AI Mode vs AI Overviews: What's Actually Different

• AI Overviews are summaries above organic results. AI Mode is a separate surface that returns no organic results.
• Both use query fan-out — Google's documented technique of issuing multiple related searches to build one answer.
• The two surfaces cite the same URLs only around 13-14% of the time, per independent analyses.
• Ranking first for the visible query is not how you get cited. Ranking for the invisible sub-queries is.
• Google is merging the two into one experience, with follow-ups from an Overview flowing into AI Mode.

The plain difference

	AI Overviews	AI Mode
How you get it	Appears automatically on a normal search	You choose to enter it
Organic results	Still there, below the summary	Not shown
Interaction	Single answer	Conversational, with follow-ups
Best suited to	One quick question	Multi-step research and comparison
What it costs you	The click, when the answer suffices	The whole results page

The distinction that matters commercially is the second row. An AI Overview leaves your blue link on the page — pushed down, often below the fold, but present and clickable. AI Mode does not have a blue link to push down. The only route to your site is a citation.

Google has said AI Mode passed one billion monthly users, with queries more than doubling every quarter since launch. Whatever one thinks of the direction, the surface is not a pilot.

Query fan-out, and why it breaks the usual mental model

Both surfaces use what Google's documentation calls a query fan-out technique: issuing multiple related searches across subtopics and data sources to develop a response.

That sentence is easy to skim past, and it changes how visibility works.

When someone asks a broad question, the system does not run one search. It decomposes the question into a set of narrower sub-queries — different facets, comparisons, definitions, qualifications — runs them in parallel, retrieves passages for each, and assembles a single answer from the pieces. Reporting through 2026 puts the count at up to roughly sixteen parallel sub-queries for a complex question.

The consequence is the part worth internalising. You are not competing for the query the user typed. You are competing for a set of queries you never see, that were generated from it, and any one of them can put a passage of yours into the answer. Independent analyses report that pages ranking for those fan-out queries are substantially more likely to be cited than pages ranking only for the visible one.

This also explains something that otherwise looks like a bug: sites that rank nowhere on page one turning up as citations. They were not retrieved for the headline query. They were retrieved for a sub-question the headline query implied, and they answered it in a self-contained passage.

The overlap problem

Here is the finding that should reset expectations.

Independent analyses through 2026 report that AI Mode and AI Overviews cite the same URLs only around 13-14% of the time, even while reaching broadly similar conclusions on the same questions. Separately, only about 14% of URLs cited in AI Mode appear in the top ten organic results for the same query.

These are vendor and agency studies rather than published research, and methodologies vary, so treat the exact percentages as indicative rather than precise. The direction, though, is consistent across sources and it is large enough that the rounding does not matter.

Two conclusions follow:

• Position one is not a citation strategy. The correlation between ranking and being cited is much weaker than the SEO reflex assumes.
• Appearing in one AI surface tells you almost nothing about the other. They are separate retrievals over a shared index, and they resolve differently.

If you want the evidence base on what does correlate with citation, that is the subject of how to get cited in AI search.

The merge

Google has now said it is bringing AI Overviews and AI Mode into one AI Search experience. In practice that means a user can ask a follow-up question directly from an AI Overview and flow into an AI Mode conversation, with the context of what they were reading carried across. Gemini 3 became the default model for AI Overviews globally.

For anyone planning around these products, the useful reading is this: the boundary between "a summary on the results page" and "a conversation with no results page" is dissolving from the user's side. The two-product mental model is already dated. What persists is the underlying mechanic — fan-out retrieval assembling answers from passages, with citations as the only route back to a site.

What actually follows from all this

Not much that is new, which is itself the finding. The tactics that survive contact with fan-out retrieval are unglamorous:

Write passages that stand alone. Retrieval operates on chunks, not pages. A paragraph that answers one specific question completely, without depending on the three paragraphs above it, is a retrievable unit. A paragraph that begins "As we saw earlier" is not.

Cover the sub-questions explicitly. If fan-out generates the sub-queries, the way to be present for them is to answer them on the page — as real sections with real headings, in the words a person would use. This is what a well-built FAQ section is genuinely for, as opposed to what most of them are used for.

Be specific and checkable. Numbers, dates, named sources, stated limitations. Systems that must attribute a claim prefer text that makes attribution safe.

Let the right crawlers in. None of the above matters if the indexing bot cannot fetch the page — see AI crawlers and robots.txt, and note that Cloudflare's default block lands in September.

Measure what you can, honestly. Clicks from AI Overviews arrive as ordinary Google organic traffic and cannot be separated in analytics; assistant referrals can be, imperfectly. Both are covered in how to track AI search traffic in GA4.

The uncomfortable summary is that there is no lever here. There is a retrieval system that rewards clear, self-contained, verifiable writing, and a measurement environment that makes it hard to prove you are winning. That is a worse deal than the last decade of search offered, and pretending otherwise is not a strategy.

Sources: Google — Expanding AI Overviews and introducing AI Mode · Google — AI Mode and AI Overviews get Gemini upgrades · Google Search Central — AI features and your website · Search Engine Land — Google updates links within AI Overviews & AI Mode