How Google AI Mode picks sources (different from AI Overviews)
AI Mode and AI Overviews share only 13.7% of their citations. Here is how AI Mode selects sources, and why being cited in one does not get you the other.
Most operators still treat Google's AI answers as one thing. They are two, and they do not agree with each other. AI Overviews is the summary block at the top of a normal results page. AI Mode is the separate conversational surface that now has more than a billion monthly users per Google's own I/O numbers. When Ahrefs lined up the two for the same queries across 540,000 citation pairs, the same URL appeared in both only 13.7% of the time. You can be the cited source in AI Overviews and be completely absent from AI Mode for the identical question. Signals runs an aged Reddit account marketplace plus an editorial network for AI brand mentions across Reddit, Quora, Product Hunt, and Threads, and the brands we help track both surfaces because winning one does not hand you the other. This is how AI Mode actually chooses what to cite, where it diverges from AI Overviews, and what that split means for the work you do to get cited.
What is Google AI Mode, and how is it different from AI Overviews?
AI Mode is a separate conversational search surface, not the summary box. AI Overviews sits at the top of a standard results page and answers in a few sentences. AI Mode is its own tab where the user asks longer, multi-part questions and gets a longer synthesized response with its own citation set. Both now run on a custom build of Gemini, but they behave like different products.
The practical consequence is that they pull from different source graphs. Ahrefs found AI Mode responses run about 4x longer than AI Overviews and reference 3.3 distinct entities per response versus 1.3. More length and more entities means more citation slots per answer. Treating the two as one surface is the first mistake we see operators make. They check their AI Overviews presence, see a citation, and assume AI Mode is covered. It usually is not.
How does AI Mode actually select sources?
AI Mode selects sources through query fan-out and passage-level retrieval, not whole-page ranking. Google describes the technique in its own words as "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf." Practitioner teardowns put the real number at roughly 8 to 15 hidden sub-queries per prompt, each one fired in parallel against the live web, the knowledge graph, and Google's index.
Retrieval then happens at the passage level. The system does not grade your page as a whole and slot it into a ranking. It pulls individual paragraphs that cleanly answer one of the sub-queries, scores those passages, and stitches the surviving ones into the answer. A single page can contribute one passage to one sub-query and never appear again. This is why a clear, self-contained answer near the top of a section matters more than the page's overall authority: the passage is the unit that gets retrieved, evaluated, and cited.
How different are AI Mode and AI Overviews citations, really?
They are different far more often than they are the same. Two independent studies landed on the same conclusion from different angles. Ahrefs, comparing 730,000 query pairs from September 2025, found a 13.7% URL overlap overall and 16.3% even among the top three citations. Victorious, analyzing 1,540 queries, found 30 to 35% of AI Overviews URLs carried into AI Mode and that not a single query produced an identical citation set. In Victorious's data, 77% of unique domains showed up in only one of the two surfaces.
The studies differ on the exact overlap figure because they measured different query sets and matching rules, but the direction is identical: the two surfaces reward different sources. The table below is the operator summary.
| Dimension | AI Overviews | AI Mode |
|---|---|---|
| Where it appears | Summary block atop normal results | Separate conversational surface |
| Retrieval style | Lighter fan-out, shorter answers | Heavy fan-out, passage-level |
| Avg domains cited per query | ~7.7 | ~9 |
| Responses with no citation | 11% | 3% |
| Wikipedia citation share | 18.1% | 28.9% |
| Quora | Baseline | ~3.5x more |
| YouTube and video | ~2x more | Baseline |
| Source posture | Selective, curated | Broad, diverse |
Numbers from Ahrefs (730K pairs) and Victorious (1,540 queries). The headline: a citation in one is not a proxy for the other.
Which sources does AI Mode favor over AI Overviews?
AI Mode tilts toward reference and community sources; AI Overviews tilts toward video and homepages. In Ahrefs's data, Wikipedia appeared in 28.9% of AI Mode citations versus 18.1% in AI Overviews, and Quora was cited about 3.5x more often in AI Mode. Health and medical sites were cited roughly twice as often. AI Overviews, in turn, cited YouTube, video, and homepages around twice as much. If your category lives on Quora threads and reference pages, AI Mode is the friendlier surface; if it lives on YouTube, AI Overviews may surface you first.
One conflict is worth naming. Semrush's three-month study of 230,000 prompts and 100M+ citations reported Wikipedia at only about 2% of AI Mode citations by raw volume, with LinkedIn near 15% and a visible tilt toward Google-owned properties. That clashes with Ahrefs's 28.9%. The gap is methodological: Ahrefs measured the share of queries where Wikipedia appeared at least once, Semrush measured Wikipedia's share of total citation volume. Both can be true. We weight Ahrefs's per-query framing for "will I get a slot" decisions and read Semrush as a reminder that a few mega-domains still eat a large slice of total volume.
Why does AI Mode cite a wider pool of sources?
Because fan-out multiplies the number of retrieval events, and more retrieval events mean more citation slots. A single AI Overviews answer might resolve two or three sub-intents. An AI Mode answer decomposes the same question into many more sub-queries, each one its own retrieval with its own winning passage. That is the mechanical reason AI Mode cites about 9 domains per query against 7.7 for AI Overviews, and why only 3% of AI Mode responses cite nobody versus 11% of AI Overviews.
