AI Overviews sits on top of classic search results and largely inherits them. AI Mode runs its own query fan-out and can name brands that rank nowhere on page one for the original query.
Which means the two disagree about you regularly, and averaging them into one Google number throws away the most useful signal either produces.
Both are Google, both are AI, both run on the same index. They are still different enough that a brand can appear reliably in one and never in the other, and knowing which one you are missing tells you what to do about it.
What does each one actually do?
Google documents the difference itself, and the key mechanism has a name.
AI Mode “expands what AI Overviews can do with more advanced reasoning, thinking and multimodal capabilities”, and critically it “uses a ‘query fan-out’ technique, issuing multiple related searches concurrently across subtopics and multiple data sources and then brings those results together”.
Sits above the classic results for the query that was asked, and draws largely on what would have ranked for it. Closer to a summary of page one than to a new retrieval.
Fans the question out into related subtopics, searches those concurrently, and assembles an answer from the union. The original query is a starting point rather than the whole search.
Why does that produce different brand sets?
Because fan-out changes which pages are eligible to be considered at all.
If AI Overviews largely summarises what already ranks, then your position for that query is close to determinative. If AI Mode issues a spread of related searches first, your ranking for the original query is only one of several doors, and you can enter through any of them.
“Best field service scheduling software for small teams”
Overviews leans on what ranks for the query as asked. AI Mode fans out first.
the original querypricing for small teamsscheduling vs dispatchalternatives to the leaderOverlapping but not identical brand sets, from the same index on the same day.
different sourcesdifferent namesWhat is the same about them?
The entry requirement, which is the part that saves you work.
Google states there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”, and that to qualify as a supporting link “a page must be indexed and eligible to be shown in Google Search with a snippet”. That applies to both. There is no separate markup, no AI-specific file, and no second technical project.
Both also draw on the same index, so anything that removes you from Search removes you from both. Your existing technical SEO is doing the qualifying work for the whole family.
What has to be true before either can name you?
One thing, and it is not AI-specific at all.
A page has to be indexed and snippet-eligible. That means the ordinary technical work still gates both surfaces: crawlable, indexable, and not suppressed from snippets. Worth knowing that robots.txt is the wrong instrument for controlling any of this, since Google states it “is not a mechanism for keeping a web page out of Google” and that a disallowed page can still be indexed when other sites link to it.
Snippet controls matter more here than people expect. Google’s snippet documentation describes nosnippet, max-snippet and data-nosnippet as ways to restrict what can be shown. Applying those broadly, which some sites did years ago to protect content, restricts exactly the eligibility both AI surfaces depend on.
The baseline. Not indexed means not eligible for either surface, whatever else is true.
An old blanket nosnippet is a quiet way to have removed yourself from both.
No AI-specific file, no special markup, no second technical project. Google says so directly.
How should you read a disagreement?
As a diagnosis rather than as noise, which is the entire argument for tracking them separately.
In Overviews, not AI Mode
You rank for the query as asked but are thin on the surrounding subtopics. The fan-out went somewhere your coverage does not reach.
Breadth gap
In AI Mode, not Overviews
You are strong on adjacent questions and weak on the head term itself. Classic ranking work on the primary query is the lever.
Head-term gap
In neither, but indexed
Eligible and not chosen. This is an evidence problem rather than a technical one, and no Google-specific fix applies.
Evidence gap
In neither, not indexed
You have failed the stated entry requirement. Nothing else on this page matters until that is fixed.
Fix first
Does one matter more than the other?
It depends on the question being asked, and that is a more useful framing than picking a winner.
AI Overviews appears on a very wide range of queries, including plenty where nobody is buying anything. AI Mode is built for complex, multi-part questions with follow-ups, which is a better description of how somebody evaluates software than a single search is.
So for a B2B buyer working through a decision, the AI Mode shape is closer to the real behaviour. For sheer breadth of exposure, Overviews reaches more queries. Track both, weight them by which one your buyers actually use, and resist any tool that quietly averages them.
What should you do differently for each?
Almost nothing separates them technically, and the one thing that does is worth knowing.
For AI Overviews, ordinary ranking work on the query itself is close to the whole game, because the surface leans on what already ranks. For AI Mode, breadth across the surrounding subtopics matters more, because fan-out is searching for those subtopics whether or not you have thought about them.
That is an argument for covering a topic properly rather than writing one page per keyword. Not because a search engine rewards thoroughness in the abstract, but because a mechanism that generates its own related searches will find you only if you have something to find on the related questions.
Does fan-out change what you should publish?
It raises the value of covering a topic completely, and lowers the value of writing one page per keyword.
If a mechanism generates its own related searches before answering, then the questions it asks are not the ones you optimised for. It will ask about pricing, about alternatives, about the specific constraint in the original question, and about things adjacent to all of those. You appear in the final answer if you have something worth retrieving on some of them.
That is an argument for depth on a subject rather than breadth across phrasings. A single thorough page covering a topic and its neighbours has more surface area for fan-out to catch than five thin pages each targeting a variant of the same phrase, which is close to the opposite of a decade of keyword-page habit.
It also raises the value of self-contained sections, since retrieval scores passages rather than whole documents. A section that answers its own heading can be caught by a subtopic search even when the page as a whole is about something broader.
Google’s third surface, Gemini, is a separate product again and belongs in its own row. All three are covered in the engine guide, the retrieval mechanics underneath are in the RAG explainer, and if you want to see for yourself whether the two disagree about you, the five-minute test works on both.