Cats & Pigions
Google has now done something rather useful, if slightly inconvenient for the cottage industry of AI-search conjurers. It has published a straight answer. Not a mystical answer. Not a new acronym wearing a cheap suit. A straight answer.
Generative AI features in Google Search, including AI Overviews and AI Mode, are rooted in Google's core Search ranking and quality systems. They use retrieval-augmented generation, query fan-out, indexed pages, crawlable content, eligibility for snippets, and the same fundamental requirements that have always sat underneath the work.[1]
So yes, SEO still matters. In fact, it matters precisely because the generative layer is not floating above Search like some celestial librarian. It is attached to Search. It retrieves. It grounds. It fans out queries. It looks for useful, accessible, high-quality material that Google can already find, process, rank, and show.[1]
That should settle a few pub arguments, although of course it will not. There will still be somebody selling an llms.txt miracle package before the kettle boils.
The Litter Box
There is a particular smell that appears whenever a new search surface arrives. It is the smell of panic dressed as innovation. First the industry invents a new acronym. Then it invents a new checklist. Then somebody, usually with a landing page, a webinar, and a microphone clipped to a linen shirt, explains that everything has changed and you now need their proprietary framework to survive.
And then, rather inconveniently, Google publishes a document saying the foundations still matter.
Google's guide to optimising your website for generative AI features on Google Search is important because it does not say what the AI-search theatre wanted it to say. It does not say that a new file, a special Markdown layer, a novel schema type, a secret chunking pattern, or a little sprinkling of inauthentic brand mentions will make your site eligible for AI glory.[1]
It says, in plain enough English, that Google's generative AI features are rooted in its core Search ranking and quality systems.[1] That one sentence should be nailed to the door of every AEO presentation currently doing the rounds.
Because if AI Overviews and AI Mode are grounded in Search, then the work is not separate from SEO. It is SEO under a new retrieval surface. Different interface. Different answer format. Same underlying need for crawlability, indexability, usefulness, quality, and trust.
Google describes two mechanisms worth paying attention to. The first is retrieval-augmented generation, where the system relies on core Search ranking systems to retrieve relevant, up-to-date pages from the Search index, then uses information from those pages to generate a more reliable response with prominent, clickable links.[1] The second is query fan-out, where the model generates concurrent, related queries to gather more information around the user's original need.[1]
That is not magic. It is retrieval, grounding, expansion, selection, and presentation. The machine may speak in softer language than ten blue links, but it still has to fetch the world before it can describe it.
The point Google has now made explicit
Google's central claim is simple: SEO is still relevant for generative AI search.[1] Not because old SEO tricks should be dragged into AI Mode like furniture from a burning house, but because the serious part of SEO was never the tricks. The serious part was making useful material discoverable, understandable, trustworthy, and satisfying.
That distinction matters. A large amount of the modern SEO industry confused technique with purpose. It took useful practices, such as structured data, internal linking, clear headings, descriptive titles, crawlable pages, and content depth, and slowly turned them into ritual. The ritual then became the product.
Google's new guidance pushes back against that ritualism. A page must still be indexed and eligible to appear in Google Search with a snippet to be eligible for generative AI features.[1] Google's generative AI models use publicly accessible, crawlable content to learn patterns and provide relevant, grounded responses.[1] JavaScript content can be processed if it is not blocked, but JavaScript SEO is more complex and still requires the usual care.[1] Page experience, duplicate content reduction, crawl budget, technical requirements, and Search Console diagnosis remain part of the work.[1]
In other words, the shiny AI layer does not rescue a bad technical foundation. If Google cannot reliably find, fetch, render, understand, index, or serve the content, then waiting for AI Mode to compensate is not strategy. It is wishful thinking with a dashboard.
| Google's AI Search guidance | Practical meaning for the Canon |
|---|---|
| Generative AI features are rooted in core Search ranking and quality systems.[1] | AI visibility is not separate from SEO fundamentals. It inherits the same crawl, index, quality, and policy constraints. |
| RAG uses Search ranking systems to retrieve relevant pages from the index.[1] | Ranking and retrievability still matter. Being useful but invisible is not enough. |
| Query fan-out generates related searches around the user's original query.[1] | Pages must answer the real underlying task, not merely exact-match one phrasing. |
| A page must be indexed and eligible for a snippet.[1] | Snippet eligibility, indexability, and accessibility are AI Search prerequisites, not optional extras. |
| Google says there is no special AI markup or llms.txt requirement.[1] | New machine-readable rituals should not be treated as Google AI ranking requirements. |
Non-commodity content is now the line in the sand
The most valuable part of the guide is Google's phrase non-commodity content.[1] It is a wonderfully blunt concept. Commodity content is the soup. It is the endlessly recycled "7 tips" article, the paraphrased explainer, the generic buyer guide, the hollow overview that could have been produced by anyone because, in truth, it was produced by nobody in particular.
Google contrasts that with content that provides a unique point of view, first-hand experience, expert takes, original insight, and information that goes beyond common knowledge.[1] A first-hand review has a kind of evidential texture that a summary of existing internet noise simply does not have.[1]
This is where the AI-search conversation becomes less technical and more uncomfortable. Because it is easy to add a file. It is easy to add schema. It is easy to generate twenty pages targeting twenty fan-out variations. It is much harder to have a real point of view, a real field of experience, real data, real mistakes, real scars, and the capacity to say something that has not already been laundered through the great content machine.
Google is not merely saying "write better content", the most useless phrase in digital marketing. It is saying that, for generative AI Search, the material most likely to matter over time is content that people find unique, compelling, and useful.[1] That is a direct challenge to scaled mediocrity.
