How SEO Works in Google AI Mode

Research basis: what this study looked at

This article is based on three groups of evidence. The first group is official Google documentation about generative AI features in Search, AI Overviews, AI Mode, technical eligibility, AI-generated content and the Search Engine Optimization (SEO) basics that Google says still matter. The second group is public analysis of query fan-out, especially how AI Mode can expand one user query into multiple related retrieval tasks. The third group is market research about AI Overviews, zero-click behavior and click-through rate pressure, including data from Ahrefs, Semrush and Pew Research Center.

The important distinction is this: Google’s official position is that SEO remains relevant because generative AI features are rooted in core Search systems. External studies focus on traffic consequences and visibility shifts. Both can be true at the same time. AI Mode can still use search and links while also changing how many users click, which sources are surfaced, and what kind of content is worth producing.

That is why the useful question is not “Is SEO dead?” That question is lazy. The useful question is: how does SEO work when the interface is conversational, the system expands the query, the answer is synthesized, and the user may never click unless your brand earns a reason to be remembered?

The short answer: SEO becomes source engineering

In classic search, SEO often looked like this: choose a keyword, build the best page, earn links, fix technical issues and compete for a ranking. That model still exists. Google has not deleted crawling, indexing, ranking systems, snippets, links, structured data, internal linking, content quality or page experience. But AI Mode adds another layer on top of the old model.

In AI Mode, the system may not simply return one ranked list. It may interpret the prompt, break it into subtopics, search for supporting sources, compare information, synthesize an answer and provide links. This means your page must do more than rank for an exact query. It must be a good source for one or more pieces of the answer.

That is what I mean by source engineering. The page has to be crawlable, indexable and eligible to appear in Search. It has to be technically clear enough that Google can process it. It has to contain useful passages that answer real sub-questions. It has to demonstrate experience or evidence. It has to connect to the rest of the site through internal links. It has to be written for people, but structured well enough that machines can extract the useful parts.

For AYSA.ai, the practical translation is simple: SEO work in AI Mode is not a one-time content sprint. It is a continuous system that finds weak pages, improves them with useful evidence, connects them internally, measures search and AI visibility signals, and asks for approval before publishing meaningful changes.

What Google AI Mode changes

Google describes AI Mode as useful for nuanced questions, complex comparisons and deeper exploration. That single sentence has huge SEO implications. Users are no longer limited to short keyword fragments like “best SEO tool” or “how SEO works.” They can ask, “How does SEO work on AI Mode if I run a small ecommerce site and already have blog content but no technical team?” That prompt contains multiple intents: explanation, platform change, business type, content state, constraint and recommended next step.

Traditional keyword research tends to flatten this into one keyword. AI Mode expands it. A good answer may need a definition of AI Mode, a difference between AI Overviews and AI Mode, an explanation of query fan-out, technical eligibility requirements, content examples, measurement advice, risks around AI-generated content, and a practical workflow for the user’s business type.

This is why exact-match SEO becomes less central. A page can be relevant even if it does not repeat the exact long prompt. Conversely, a page that repeats the words but lacks evidence, structure and usefulness may still be weak. AI systems are better at matching meaning than old keyword tools, but they still need available, reliable, crawlable content to ground answers.

The change also affects page architecture. Thin pages built for every keyword variation are a poor fit. Google’s own guidance warns against creating separate pages for every possible search variation just to manipulate rankings or AI responses. A stronger approach is to build fewer, better pages that cover a topic with real expertise, then use internal linking to connect related questions and commercial actions. For a broader explainer, see our complete Google AI Mode SEO guide for SMEs; this study focuses specifically on how SEO works when AI Mode expands and selects sources.

Query fan-out: the hidden reason one article can rank for many AI needs

Query fan-out is the concept that matters most for this keyword. In simple terms, AI Mode can take one prompt and generate related searches behind the scenes. If someone asks how SEO works in AI Mode, the system may need information about AI Mode itself, Search indexing, retrieval-augmented generation, AI Overviews, content quality, query expansion, ecommerce visibility, local SEO, measurement and user behavior. One visible prompt can hide ten invisible retrieval tasks.

This changes how we should plan content. The goal is not to stuff every sub-query into one chaotic page. The goal is to understand the user’s task deeply enough that the page contains the most important answer components and links to supporting assets. A strong AI Mode article should answer the core question, then point to deeper pages about technical SEO, AI search visibility, content strategy, entity SEO, structured data, ecommerce SEO and automation.

For example, a page about “how SEO works in AI Mode” should not become a complete tutorial on every schema type. But it should explain that structured data is still useful as part of normal SEO, while avoiding the false promise that there is a magic AI Mode schema. It should not claim that llms.txt is required for Google AI Mode, because Google says it does not use such files for Search. But it can explain why some businesses may still maintain those files for other AI systems. That nuance is exactly the kind of answer a serious reader needs.