For a brand that is not yet a category leader, this is the good news in the data. AI Overviews is a near-zero-sum block where a curated handful of sources win. AI Mode's wider net means a relevant passage on a mid-authority page can earn a slot on a long-tail sub-query that the big domains did not bother to cover. The breadth is the opening. The question is whether your content is structured so a single passage can stand on its own when the fan-out pulls it.
What does this mean for getting your brand cited in AI Mode?
Optimize the passage, then optimize your presence across third-party sources. Two levers move AI Mode citation, and neither is your Google ranking. The first is passage-level structure: every section needs a direct, self-contained answer in the first 40 to 60 words, written so it makes sense lifted out of the page with no surrounding context. The fan-out retrieves passages, not pages, so a passage that only resolves mid-paragraph after three sentences of setup loses to a competitor's clean opener.
The second lever is breadth of presence on the sources AI Mode actually pulls. Given its lean toward Wikipedia, Quora, and reference pages, the brands that get cited are the ones with a confirmed entity across multiple third-party surfaces, not just a strong owned domain. This is the same causal chain that runs through our analysis of why a competitor shows up in ChatGPT and you do not: unlinked brand mentions predict AI citation far more strongly than backlinks. For the underlying mechanics, see our piece on backlinks vs brand mentions for AI visibility.
How do you optimize for AI Mode without abandoning AI Overviews?
Run one content standard that serves both, then diversify where you place signals. The overlap data does not mean you need two separate content operations. The structural fundamentals that help AI Mode retrieve a clean passage, such as answer-first sections, 120 to 180 word chunks, comparison tables, and FAQ schema, are the same fundamentals that help AI Overviews. Build content to that standard once and both surfaces can extract from it.
Where the surfaces genuinely diverge is source placement. Because AI Mode leans on Quora and reference pages while AI Overviews leans on video, a category that wants both should not put all its eggs in one format. Seed Quora answers on the high-intent questions in your space, since Quora is cited several times more often in AI Mode and we have a full playbook on getting a Quora answer into Google AI Overviews. Keep video in the mix for the AI Overviews tilt. Then map the third-party domains AI search already cites in your vertical, using our reference list of the 50 domains that drive most AI citations, and earn presence on the ones that match.
How do you test whether AI Mode cites you?
Run the same prompt in both surfaces and read the citations, not the answers. Open AI Mode and AI Overviews side by side for your top three category queries, phrased the way a user types them, such as "best [category] for [use case]." Record which surface cites you, which cites your competitor, and the exact source URL next to each citation. Because the two surfaces overlap on so few URLs, you will often find you are present in one and invisible in the other, and the source type usually tells you why.
If your competitor's AI Mode citation is a Quora thread or a Wikipedia entry you do not have, that is a retrieval-source gap, not a content-quality gap. If you are cited in AI Overviews via a YouTube result but missing from AI Mode, the fix is reference and community presence, not a better homepage. Track this monthly, since cited sources shift heavily month to month across both surfaces. The diagnostic takes 20 minutes and tells you whether your next move is a structure fix or a presence-building one.
Frequently asked questions
Is AI Mode the same as AI Overviews?
No. AI Overviews is the summary block at the top of a normal Google results page. AI Mode is a separate conversational search surface with its own longer answers and its own citation set. Ahrefs found the two cite the same URL only 13.7% of the time for identical queries, so they function as distinct surfaces that need to be tracked and optimized separately.
How does AI Mode decide which sources to cite?
AI Mode uses query fan-out: it breaks your question into roughly 8 to 15 sub-queries, runs them in parallel against the live web, knowledge graph, and Google's index, then retrieves at the passage level rather than the page level. It scores individual paragraphs that answer each sub-query and synthesizes the winning passages into one response. A clean, self-contained answer near the top of a section is what gets retrieved, not the page's overall authority.
Does ranking in Google's top 10 get me cited in AI Mode?
Not reliably. AI citation is a different graph from organic ranking. Only a minority of URLs cited across AI surfaces rank in the top 10 for the original prompt, and AI Mode in particular draws from a wider, more diverse pool, citing roughly 9 domains per query. A strong ranking helps but does not guarantee a citation, and many cited passages come from pages that do not rank on page one.
Which sources does AI Mode favor compared to AI Overviews?
AI Mode cites Wikipedia more heavily (28.9% of citations versus 18.1% for AI Overviews), cites Quora about 3.5x more often, and cites health and medical sites roughly twice as much. AI Overviews leans about 2x more on YouTube, video, and homepages. So a brand strong on community and reference content tends to do better in AI Mode, while video-heavy brands may surface first in AI Overviews.
Why am I cited in AI Overviews but not in AI Mode?
Because the two surfaces overlap on only 13.7% of URLs. Being cited in one is statistically not a predictor of the other. Usually the gap is a source-type mismatch: AI Overviews surfaced you through a format it favors, such as video, while AI Mode is pulling Quora threads, Wikipedia, and reference pages where you have no presence. The fix is to build presence on the source types AI Mode actually retrieves for your category.
How often do AI Mode citations change?
Frequently. Cited sources shift substantially month to month across both AI Mode and AI Overviews, so a single citation snapshot is not durable. Treat AI visibility as a tracked metric, not a one-time win. Re-run your prompt sweep monthly, watch which source types are gaining, and keep feeding fresh signals into the third-party sources the engines pull from rather than assuming an early citation will hold.
:::
