Google just confirmed what we've been telling clients for a year

‍

TL;DR

Google published its first official guide to AI search this week:

• AI search retrieves from the same index as classic search.
• If you don't rank, you don't get cited.
• The popular hacks (llms.txt, content chunking, AI-readable rewrites, manufactured mentions) don't work. Google said so directly.
• What wins is the same as it always was: a technically solid website, non-commodity content, specific positioning.
• Write for humans.
• One of our clients, Thermax, proves the point.
• 175 leads and ChatGPT/Gemini referrals in four months.
• AI search hasn't changed the playbook. It's made a broken website more expensive to ignore.

AEO and GEO have been everywhere on LinkedIn this year. And a lot of them have been screaming that "SEO is dead" and created a form of scarcity. Google just said they aren't real disciplines. Consultancies have built service lines around it. Vendors have shipped tools. The pitch is always the same. AI search has changed everything, traditional SEO is dead, and you need a new playbook, whitch they happen to sell.

This week (15th of may) Google published its first official guide on optimizing for generative AI features in Search. The headline finding, in their own words: optimizing for generative AI search is optimizing for the search experience, and thus still SEO.

‍

What Google actually said

Their guide makes four points worth reading carefully, because the entire AEO/GEO discourse rests on the opposite assumption.

AI search is built on the same ranking system as classic search.
AI Overviews and AI Mode use retrieval-augmented generation. Google's existing ranking system retrieves the pages. The AI generates a response from them. If you don't rank, you don't get retrieved. If you don't get retrieved, you don't get cited. The path to AI visibility runs through the same index it always did.

The content principle hasn't changed. It's gotten stricter.
Google's word for what wins is non-commodity. Not "long-form." Not "10x content." Non-commodity. Their own example: a "7 Tips for First-Time Homebuyers" article is commodity. "Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line" is non-commodity. The first one any AI can produce in three seconds. The second one requires that you actually waived an inspection, lost money on a sewer line, and learned something from it.

The popular hacks don't work.
Google explicitly tells you to ignore them. llms.txt files: not used. Content chunking for AI readability: not needed. Rewriting copy in some special AI-friendly format: not needed. Manufactured "mentions" across the web to influence what AI says about you: violates spam policy. Special schema markup for AI: not required.

The technical foundations are the same.
Indexable. Crawlable. Fast. Semantic HTML where it helps humans. JavaScript that doesn't block content. None of this is new. All of it is still the foundation everything else depends on.

That's the whole guide. If you've been doing real SEO, none of it surprises you. If you've been selling AEO retainers, parts of it should sting.

‍

Why so much of this industry exists anyway

Because new acronyms and scarcity sells. That's the honest answer.

When ChatGPT and AI Overviews emerged, two things happened at once. Real referral traffic from AI surfaces became measurable, which is genuinely new and genuinely matters. And an entire layer of the marketing services economy realised it could re-package existing SEO work under a fresh label and charge more for it. We can say that AI serach is real, but all new sayings about it, its just repackaging.

The tell is in the objects. llms.txt is the clearest example: a file format proposed by one person, adopted by zero major AI systems, and now sold as a critical AEO deliverable by agencies who'd rather invoice for "AI infrastructure setup" than have the harder conversation about whether the client's content is any good. Google has now explicitly said llms.txt does nothing for Search. Anthropic doesn't use it. OpenAI doesn't use it. It is a file you create so you can tell a buyer you did something.

Same story with chunking. Same story with "AI-readable rewrites." Same story with the cottage industry around tracking which LLMs cite which brands and selling reports on it. These are activities that look like work, that produce dashboards, and that very rarely change a single revenue number for a B2B company.

‍

The thing nobody says out loud

The AEO/GEO industry exists because nobody wants to tell a CMO that the website is the problem.

It's awkward. The website was somebody's project. Somebody approved the agency. Somebody signed off on the redesign two years ago. To say "the foundation is broken" is to say something uncomfortable about decisions that have already been made and people who are still in the room.

So the conversation moves to a layer above the website, where everything is new and nobody is accountable yet. AI search. Schema. llms.txt. These are safer topics. They don't require admitting that the foundation isn't load-bearing.

‍

What we've watched happen when companies fix the foundation

Thermax is one of our clients. They're a Norwegian fire safety company. Their website was effectively invisible. Almost no traffic. No leads being tracked. We rebuilt it. Not as a marketing project, but as infrastructure. Pages organised around what a building manager would actually search for. Technical foundation tightened. Content rewritten for the buyer.

Four months later: 9,113 users. 175 qualified leads. And referrals from ChatGPT and Gemini, without doing a single thing on the AEO/GEO playbook. No llms.txt. No chunking. No AI optimisation. The site is good, so AI assistants retrieve from it. That is the entire mechanism Google described in their guide, working as advertised.

Thermax isn't an isolated result. It's what happens when a B2B company stops paying for a layer above the website and pays for the website itself.

‍

What this means if you're a B2B company right now

If your sales team is struggling to find pipeline, do not start with an AI search strategy. Start with the website and its foundation, structural and technical.

Open Search Console. Are your pages indexed? Are they being crawled? Are they slow? Are the pages your buyers would actually search for among the ones that exist?

If the answer is no to any of that, every penny you spend on AEO retainers, content programmes, paid media, or anything else is being spent on top of a foundation that doesn't work. This is what we mean when we say your website owes you money. AI search has changed the playbook, but not in a "SEO IS DEAD"-way. It's made your website more expensive to ignore.

If you want to read more about googles guide, you can find it here: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide

Ish Studio is a B2B growth and go-to-market agency in Oslo. We design, build, and operate websites for b2b companies in Europe.