Google has published its official guide on AI search optimisation. And their position is noticeably different from much of the marketing around “AEO” (Answer Engine Optimization) and “GEO” (Generative Engine Optimization).
The short version: do the foundational work well. The rest is secondary.
That’s good news for local businesses who’ve been offered “AI SEO packages” or felt worried they need to hire expensive specialists. You probably don’t need a new service. You probably just need to do what you’re already doing, slightly better.
Here’s an interpretation of the guide from a local business perspective.
What Google says in three sentences
-
The SEO foundation still applies. AI features in Google Search use the same ranking systems as regular search, supplemented with new techniques like retrieval-augmented generation (RAG) and query fan-out.
-
Quality and perspective win. Unique first-hand insights beat commodity content like “10 tips for X”.
-
Technically: be indexable, crawlable and snippet-friendly. None of that is AI-specific, it’s regular SEO.
What Google explicitly says you DON’T need
For local businesses considering investing in “AI SEO services”, this is particularly worth reading.
You don’t need llms.txt for Google
Google’s own AI doesn’t use it. Nor have we found documentation from OpenAI, Anthropic or Perplexity saying their search services read it (as of October 2026).
Practically: if you already have an llms.txt, don’t remove it. It costs nothing to keep. But don’t invest money in “optimising” it for Google.
You don’t need structured data as a requirement
JSON-LD is valuable, it can give a page rich results and better snippets in regular search. But it isn’t a requirement to show up in AI answers.
Many agencies sell schema markup as an “AI feature”. That’s misleading. Schema markup is useful for traditional SEO. The AI benefit is secondary.
You don’t need to write “for AI”
Google’s guide says, verbatim: “You don’t need to write in a specific way just for generative AI search.” Write for readers. AI models are good enough at understanding normal prose.
That goes directly against “AEO copywriting” services promising to write in ways “AI understands better”. According to Google itself, it isn’t needed.
You don’t need to break content into thin chunks
There’s no magic paragraph length. “AI-optimised chunking” is pseudo-science. Write natural sections with clear headings, that’s enough.
You don’t need to manufacture brand mentions
Invented mentions and fake citations are no shortcut. Google’s guide says that seeking inauthentic mentions isn’t as helpful as it might seem.
What Google says you DEFINITELY should do
Write uniquely and from a perspective
First-hand insights beat commodity content. For local businesses, that means:
- Share cases from your own customer projects
- Tell about a specific renovation you did
- Explain why you charge what you charge
- Describe what your area looks like, what customers you work with, what works in Uppsala but not in Stockholm
Generic “10 tips for choosing a hairdresser” lists are worthless. A specific story from your business is valuable.
Be indexable, crawlable, snippet-friendly
Boring but decisive. A lot of basic SEO that local businesses actually miss:
- The page must be indexed in Google (check Search Console)
- robots.txt must not accidentally block Googlebot
- The page must load fast (LCP of 2.5 seconds or less on mobile, according to web.dev)
- Mobile-friendly design that doesn’t lose visitors
This isn’t AI-specific. It’s the prerequisite for being included in AI answers at all.
Use Google Business Profile
Our assessment is that GBP is one of the most important levers for local AI search. When someone asks Google’s AI about “good hairdresser in Vasastan” or “dentist near me”, the answer is about details such as name, address, opening hours and reviews, and you control those in GBP.
That means:
- Verified profile with the right category
- 20+ photos showing what you actually do
- Current opening hours, including for holidays
- Regular posts (one per week is a good rhythm)
- Active review work
All of it is free. That’s the whole point.
Allow crawling
Nothing mysterious. Allow Googlebot in robots.txt. And if you also want to show up in search answers in ChatGPT, Perplexity and Claude, allow OAI-SearchBot (ChatGPT’s search feature), PerplexityBot, Claude-SearchBot and Claude-User.
GPTBot and ClaudeBot collect material that may be used to train the models. OpenAI says the settings for GPTBot and OAI-SearchBot are independent of each other, so you can opt out of training and still be able to appear in ChatGPT search. Google-Extended controls whether content may be used for Gemini and, according to Google, doesn’t affect whether or how the page appears in Google Search.