The commercial implication is powerful. If query fan-out creates multiple paths into the answer, then a brand can win by owning the cluster, not just the keyword. The homepage, blog article, product pages, glossary, help docs, pricing page and case studies should all reinforce the same entity: AYSA.ai as an SEO automation platform for SMEs that helps turn SEO recommendations into approved execution.

From ranking to selection: the new visibility model

Classic SEO asks, “Where do we rank?” AI Mode forces a second question: “Are we selected as a useful source?” Ranking and selection overlap, but they are not identical. A page can rank well and still be ignored by an AI answer if it does not provide extractable value. Another page may be selected because it contains a precise explanation, data point, comparison or product fact that helps the generated response.

Selection has several layers. First, the page needs technical eligibility: crawlable, indexable, not blocked, and eligible for snippets. Second, the page needs relevance to the user task or one of the fan-out subtopics. Third, it needs quality: experience, originality, clarity, trust and usefulness. Fourth, it needs answer utility: passages that can support a synthesized response without forcing the system to infer too much. Fifth, it needs entity clarity: the brand, product, author, topic and relationships should be easy to understand across the web.

This is why “write longer content” is not the answer. A long page can still be vague. A short page can be useful if it answers a narrow question precisely. The stronger rule is this: build pages with high signal density. Every section should help the user decide, understand, compare, execute or trust. If a section only repeats common knowledge, it should be improved or removed.

External click studies make this more urgent. Ahrefs found a large CTR reduction for top-ranking pages when AI Overviews are present, while Semrush and Pew data show that AI summaries are becoming common and change user behavior. The exact numbers vary by methodology, query type and date, but the direction is clear: winning search is no longer only about getting the click. It is about earning visibility, trust, citations, branded recall and the right clicks when the user does need to act.

The technical layer: no AI Mode magic, just fewer excuses

Google’s AI guidance is surprisingly conservative on technical SEO. There is no special AI Mode tag. There is no required llms.txt file for Google Search. There is no requirement to break every page into tiny chunks. There is no promise that structured data alone gets you cited. The foundation is still the boring part many websites neglect: crawlability, indexing, canonical consistency, clean internal linking, readable HTML, strong page experience, useful media, accurate metadata and policy-compliant content.

That should be good news for serious operators. It means the basics still compound. It also means technical debt becomes more expensive. If important pages are blocked, duplicated, slow, orphaned, thin or buried behind poor JavaScript rendering, they are less likely to participate in any search experience, AI or classic.

For WordPress-based SMEs, the most common failures are practical: too many weak category/tag archives, poor internal links to money pages, duplicate service pages, bloated themes, plugin conflicts, missing canonical discipline, weak headings, images without useful context, slow mobile performance and content written without a clear buyer journey. None of these problems are glamorous, but AI Mode does not make them disappear.

AYSA.ai should treat the technical layer as the entry ticket. The system should detect whether a page is eligible, understandable and connected before proposing fancy AEO or GEO improvements. A page that cannot be crawled does not need a thought leadership angle. It needs to be fixed.

The content layer: non-commodity beats volume

Google’s AI optimization guide emphasizes unique, valuable, non-commodity content. This matters because generative AI makes average content cheap. If your page is only a paraphrase of the top five results, it does not deserve to be a source. AI Mode can summarize the average. Your job is to provide what the average does not contain.

Non-commodity content can take several forms. It can be original research, as in a benchmark or dataset. It can be first-hand experience, such as what broke during an implementation. It can be a comparison framework that helps a buyer choose. It can be a case study with clear context. It can be a product teardown. It can be a pricing explanation. It can be a checklist based on real audits. It can be an expert point of view with operational consequences.

For the keyword “how does SEO work on AI Mode,” a commodity article would say: “AI Mode uses AI, so optimize for helpful content and schema.” That is too shallow. A non-commodity article explains the retrieval model, the role of fan-out, why exact-match keywords are weaker, how technical SEO still matters, how content must serve sub-questions, how to measure visibility when clicks fall, and what an SME should do in the next 30 days.

Content should also be internally connected. A single article cannot carry the whole strategy. This page should link naturally to AI search visibility, technical SEO, research, on-page SEO, off-page SEO, monitoring and the AYSA.ai registration flow. That internal architecture helps users and search systems understand what AYSA.ai actually does.

Measurement: stop judging AI Mode only by clicks

Clicks still matter. Revenue still matters. Leads still matter. But AI Mode can create value and risk before a click happens. A user may see your brand as a cited source, remember it, search for it later, compare it inside another AI tool or use the answer to shortlist vendors. A user may also see a competitor cited repeatedly and assume that competitor is the authority. If you only measure last-click traffic, you will miss both outcomes.

