More and more people no longer ask Google. They ask ChatGPT, Gemini, Copilot or Perplexity, get one answer instead of ten blue links, and inside that answer a handful of companies are named. If yours is not among them, you do not exist for that searcher.
The discipline that deals with this is called GEO: Generative Engine Optimization. This article explains what it is, how it differs from classic SEO, what AI search engines actually weigh, and what you can do today without opening a new budget.
What is GEO?
GEO is optimizing your website and your online presence so that generative AI systems understand, trust and cite your content in their answers. Where SEO is about a position in a list of results, GEO is about being mentioned in a composed answer.
That difference is bigger than it looks. A search engine shows ten results and lets the user choose. An AI assistant reads dozens of sources, weighs them and writes one answer that names perhaps three of them. Competition for those three places is fiercer, but so is the reward: whoever gets named inherits the trust of the answer.
The terminology is messy in practice. You will also see AEO (Answer Engine Optimization), AI SEO and LLM SEO. They describe the same goal from slightly different angles. In this article we use GEO as the umbrella term.
How AI search engines choose their sources
Nobody outside OpenAI or Google knows the exact formula, but the behaviour has been mapped well enough by researchers and by people who test it daily. Four patterns keep coming back.
They read, they do not rank. A language model usually first retrieves a set of pages through an ordinary search engine and then reads them. If your page is not near the top of a classic search, it rarely ends up in that reading set. Good SEO is therefore not an alternative to GEO but the ticket in.
They cite what is clear. A page that answers the question directly in its first paragraph, with concrete figures and plain definitions, is cited more often than one that gets to the point after five paragraphs of atmosphere. The model is looking for text it can almost quote.
They trust consistency. If your company name, address, services and price level are the same everywhere, on your site, in the Chamber of Commerce register, on LinkedIn, in reviews, the model dares to name you. Contradictory information makes a source unreliable and therefore unusable.
They like structure. Structured data (schema.org), clear headings, tables and lists of frequently asked questions help the model understand your content without guessing. What works for a screen reader usually works for a language model.
Why this matters now for Dutch businesses
Google increasingly shows AI overviews at the top of Dutch results. Whoever is named in them gets the click or the mention; whoever sits underneath loses part of the traffic that used to arrive on its own. At the same time, business decision-makers increasingly put their exploratory questions straight to an assistant: “which agency in Brabant builds web shops with iDEAL and Mollie” used to cost ten tabs and now yields one answer.
The market is also young. Most competitors have not done anything about it yet. That is exactly the moment when a mid-sized company with a well-maintained site can overtake larger players.
What you can do in practice
1. Answer the question in the first two sentences
Open every important page with a direct definition or a direct answer. What you do, for whom, in which region, at what price. The explanation and the nuance come afterwards.
2. Publish real numbers
Price ranges, delivery times, project durations, quantities. Language models much prefer citing specific facts over general promises. “Websites from €3,450, delivered in two to four weeks” is usable; “custom work at a fair price” is not.
3. Make your entity recognisable
Make sure your company appears everywhere with the same name, the same address and the same description. Add Organization schema with logo, contact details and service area, and link to your official profiles. That is how a model knows the source on LinkedIn and the source on your site are the same company.
4. Add an FAQ to substantive pages
Questions and answers in the form people actually ask them, with FAQPage schema underneath. This is the content that most often reappears verbatim in AI answers.
5. Get mentioned by others
Mentions in industry directories, customer reviews, guest articles and local media count as confirmation. A model that encounters your name in five independent places names you with more confidence than a model that only knows your own site.
6. Keep the technical foundation in order
Fast loading, a clean sitemap, no blocked crawlers and correct hreflang tags for your Dutch and English pages. Also check that your robots.txt does not accidentally shut out the crawlers of OpenAI and Google.
What GEO is not
It is not a trick and not a separate channel. Anyone who tries to mislead models with hidden text or fake reviews is punished as quickly as in classic SEO. Nor is it a replacement for your website: the model still sends the searcher to a page in the end, and that page has to convert.
Frequently asked questions
Is GEO different from SEO?
It is an extension. The same foundations, findability, authority and clear content, still apply. GEO adds the question of whether a language model can understand your content and cite it safely. Whoever ranks well but writes vaguely gets found but not named.
Can I measure whether I appear in AI answers?
Partly. Google Search Console does not yet report traffic from AI overviews separately, but you can periodically ask a fixed set of questions to ChatGPT, Gemini and Perplexity yourself and record whether you are mentioned. These assistants also show up more and more often as referring sources in your analytics.
Do I need to write separate content for AI?
No. Content that gives a human a clear answer quickly is exactly what a model needs. Write for your customer, structure for the machine.
How long before it works?
Technical and structural changes can take effect within weeks in Google’s AI overviews. Mentions in models that do not search live, such as some ChatGPT answers, follow the training cycles and can lag by months. Expect an investment that pays back over quarters, not days.
Where to start
Ask ChatGPT and Gemini the five questions your customers ask you most often and see whether you are named. If you are not, you immediately see who is, and why their pages are clearer than yours. That is your to-do list.
We build websites with this structure as the starting point and carry out GEO as part of our SEO and content service. If you are building AI applications yourself, we also help you become visible in the systems your customers use. Tell us what your customers are searching for and we will show you where you stand today.