That’s two minutes of work with big upside.
What this means concretely for local businesses
Three practical interpretations.
1. Pay for the work, not the label
If an agency sells “AI optimisation” as a separate service, ask what they do. If the answer is “schema markup, llms.txt, AI-optimised copy and chunking”, that’s the same thing as regular SEO with a new label.
Invest instead in a good website + good GBP + honest content production. That’s what Google says works.
2. Invest in Google Business Profile
For local businesses, GBP is the single biggest lever for visibility in AI answers. It’s also the biggest for traditional local search. Two benefits from one investment.
We go deeper on this in Google Business for local businesses.
3. Keep focus on local signals
Local search is how Google’s AI answers local questions. NAP consistency (Name, Address, Phone), local reviews, local content production. That’s where the lever sits.
We go deeper on this in Visibility on Google in your city.
The difference between Google’s AI and third-party models
An important nuance in Google’s guide: it’s specifically about Google’s own AI (AI Overviews / Gemini). That isn’t the same as ChatGPT, Perplexity or Claude.
| Model | How they find information | What helps |
|---|---|---|
| Google AI Overviews / Gemini | Google’s current search index | Traditional SEO + GBP + local signals |
| ChatGPT | Training data + web search | Own site + allow OAI-SearchBot (GPTBot is for training) |
| Perplexity | Searches the web for every question | Own site + allow PerplexityBot |
| Claude | Training data + web search in certain contexts | Own site + allow Claude-SearchBot and Claude-User (ClaudeBot is for training) |
For Google’s AI, traditional SEO + local signals are enough. For third-party models, their search crawlers also need to be let in through robots.txt. We have found no documentation from OpenAI, Anthropic or Perplexity saying their search services read llms.txt (as of October 2026).
Practical conclusion: cover both. The cost is low, the upside of being visible in more places isn’t small.
How third-party models find and choose sources is covered in How your business shows up in ChatGPT and Optimise for Perplexity. The practical steps, with timing, measurement and cost, are in AI SEO in practice.
What you do, in three time windows
None of this is AI-specific. It’s the same thing as good local SEO. Which is exactly the point Google is trying to make in their guide.
What we take from this
Google’s guide confirms at the core what we’ve said all along at Synlighetsverket: visibility is about craft, not hacks. Clear text, alive website, correct business information, real reviews.
“AEO” and “GEO” as concepts exist. They’re sold as disciplines. But for local businesses in Sweden, they’re not the biggest lever. It’s local search + Google Business Profile + honest content production. And that, as always, can be done well.
The only new thing in the AI era is that the lever for good foundational work has grown larger, not that new levers have appeared.
How we work with this
At Synlighetsverket we build websites and digital presence based on the same principles Google itself advocates in their guide. It’s not a coincidence that it matches either, we’ve always worked from the position that good foundational craft beats technical tricks.
No guarantees of showing up in Google AI Overviews. No guarantees of showing up in ChatGPT or Perplexity. Just the conditions for showing up where it actually is possible to show up, using the same methods that build good regular SEO.
Frequently asked questions
Do I still need to think about “AI SEO” at all?
Yes, but not as a separate strategy. Think of it as “quality level of your regular SEO”. Good SEO for local businesses = good AI visibility. That’s what Google says about its own AI features.
My agency sells AI optimisation. Are they shady?
Not necessarily, it depends on what they deliver. If they optimise Google Business Profile, help with structured data, improve website speed and content, that’s valuable and helps both regular and AI search. If they just sell “llms.txt + chunking + AEO copy” as separate items, they’re not doing much good.
What do I do about the fact that Google constantly changes things?
Build on fundamentals. Good website, correct GBP, honest content production, reviews. The principles Google itself emphasises in their guide don’t change even if individual algorithms do. That’s why we always talk about craft instead of hacks.
Will AI Overviews replace regular Google search?
Probably not. AI answers are often shown as a complement, not a replacement. For many questions (especially local ones), people still click links to verify, compare and book. Local search will keep being central, but the result format is evolving.