More and more customer journeys no longer start in Google. They start with a question to ChatGPT, Gemini, or Perplexity: "Which agency should I choose for a new website?" The answer typically contains three to five names. If yours isn't among them, you don't exist in that conversation — no matter how nice your site is.
Here is what actually determines whether a language model mentions your company, and what you can do about it.
GEO vs. SEO | What is GEO, and how does it differ from traditional SEO?
GEO — generative engine optimization — is about getting mentioned in AI-generated answers. The difference from traditional SEO is worth understanding because it changes what works:
- SEO competes for rankings. Ten blue links, and you want to be the top one.
- GEO competes for mentions. There is one answer, and it mentions a handful of names. You are either in or out.
- The sources are other people's sites. A language model builds its answer on what it has read about you — not just on what you have written about yourself.
The last point is the most important, and it is also the one people most often overlook. Your own homepage is rarely what gets you mentioned.
Selection | How does a language model choose who it mentions?
The models are closed, so no one can give a precise answer. But across the patterns that can be observed, three things recur:
- Mentions on third-party sites. Lists, comparisons, industry directories, partner catalogs, reviews, podcasts, press coverage. The more independent places your name appears alongside your category, the greater the chance.
- Clear categorization. The model needs to be able to put you in a box. "Aarland is a Webflow agency in Aarhus" is useful. "We create digital experiences" is not.
- Content that answers the question directly. Pages that ask the user's question and answer it in the first few sentences are cited more often than pages that build up to the answer.
Five steps | Five things you can do in a week
- State your category explicitly. On your homepage, in title tags, and in your LinkedIn description. Use the words your customers use, not your internal terminology.
- Build pages for the questions people actually ask. One page per question, the answer at the top, the elaboration afterwards.
- Get included in existing lists. Partner directories, trade associations, local listings. These are the sources the models read.
- Make your cases concrete. