AI Mode SEO is regular SEO. At Search Central Live in Barcelona, Google said AI Overviews and AI Mode use the same crawler (Googlebot), the same index and the same quality systems as classic Search. The AI adds one step on top: it sends fan-out queries to Search and builds an answer from the pages that come back.
Key takeaways
- Gary Illyes closed the three-day event with the line "AI on Google is just SEO", as reported by attendees.
- AI Overviews and AI Mode don't have their own crawler. Gemini does.
- AI answers are grounded with fan-out queries the model sends to Search. Those queries don't appear in Search Console.
- Structured data isn't required for AI features. Google passes it to AI Overviews and AI Mode as context, and said it's never a negative to have it.
I've argued in GEO vs SEO that most of "GEO" is SEO with a new label. Three days in Barcelona gave that argument Google's own diagrams. Here's what was said, where it came from, and what I'd do with it.
A sourcing note: Google hadn't published the slides when I wrote this. Quotes come from attendee recaps, mainly ROAST's day-by-day notes, plus Google's own AI optimisation guide, which says the same things in writing.
How AI Mode actually gets its answers
Picture a user asking AI Mode: "Which hiking trails near Barcelona are good for kids in October, and where can we eat after?"
That's three questions in one. No single page answers it. So the model breaks it into smaller searches (trails near Barcelona for kids, October weather, family restaurants near those trails) and sends each one to Google Search. The results come back, the model reads them, and it writes an answer with links.
John Mueller described this on day 3. After normal query understanding (language detection, stop words, entities, synonyms), AI search adds a step: the LLM creates fan-out queries and sends them to the search engine to ground the answer.
Two consequences:
- Your page is retrieved by Search, not by the model. If you don't rank for the fan-out query, you're not in the pool the model reads.
- You can't see the fan-out queries. They don't come from users, so they don't show in Search Console. You can only infer them from the sub-questions a long query contains.
Gary Illyes' numbers explain why this matters now. Per ROAST, AI Mode queries have doubled every quarter since launch, are about 3 times longer than classic queries, and US follow-up queries grew more than 40% a month on average. Longer, multi-part questions mean more fan-out, and more chances for a focused page to be pulled in.
Same crawler, same index
Cherry Prommawin and Gary Illyes said on day 1 that Google Search, AI Overviews and AI Mode rely on the same main crawler, Googlebot. Gemini uses a separate one, "for now at least" according to ROAST.
| Google surface | Crawler | Index | Needs special markup? |
|---|---|---|---|
| Classic Search results | Googlebot | Google Search index | No (structured data helps rich results) |
| AI Overviews | Googlebot | Google Search index | No |
| AI Mode | Googlebot | Google Search index | No |
| Discover | Googlebot (no separate bot) | Google Search index | No (large images help) |
| Gemini app | Separate crawler | Not covered in detail | Not covered |
So the old checklist still decides whether you can be cited at all:
- Googlebot can reach the page (check CDN and firewall rules; day 1 flagged CDNs blocking bots more often).
- The page renders without needing scrolls or clicks (Google's renderer uses a viewport about 10,000 px tall and doesn't interact, per the day 2 rendering talk).
- The page is indexed. If it's sitting in crawled – currently not indexed, no AI feature will cite it.
Structured data: context, not a ticket in
This was one of the clearer answers of the event. Ryan Levering, a Google engineer, said on day 2 that Schema.org data is cleaned and filtered, then either powers rich results or is passed to AI Overviews and AI Mode as context. Gary Illyes added on day 3 that rich results usually need structured data, but AI results are built from text.
Google's guide says the same in writing: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add."
My read: keep your structured data accurate because rich results still depend on it, and Levering said "It's never a negative to have structured data." But don't expect a new schema type to get you into AI answers. The visible text does that job.
llms.txt and other "AI files"
Google's AI optimisation guide is blunt: Google Search doesn't use llms.txt, and you don't need new machine-readable files, AI text files or Markdown versions of pages to appear in Search. Keeping an llms.txt for other services is fine.
At the event, community speaker Carlos Ortega Roldán made the same call in his lightning talk on preparing sites for AI agents. He also warned that Markdown copies of pages risk duplicate content. What he did recommend is plain good practice: correct heading nesting, semantic HTML landmarks, ARIA labels and low layout shift.
If you've already built an llms.txt, there's no need to remove it. My earlier piece on what llms.txt is covers where it might still help outside Google.