A modern AI search measurement stack should include classic SEO metrics and AI visibility checks. Classic metrics include impressions, clicks, CTR, indexed pages, rankings, internal links, Core Web Vitals and conversions. AI visibility metrics include whether the brand is mentioned for priority prompts, whether the description is accurate, which pages are cited, whether competitors dominate the answer, whether sentiment is positive or negative, and which content gaps appear repeatedly across ChatGPT, Gemini, Perplexity and Google AI surfaces.

This is where many SMEs struggle. They do not have time to manually test prompts, review pages, rewrite content, fix internal links and watch Search Console every week. They need a workflow. AYSA.ai’s role is to turn signals into action: identify the opportunity, prepare a recommendation, explain the reason, ask for approval and execute accepted changes.

The best KPI is not “we published 20 AI articles.” The better KPI is: how many commercially important pages became stronger, clearer, more useful, more connected and more likely to be selected as sources?

Marius Dosinescu view: AI Mode rewards operational SEO

After more than 25 years in SEO, ecommerce and digital products, I do not see AI Mode as the death of SEO. I see it as a pressure test. It exposes whether a company has real expertise, whether the site is technically healthy, whether the content helps buyers decide, whether the brand is understood consistently, and whether the team can execute improvements fast enough.

The old SEO mistake was to treat rankings as the goal. Rankings were never the final goal. The goal was profitable discovery. AI Mode makes that clearer. If a generated answer explains the market and cites three competitors, your problem is not that SEO died. Your problem is that your website did not become a strong enough source for the system to use.

For SMEs, the opportunity is practical. Do not try to out-publish large companies. Out-specific them. Use your customer questions. Use your implementation experience. Use your product knowledge. Use your mistakes. Use your comparisons. Use your local market understanding. Turn those into pages that help humans and AI systems understand why you are credible.

That is the reason AYSA.ai exists. Most companies do not fail because they lack another dashboard. They fail because recommendations do not become approved execution. SEO in AI Mode needs the full loop: monitor, diagnose, recommend, approve, publish, measure and repeat.

The 30-day playbook for SMEs

1. Map the real prompt, not just the keyword. Take a query like “how does SEO work in AI Mode” and list the underlying needs: definition, mechanism, examples, risks, measurements, actions and vendor implications.
2. Build one source page, not ten thin pages. Create a strong page that answers the core question and links to deeper supporting pages instead of generating doorway variations.
3. Fix technical eligibility. Check indexing, canonical tags, robots rules, internal links, mobile experience, structured data validity where relevant and snippet eligibility.
4. Add evidence. Include first-hand examples, data, comparison logic, case notes, product screenshots, methodology, expert commentary or original observations.
5. Connect the cluster. Link the article to commercial pages, product pages, help content, pricing, glossary terms and signup flows.
6. Measure AI visibility. Track priority prompts across Google AI surfaces, ChatGPT, Gemini and Perplexity. Record mentions, citations, competitors and accuracy.
7. Turn insights into approved execution. Use AYSA.ai to prepare changes, review them with human judgment and publish improvements that actually strengthen the site.

Conclusion: SEO still works, but the work is more serious

Google AI Mode does not remove SEO. It raises the standard for SEO. The page still has to be discoverable. The site still has to be technically sound. The content still has to be useful. The brand still has to earn trust. What changes is the route between the query and the click. AI Mode can expand the query, synthesize the answer and show supporting sources, which means weak pages may never be noticed even if they target the right keyword.

The answer to “how does SEO work on AI Mode?” is this: SEO works by making your website a better source for generated answers and a better destination for the users who need depth, proof or action after the answer. That requires technical clarity, non-commodity content, topical coverage, entity consistency, internal linking, measurement and execution discipline.

For SMEs, the winning move is not panic. It is focus. Build pages that answer real buyer questions better than generic content can. Add evidence. Connect the site. Measure visibility. Improve continuously. That is modern Search Engine Optimization (SEO), and it is exactly the kind of workflow AYSA.ai is built to support.

Sources and further reading

• Google Search Central: Optimizing your website for generative AI features on Google Search
• Google Search Central: AI features and your website
• Google Blog: AI Mode in Google Search
• Google Search Central: Search Engine Optimization (SEO) Starter Guide
• Google Search Central: Guidance on using generative AI content
• Aleyda Solis: Google AI Mode query fan-out
• Ahrefs: AI Overviews reduce clicks study update
• Semrush: Zero-click search market analysis
• Pew Research Center: Americans and AI summaries in search results