and SE Ranking.
AI Mode: 15 domains, and it leaned on Reddit and YouTube
AI Mode answered the same query with 15 cited domains and a visibly different mix: Profound, Reddit, YouTube, Mangools, Otterly.AI, Frase, Zapier, Siftly, eesel, and AIclicks, plus the five it shared with the AI Overview. Reddit appeared through eight separate links. YouTube appeared through creator review videos. AI Mode also wrote roughly twice as much, built its own comparison table inside the answer, and finished by asking a follow-up question about budget and team size. That conversational close is core to how AI Mode handles follow-up questions, and it is why the surface behaves less like a snippet and more like a chat.
The overlap was only 5 domains out of 22
Five domains were cited by both surfaces: Whatagraph, Rankability, OneLittleWeb, Surfer SEO, and Freddie Chatt. Everything else split. The table below puts the two answers side by side.
| Attribute | AI Overview | AI Mode |
|---|---|---|
| Domains cited | 12 | 15 |
| Shared with the other surface | 5 | 5 |
| Cited Reddit | No | Yes (8 links) |
| Cited YouTube | No | Yes |
| Answer length | About half | About 2x longer |
| Built-in comparison table | No | Yes |
| Asked a follow-up question | No | Yes |
| Top brands named | Semrush One, AthenaHQ, SE Ranking | Profound, Otterly.AI, Surfer, Rankability |
Why the two surfaces pull different sources
The split is structural, not random. Google describes AI Mode as using a query fan-out technique: it breaks your question into several related searches and synthesizes across many more results than a single query returns. That breadth is what reaches into forums, video, and niche tools. The AI Overview draws on a narrower retrieval tied closely to the top web results for the one query, and for a commercial “best X” search those top results are vendor listicles.
So the two surfaces reward different things. AI Mode rewards breadth of presence: community threads, video reviews, and mentions across many sites. It is the same pattern behind the YouTube, Reddit, and Wikipedia dominance we see across AI answers, and behind the citation behavior in our AI Mode study. The AI Overview rewards being inside the canonical round-ups and structuring your own pages as clean, extractable comparisons. For the official version of the latter, Google publishes its own guide to optimizing for AI features.
What this means for your GEO strategy
One query produced two answers with mostly different sources. Five takeaways follow directly.
- Track the two surfaces separately. A citation in the AI Overview is not evidence you appear in AI Mode. Measure each as its own visibility metric.
- For AI Mode, build community and video presence. Reddit threads and YouTube reviews fed the answer directly. Earned mentions move this surface more than polished pages do.
- For AI Overviews, get into the round-ups. Earn placement in the listicles the surface cites, and format your own pages as clear comparison templates AI can extract.
- Cover the adjacent sub-questions. AI Mode fans out into related queries, so pages that answer nearby questions get pulled in. This is core generative engine optimization.
- Expect different competitors in each surface. Your AI Mode rivals are not your AI Overview rivals. Map both with competitive intelligence built for AI search.
The headline is simple. Google now answers the same question two different ways, from two different source lists, on the same results page and one tab over. A GEO program that optimizes for only one of them, or measures them as if they were one, is reporting half the picture. Pair this with your share of voice tracking, watch the click-through impact of AI Overviews, and run the same audit for Perplexity to see how far the divergence goes.
Want to watch your presence in both surfaces? These are the tools that track AI Overviews and AI Mode.