SEO for Google AI Mode has attracted a lot of confident advice for a feature whose inner workings Google has described in only a few sentences. Those sentences are worth reading closely, because they say more than the checklists built on top of them. This post sticks to what Google has said, what follows from it, and what is still unknown.
What Google has said about AI Mode
Google began rolling out AI Mode in the U.S. on 20 May 2025, with no Labs sign-up required. The announcement, by Elizabeth Reid, VP and Head of Search, is the Google I/O 2025 update on AI Mode. The parts that matter for SEO:
| What Google said | What it means for your pages |
|---|---|
| AI Mode “uses our query fan-out technique, breaking down your question into subtopics and issuing a multitude of queries simultaneously” | Your page can be retrieved for one subtopic of a larger question, not only for the query the user typed |
| Deep Search “can issue hundreds of searches” and create “an expert-level fully-cited report” | For complex questions, many more pages are consulted, and citation is the visible outcome |
| AI Mode offers follow-up questions and “helpful links to the web” | The conversation continues, so pages covering the next question can be pulled in too |
| Ranking in AI features needs normal SEO (Gary Illyes) | No separate playbook has been documented; crawlability, quality and relevance still decide |
Query fan-out is the part that matters
A traditional search returns results for the query typed. Fan-out means AI Mode runs many related searches behind the scenes and assembles an answer from what they return. A question like “best way to finance a used car with poor credit” can become separate searches on lender types, interest rates, credit thresholds and alternatives.
That has two consequences. First, a page that would never rank for the full question can still be cited if it is the clearest answer to one of its parts. Second, a page that tries to cover everything at once may lose to narrower pages that answer each subtopic cleanly. Fan-out rewards being findable for specific sub-questions.
SEO for Google AI Mode: structure pages around subtopics
If retrieval happens at the subtopic level, structure matters more than length. The practical changes are modest:
- One subtopic per section. Descriptive H2s and H3s that name the sub-question, followed by a direct answer in the first sentence or two.
- Map the questions before you write. List the sub-questions a searcher would need answered, then check each has a clear home on your site.
- Separate pages for separate intents. If a subtopic has its own search intent, it probably deserves its own page. Our guide to search intent covers how to tell.
- Link the cluster together. A set of pages that cover a topic’s parts, linked sensibly, is what topical coverage looks like in practice; see how to build topical authority.
None of this is AI Mode-specific advice. It is good site architecture, and it was good before fan-out existed. Fan-out simply makes the cost of a muddled page more visible.
Follow-up questions push in the same direction. A user who asks a first question and then a second is walking through a topic in steps, and each step is a fresh chance for a different page to be cited. A site that answers the obvious next question, on the same page or one clearly linked from it, has more ways in than a site that stops at the first answer. The internal linking between those pages is what lets Google see they belong together.
What has not changed
Google has said that ranking in its AI features needs normal SEO. That is the most useful sentence in this whole area, and it is routinely ignored by people selling AI Mode optimisation. The pages fan-out retrieves come from Google’s index, so the basics still apply:
- The page has to be crawlable and indexed.
- It has to be relevant to at least one of the generated sub-queries.
- It has to be good enough to be chosen over the alternatives.
We are not aware of any documented AI Mode-specific ranking factor, and we will not list one. Advice that names special markup, prompt-style phrasing or secret signals for AI Mode is guesswork unless it cites Google.
The same applies to the claim that AI Mode favours big brands, or small sites, or long pages. Any of those may turn out to be true in a given niche. None has been established, and a single screenshot of one answer proves nothing about the next.
Measuring AI Mode traffic
This is where the limitation sits. Whether Search Console separates AI Mode clicks and impressions from the rest of your search traffic is not verified here, so do not plan a reporting process on the assumption that it does. Check what your own Search Console shows before building a dashboard around it.
What you can do is watch the shape of your search data over time: queries getting longer and more conversational, pages earning impressions for questions they only partly answer, and changes in click-through on informational queries. None of these proves AI Mode is involved, but together they tell you where the pressure is. Our comparison of Search Console and Google Analytics covers which tool answers which question.
The traffic question
A fully cited report is still a report. If AI Mode answers the question well, some users will not click any of the links, cited or not. That is the same trade-off publishers have faced with AI Overviews, and the evidence there is worth reading before you decide how much to invest; see our analysis of AI Overviews and traffic.
The pages most exposed are the ones that answer a simple factual question and little else. The pages least exposed are the ones a user still needs after reading a summary: comparisons with real detail, tools, original data, and anything that requires a decision the user wants to check for themselves. If your content plan leans heavily on the first type, AI Mode is a reason to rebalance it, whatever the citation counts look like.
The honest summary is short. Google has described the mechanism, said normal SEO applies, and documented nothing else. Structure your pages so each subtopic has a clear answer, keep the technical basics clean, and judge AI Mode by the traffic it sends rather than by the promises around it. For the wider argument about optimising for AI answers, read our post on GEO vs SEO, and the algorithm archive for related coverage.
Frequently asked questions
When did Google launch AI Mode?
Google began rolling out AI Mode in the U.S. on 20 May 2025, with no Labs sign-up required.
What is query fan-out?
Google describes it as breaking a question into subtopics and issuing many queries simultaneously on the user’s behalf, then building the answer from the results.
Are there special ranking factors for AI Mode?
None that Google has documented. Google has said ranking in its AI features needs normal SEO.
Can I see AI Mode traffic in Search Console?
That is not verified here. Check what your own Search Console reports before relying on it for AI Mode reporting.
The takeaway AI Mode retrieves pages through query fan-out, so cover each subtopic clearly. Beyond that, Google says normal SEO applies, and no AI Mode-specific ranking factor has been documented.