Quality decides which pages make the cut
Retrieval narrows the candidates by language and country. Then, in Gary Illyes' words as quoted by ROAST: "But quality is what makes the final decision."
Duy Nguyen summarised quality as four things on day 3:
- Effort: how much work went into the page.
- Originality: whether it adds something not already in the results.
- Talent or skill: whether the creator knows the subject.
- Accuracy: whether it's correct.
And on AI-written content, ROAST recorded his point that "The point is the quality, not who wrote it." Digitaloft's recap reported Google saying it has no system to detect whether content was AI-written.
That doesn't make mass AI content safe. Gary said spam updates are increasing because AI has inflated content volume, and that scaled content is now a bigger problem than link spam. A "We're listening!" slide also listed manipulating AI responses among spam behaviours Google is watching.
What "AI on Google is just SEO" doesn't mean
I'd push back on one reading of the quote. It doesn't mean nothing changes.
Three things are different, and Google's own talks showed them:
- The unit is the passage, not the page. A fan-out query asks a narrow question. Pages that answer sub-questions clearly in one section are easier to ground on. My guide to ranking in AI Overviews shows how I structure those sections.
- Measurement is weaker. Fan-out queries are invisible in Search Console, and Ariel Kroszynski gave no dates for new generative AI reporting. Use the Generative AI performance report you already have, plus manual checks; see AI search visibility.
- Trust is part of the journey. Pablo Pérez showed "is [brand] legit" searches rising since 2020 and peaking in 2026. People research more before deciding, and AI answers often summarise what others say about you.
The community talks made the same point from the other side. Thiago Pojda's lightning talk was titled "GEO is not 'Just SEO'", arguing that GEO is much harder to measure. Both things can be true: the system is SEO, the reporting isn't there yet.
What I'd do this month
- Make sure Googlebot isn't blocked or challenged by your CDN or WAF, and that challenge pages don't return 200.
- Take your 10 most important pages and list the sub-questions a long AI query would split them into. Give each one a clear, self-contained section.
- Check rendering: nothing important should load only on scroll or click.
- Keep structured data accurate for rich results, but stop treating it as an AI ranking lever.
- Review pages against effort, originality, skill and accuracy, and cut or merge thin ones.
- Track AI visibility with Search Console's Generative AI report plus a monthly manual prompt check.
For the full three-day recap, read my Search Central Live Barcelona summary. For the deeper version of this approach across ChatGPT, Gemini and Perplexity, see LLM SEO.
Frequently asked questions
Is AI Mode SEO different from regular SEO?
Mostly no. Google said AI Mode and AI Overviews use Googlebot, the Google index and the same quality systems as Search. The difference is that AI answers are built from several pages through fan-out queries, so clear, self-contained sections help.
What are fan-out queries?
They are smaller searches an AI model creates from a long user question and sends to Google Search to ground its answer. Google said they don't appear in Search Console because they don't come from real users.
Does Google use llms.txt for AI Overviews?
No. Google's AI optimisation guide says Google Search doesn't use llms.txt and you don't need special AI files or Markdown versions of pages. Keeping one for other services won't hurt your Google visibility.
Do I need structured data to appear in AI Overviews?
No. Google says no special schema is required for generative AI features. Structured data is passed to AI Overviews and AI Mode as context and still matters for rich results, so keep it accurate.
Who said "AI on Google is just SEO"?
Gary Illyes of Google, in the closing session of Search Central Live Deep Dive Europe in Barcelona on 2 October 2026, as reported by attendees including ROAST and aioseo.fr.
.webp)
Connect with Akif on LinkedIn, Instagram and Facebook.
Driven by advanced SEO expertise, deep marketing analytics, high-impact content strategy
With 5+ years of hands-on experience, I specialize in holistic search strategies that don’t just rank—they drive real, measurable business growth. I’ve worked across industries including healthcare, hospitality, legal, e-commerce, and professional services, helping brands dominate their target markets. My approach bridges the gap between raw data and creative execution. Every strategy I build is rooted in rigorous market analysis, structured SEO frameworks, and tailored content ecosystems—no templates, no shortcuts. Whether you’re a single-location brand or scaling across multiple cities, I create data-driven marketing systems designed to compound results and grow with you.
Want Your Brand Cited in AI Answers?
Get an AI search visibility audit covering ChatGPT, Perplexity and Google AI Overviews — and a plan to earn more citations.
Book a Free SEO Audit →