There's a lot of noise right now. AI Overviews. AI Mode. ChatGPT. Perplexity. And right behind the noise, a queue of people selling you a brand-new discipline with a brand-new acronym (GEO, AEO), and a brand-new invoice to match.
So here's the honest version, straight from the source: Google published its own guidance on optimizing for generative AI search, and the punchline is almost disappointing.
There is no secret new playbook. It's still SEO. You just have to actually run it properly.
Google's own words are that optimizing for its generative AI features is optimizing for the search experience, and that's still SEO. The AI features (AI Overviews, AI Mode) aren't a separate engine bolted onto Search. They're built on top of the same ranking and quality systems that have always decided who shows up. If your content earns its place in normal Search, it's eligible for the AI features. If it doesn't, no amount of "GEO hacks" will rescue it.
Let me unpack what that means in practice, what to ignore, and how you'll actually measure it.
First, how AI search actually finds you
Two mechanics are worth understanding, because they explain why the old fundamentals still win.
Grounding (RAG). When Google generates an AI answer, it doesn't invent it from thin air. It pulls relevant, up-to-date pages from its Search index, reads them, and builds the response on top of them, with clickable links back to the sources. That retrieval step uses the same ranking systems as normal Search. Translation: to be quoted by the AI, you first have to be findable and trusted by Search.
Query fan-out. The model doesn't just answer the question you typed. It quietly fires off a fan of related questions and pulls results for those too. Ask "how do I fix a lawn full of weeds" and behind the scenes it's also chasing "best herbicides for lawns," "remove weeds without chemicals," and so on. This is why genuinely useful, thorough content does well: it naturally answers the questions around the question, without you having to game anything.
What to DO (the part that moves the needle)
Content: this is the whole game
- Have a point of view. A first-hand review or genuine expert take beats a summary of what's already online. If a generative model could've written your page itself, it's not going to bother citing it.
- Write non-commodity content. "7 Tips for First-Time Homebuyers" is common knowledge anyone could produce. "Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line" is experience nobody can copy. Aim for the second kind.
- Structure it for humans. Clear headings, sections, and paragraphs. If a person can navigate it easily, so can the machine.
- Add real images and video. AI features surface visuals too, so good media is another way onto the page. Follow normal image and video SEO and you're already covered.
- Don't spin up a page for every keyword variation. Mass-producing near-duplicate pages to chase every phrasing trips Google's scaled content abuse policy and doesn't work anyway. Quantity isn't quality.
- Use the simplest test there is: "Would a visitor be genuinely satisfied after landing on this?" If yes, you're aligned with exactly what Google's helpful, people-first content systems reward.
Technical structure: make sure you're even in the room
- Be indexable. To appear in an AI feature, a page has to be indexed and eligible to show in normal Search with a snippet, per the Search technical requirements. No indexing, no AI visibility. Full stop.
- Be crawlable. The AI features lean on publicly accessible, crawlable content. Don't accidentally block the very pages you want surfaced.
- Use sensible, semantic HTML. Not for perfection's sake (the whole web is messy and Google copes), but because clean structure helps screen readers, AI agents, and everything in between.
- Mind your JavaScript. Google can read content rendered in JS as long as it isn't blocked, but JS-heavy sites are trickier, so follow JavaScript SEO basics.
- Give a good page experience. Fast, mobile-friendly, easy to tell the main content from the clutter.
- Reduce duplicate content so you're not wasting crawl budget on URLs you don't even care about.
- Verify the site in Search Console so you can actually see and diagnose issues.
Local & ecommerce: if it applies to you
- Keep your Google Business Profile current, because AI answers can pull in local business info.
- For products, keep Merchant Center feeds clean so your products can surface in AI responses and across Search.
What to IGNORE (Google says so directly)
This is the bit that saves you money and time. Google explicitly lists tactics you can skip for Search:
llms.txtand other "special" AI files. You don't need them. Google may crawl all sorts of file types, but that doesn't mean they get special treatment.- "Chunking" your content into tiny pieces. No required page length. Google understands multiple topics on one page just fine.
