A few years ago the search engine was the only door. Someone typed a question into Google, got a list of blue links, clicked. For search engine optimisation — SEO — everything was about ranking high in that list, because that’s where the traffic came from.
That’s still true — but it’s no longer the whole truth. More and more people — not everyone, but more and more — start their search somewhere completely different. They open ChatGPT and ask “which is the best hairdresser in Östermalm?”. They open Perplexity and ask “what’s the difference between a general practitioner dentist and a specialist?”. They see Google’s AI answer at the top of the search results and read it instead of clicking the links below.
And in all those cases they get answers. Answers that mention specific businesses. Answers that don’t mention others. And if your business isn’t mentioned where it should be mentioned, you’ve lost a customer you’ll never know you lost.
That’s what AI SEO is about — or more correctly answer engine optimisation. Making your business understandable and describable for the models that answer people’s questions.
It isn’t a trick — it’s a foundation
The first thing to say: AI SEO isn’t a set of tricks that fools AI models. Models like ChatGPT, Claude, Perplexity and Google’s Gemini can’t be fooled in the way older search engines could. They draw conclusions from the text in their available information, and build answers based on what’s clear and consistent across many sources.
That means the foundation for AI SEO is exactly the same thing as the foundation for good traditional SEO: clear language, clean structure, correct business information, and content that actually answers what people are wondering about. If you do the basics well, you also become easier for AI models to understand and cite.
The difference lies in a few specific technical and editorial choices that strengthen AI readability — not replacing the SEO foundation, but complementing it.
How AI chooses what to mention
To understand what matters, we need to look a bit at how AI models choose what to include in their answers. Simplified, it’s like this:
They’ve been trained on huge volumes of text from the internet, books, articles, forums. That training decides what the model knows about the world at a certain point in time. If your business didn’t appear in any source during the training period, the model doesn’t know you exist.
They also, increasingly, have real-time access to the web via tools and plugins — but not all, and not always. Perplexity is built from the ground up to search the web for every question. ChatGPT searches when the browsing feature is enabled or when the model judges that the question requires fresh information. Claude uses web search in certain contexts. Google’s AI answers (AI Overviews / Gemini) draw from Google’s current search index. That means even if a model didn’t “learn” about you during training, it can find you now — if you exist on places it pulls from and if the model being used is configured to search.
Many users, however, interact with AI models without web access — then the model relies entirely on its training data, and what wasn’t there is invisible.
They cite sources that seem credible and clear. A page with clear structure, clear language, named sender and consistent information is more likely to be cited than a page with messy layout, vague information or anonymous claims.
They prioritise fresh sources when relevant. If someone asks about current prices, opening hours or contact details, the model prefers a source that looks recently updated.
These aren’t rules. They’re tendencies. But they’re consistent enough to be worth understanding.
The central principle — write for understanding
If I could only give one piece of advice for AI SEO, it would be this: write so that it’s impossible to misunderstand what your business is, what it does, for whom and where.
It sounds obvious. It’s still something most websites miss. Many company pages start with a hero headline saying “We help you grow” or “Together we create opportunities” — beautiful, but saying nothing about what the business actually does. For a human who already knows who you are it works. For an AI model trying to describe your business, it’s information-poor — and you either get left out or described in a vague and useless way.
Good AI SEO text instead says things like: “We’re a hair salon in Vasastan, Stockholm, specialising in men’s cuts and beard trims. We’ve been at the same address since 2018 and booking happens via our website or Bokadirekt.” That’s a couple of sentences that give an AI model everything it needs to answer a dozen different kinds of questions about you — what kind of business, where, what services, how long, how to book.
This kind of clarity also helps people. AI readability and human readability converge. When you write so a machine understands, it also becomes easier for a stressed visitor on mobile to understand.
The technical foundation — structured data and llms.txt
Beyond the text itself, there are a few technical elements that directly help AI models interpret your site:
Structured data (JSON-LD) is invisible code on your page that describes what the page contains in a format machines easily understand. For a local business the most important thing is to have Organization, LocalBusiness or ProfessionalService schema with the correct name, address, phone, opening hours, and a FAQPage schema for your most common questions. It gives AI models (and Google) a structured key to the information.
llms.txt is a relatively new standard — a simple text file at the root of your domain (/llms.txt) intended as a condensed summary for AI models. It’s like an “elevator pitch” in plain text: who you are, what you do, where you are, what you offer, how to contact you. Not all AI models use it yet, but the cost of adding it is minimal and the upside is being prepared as the standard matures.
Robots.txt with permission for AI crawlers is something many miss. Some sites block (deliberately or accidentally) AI crawlers like GPTBot, ClaudeBot, PerplexityBot, Google-Extended. If you want to be cited you need to allow these to read your site. For most local businesses this is nothing other than a deliberate choice to exist — there’s no reason to be invisible to AI answers.
Semantic HTML — using <header>, <main>, <article>, <section> instead of just <div> everywhere — lets machines understand what is what on the page. It’s basic web development, but many modern sites generated by “drag-and-drop” tools lack it entirely.