And Google is also explicit about the boundary. Creating separate content for every possible variation of how people might search, including fan-out queries, primarily to manipulate rankings or generative AI responses, can violate Google's scaled content abuse spam policy.[1] The machine can generate fan-out queries. That does not mean publishers should manufacture thin pages for every fan-out shadow.
There is always the useful version and the counterfeit version. The useful version is understanding the deeper information need and creating a page that genuinely satisfies it. The counterfeit version is spinning up hundreds of pages because a tool gave you a fan-out list and you mistook volume for substance.
Google has now made that distinction explicit.
The myth-busting section should kill several bad products
The guide's myth-busting section is the part that will age into a Canon reference because it names the things people are already selling as if they were obligations.[1]
Google says you do not need to create new machine-readable files, AI text files, special markup, or Markdown to appear in generative AI search.[1] It specifically names llms.txt and other "special" markup as something you can ignore for Google Search.[1]
That does not mean a clean documentation file can never be useful for other systems, users, internal workflows, or non-Google agents. But it does mean the Canon should stop treating llms.txt as a Google AI visibility requirement. If it is used, it should be framed as auxiliary documentation, not as a ranking or citation lever for Google.
Google also says there is no requirement to break content into tiny pieces for AI to understand it.[1] Its systems can understand the nuance of multiple topics on a page and show the relevant piece to users.[1] This is important because it corrects a slightly over-excited interpretation of retrieval and chunking. Yes, retrieval systems may operate over passages or sections. No, that does not mean a publisher should butcher a good page into tiny artificial fragments for Google's benefit.
The human reader is still the point. If shorter pages work for the audience and subject, use them. If longer pages work, use them. There is no ideal page length.[1]
Google further says you do not need to rewrite content in a specific way for generative AI Search, because AI systems understand synonyms and general meanings.[1] You do not need to capture every long-tail variation like some desperate fisherman casting nets into a puddle.[1]
And it says inauthentic mentions are not helpful because AI features depend on core ranking systems and spam-blocking systems.[1] This matters for brand and entity work. Real mentions, reviews, citations, coverage, and discussion are useful because they reflect the world. Manufactured mentions are not the same thing. They are trust theatre.
Finally, Google says structured data is not required for generative AI Search and there is no special schema.org markup for it, although structured data remains useful as part of ordinary SEO and rich-result eligibility.[1]
Schema is still a label on the bottle. It is not the wine.
Technical clarity still matters, just not as theatre
The interesting thing about Google's technical advice is that it is deliberately boring. That is not an insult. Boring is often where the truth lives.
The guide tells site owners to meet Search technical requirements, follow crawling best practices, make content crawlable, manage crawl budget for large and frequently updated sites, follow JavaScript SEO guidance, provide a good page experience, reduce duplicate content, and verify the site in Search Console.[1]
None of this is fashionable. None of it will give the conference circuit a dopamine hit. But it is the infrastructure that allows everything else to work.
Semantic HTML is treated with a similar level-headedness. Google says perfect semantic HTML is not required and that Google can understand imperfect HTML, while also saying semantic HTML is generally a good idea because it helps users such as screen reader users parse and navigate a page.[1]
That is the sane position. Use semantic HTML because it helps people and machines, not because you believe a validator score is a sacrament.
The same logic applies to images and video. Google says generative AI Search can bring in relevant images and video, creating more opportunities for visibility beyond ordinary page links.[1] If image SEO and video SEO are already being handled properly, those assets are already being optimised for generative AI Search as well.[1]
Again, the Canon point is not "invent a new AI image tactic". The point is to do the existing work properly: descriptive media, accessible pages, crawlable assets, useful surrounding context, and content that actually satisfies the user.
Local business and ecommerce data now belong in the AI Search conversation
Google's guide also brings local business and ecommerce data into the AI Search frame. Generative AI responses may include product listings, product information, and local business information.[1] Google specifically points to Merchant Center feeds and Google Business Profiles as ways to help products and services appear in AI responses and other Search results.[1]
This is not a small operational note. It means AI Search visibility is partly a data-maintenance problem. A business with incomplete product feeds, stale inventory, weak Merchant Center hygiene, poor Business Profile data, inconsistent local information, or neglected service details is not merely under-optimised for traditional Search. It is under-feeding the systems that may generate product and local answers.
For clinics, local service businesses, ecommerce retailers, and multi-location brands, this is ground zero. The AI layer does not remove the need for clean business data. It makes the consequences of messy data more visible.
Agentic experiences are adjacent, not the same thing
The final section of the guide touches on agentic experiences, including browser agents that may analyse visual renderings, inspect the DOM, and interpret the accessibility tree.[1] Google frames this as something relevant businesses may explore if they have extra time, not as the main path to AI Search visibility.[1]
That distinction matters. Agent-readiness is important, but it should not be collapsed into Google AI Overview optimisation. A browser agent trying to complete a task has different needs from a Search feature grounding an answer. Accessibility, DOM clarity, visual rendering, form usability, product data, and transaction flows become more important when agents begin acting on behalf of users.
The practical response is not panic. It is preparation. Make websites usable for humans first, because a site that is hostile to humans will usually be hostile to agents too. Use accessible markup, predictable navigation, clean product and service data, sensible forms, and pages that can be understood without interpretive gymnastics.
AI Search is not an escape from SEO. It is a new expression of the same unforgiving basics.
The people who were building useful, crawlable, original, technically clean, satisfying content have not been made obsolete. They have been vindicated.
The people selling shortcuts will, as always, keep selling shortcuts. That is the culture soup. New acronym, old behaviour.
But Google's guidance gives us a useful Canon line: if a tactic exists only to impress "AI" and does not make the page more useful, accessible, crawlable, trustworthy, or satisfying for a human visitor, it is probably theatre.
And theatre is not strategy.