- Rewriting everything in some special "AI-friendly" style. The systems understand synonyms and intent. You don't need to capture every long-tail variation of how someone might phrase a search.
- Chasing inauthentic "mentions" across the web. Manufactured buzz isn't the lever it's sold as. The quality systems and the spam systems both still apply.
- Over-indexing on structured data. It's not required for AI features and there's no magic schema. Keep using it (it helps with rich results), but don't treat it as the secret key.
If someone's pitch leans heavily on any of the above, you now know more than they're hoping you do.
The line worth remembering: There is no GEO. There's just SEO that somebody rebranded to sell you a course.
How you'll measure it: the new Search Console AI reports
Here's the genuinely new development, and it changes the conversation with clients.
In June 2026 Google rolled out Generative AI performance reports inside Search Console: a dedicated view of how your pages show up in AI Overviews, AI Mode, and AI features in Discover. Until now this data was blended invisibly into your normal Search totals; now you can isolate it.
What you get:
- Impressions: how often your URLs appeared inside an AI feature
- Pages: which specific URLs are being surfaced
- Countries: where that visibility is coming from
- Devices: what people were on when your page appeared
- Dates: trends down to hourly, daily, weekly, or monthly
What you don't get (yet): no clicks, no click-through rate, no average position, and no query data. Google has said more metrics may come over time, but for now the report measures visibility, not value. Knowing you were shown 10,000 times without knowing how often you were clicked is half a story, so pair it with GA4 to see what actually lands on the site.
Two practical notes:
- Deduplication. If the same URL appears in both an AI Overview and a normal blue link on the same page, that's counted as a single impression, not two. So don't expect the AI numbers to stack neatly on top of your existing ones.
- There's also an opt-out toggle. You can now exclude your content from AI features without hurting your normal organic rankings. Most service businesses won't want to, since visibility is visibility, but it exists if a client has a reason.
One thing to flag for us here in SA: the rollout started with a subset of UK site owners and is expanding globally from there. So if you log into a client's Search Console and don't see the report yet, that's the rollout, not a problem with the site. In the meantime, the impression data is still feeding your overall performance report.
And to be clear: none of this tracks ChatGPT, Perplexity, or other external tools. Search Console only sees Google. Measuring visibility on the other platforms is a separate exercise.
The checklist
Print this. Run it on any site you care about.
Content
- Every key page offers a genuine point of view or first-hand experience
- No commodity "anyone could've written this" filler
- Clear headings, sections, and readable structure
- Relevant, high-quality images/video where they help
- No mass-produced near-duplicate pages chasing keyword variants
- Passes the test: "Would a visitor leave satisfied?"
Technical
- Pages are indexed and eligible to show with a snippet
- Nothing important is blocked from crawling
- Reasonable semantic HTML
- JavaScript content renders and isn't blocked
- Fast, mobile-friendly, clean page experience
- Duplicate content minimised
- Site verified in Search Console
Local / ecommerce (if relevant)
- Google Business Profile complete and current
- Merchant Center product feeds clean
Don't waste time on
-
llms.txt or special AI files -
Chunking content into tiny pieces -
Rewriting in a special "AI" style -
Buying or chasing fake mentions -
Treating structured data as a magic key
Measure
- Check the Generative AI performance report in Search Console (once rolled out to the property)
- Track impressions by page and country
- Pair it with GA4: impressions show visibility, GA4 shows what converts
- Remember: clicks, CTR, position, and queries aren't in the AI report yet
Bottom line
The brands that win in AI search are going to be the ones that were doing SEO properly all along: useful content, clean technical foundations, and a real reason to be cited. The AI layer didn't rewrite the rules. It just raised the stakes on getting the basics right.
So before you pay anyone for "Generative Engine Optimization," ask them one question: how is this different from good SEO? If they can't answer it honestly, you already have yours.
Resources
- Optimizing for generative AI features on Google Search (Google Search Central)
- Google Search Essentials
- Creating helpful, reliable, people-first content
- Google Search spam policies
- Introducing Search Generative AI performance reports in Search Console (announcement)
- Generative AI performance report (Search Console Help)
Need a hand running this checklist on your own site? That's literally what we do. Give me a shout.