Content focus — answer real questions
The biggest difference in how you write content for AI answer engines compared to traditional SEO is that you need to think in questions and answers, not in keywords.
Earlier SEO text was often written as: “We offer quality dental care in Stockholm with modern methods.” It’s a sentence stuffed with keywords — “dental care”, “Stockholm”, “modern methods” — but no real person would ask a question matching exactly that.
AI-friendly text is instead written as: “How long is the wait for an emergency examination? Usually the same day or the next — we always hold an emergency slot every morning.” It’s a concrete question and a concrete answer. When someone asks the AI “can I get an emergency dental slot in Stockholm today?” and the AI needs to build an answer, it can cite your business as a concrete example.
This is also why FAQ pages are so powerful for AI SEO. A well-written FAQ section is essentially training data for AI models — exactly the format they prefer to cite.
What you never should do
There are a few things sometimes marketed as “AI SEO tricks” that you should stay far away from:
Prompt injection — attempts to hide instructions on your web page meant to “trick” AI models into saying certain things about you. It’s destructive on several levels: it rarely works because modern models are robust against it, it risks being blacklisted by the platforms, and it’s intellectually dishonest in a way that contradicts the whole point of building credibility.
Fake citations and fabricated stats — making up numbers, customer quotes or awards to look more authoritative. It doesn’t work over time, because other sources won’t confirm your claims, and the consequence is that the model starts noticing that your site is unreliable.
Generating massive amounts of AI-written filler text to cover many keywords. It leads to pages poor in actual insight, which causes models to deprioritise your sources as they notice they don’t give valuable answers.
Promising guaranteed placement in AI answers — neither to yourself nor to customers. It can’t be guaranteed. What you can guarantee is that the technical and editorial conditions are in place — the actual outcome is up to the models.
How to test if you appear in AI answers
This is the simplest and most underrated step. Open ChatGPT, Claude, Perplexity and Google’s AI mode once a month and ask about your business.
Concrete questions to test:
- “What is [your business name]?”
- “Tell me about [your business name].”
- “What is a good [your industry] in [your city]?”
- “Which [service] can you recommend in [neighbourhood]?”
Note what shows up. Are you mentioned? Are you described correctly? Are you described at all? If you’re mentioned but the description is wrong — consider where the wrong picture comes from. Is it an old version of your website? A third-party description? An outdated review?
This is ongoing work, not a one-off project. AI models update their knowledge, get retrained, change sources. Whoever appears today can be invisible in six months if they don’t maintain the foundation.
Where this is going
It’s still early. The share of searches that start in an AI answer engine is still smaller than the share that starts in traditional search. For many local businesses, it’s still less important today than Google Business Profile.
But the curve is clear. The share is growing quickly, especially among younger users and within certain types of searches (recommendations, comparisons, “what’s the best…” questions). In five years it’s reasonable to imagine a significant share of local-business-relevant searches starting with AI instead of traditional search.
And then the businesses that have the foundation in place — clear text, structured data, consistent information, honest communication — have a head start they’ve built over years. That’s why it’s worth starting now, even if the effect isn’t visible immediately.
Questions we get about AI SEO
How quickly will I start showing up in AI answers? For an AI model to consistently cite your business, the information about you needs to exist on several sources — your own site, Google Business Profile, industry sites, possibly articles — in consistent form. For new businesses it takes months. For established businesses it goes faster when the foundation is in place.
Do I need a blog to appear in AI answers? Not necessarily. A clear FAQ section on your main site is often just as valuable as a blog, especially for local service businesses. The blog becomes more relevant if you often get complex questions that require longer explanations.
Does it affect the number of visits to my site if I appear in AI answers? Yes and no. Direct clicks from AI answers are often lower than from traditional search, because the AI has already given an answer. But the quality of the visits is higher — those who do click already have strong intent. And several AI answer engines (Perplexity, Google AI Overviews) include source links that drive direct visits.
Is AI SEO a replacement for traditional SEO? No. It’s a complement. For local businesses, Google Business Profile, local SEO and reviews continue to be the bedrock. AI SEO builds on top of that, to also cover the growing share of searches happening in AI interfaces.
How much does it cost to work with AI SEO? For the most part it isn’t a separate cost — it’s a quality boost to the foundation work you do anyway with a website and Google Business Profile. It’s about writing clearly, structuring the data correctly, and having honest and fresh information. It’s included in all our packages because we don’t see AI visibility as a separate service but as part of modern digital presence.
How we work with this
At Synlighetsverket we build AI readability into the website from the first pixel. Clear language about what you do, for whom and where. Structured data describing your business in a format AI models understand. A clean llms.txt giving a condensed summary. FAQ text that answers the real questions people ask.
No guarantees of appearing in ChatGPT or Perplexity — no serious player can promise that. Just the conditions so that, when these models build their answers, your business is possible to find, possible to understand and possible to recommend.
Book 15 minutes and we’ll review how you’re described today — and what would make the biggest difference to become visible also where people search with natural language instead of keywords.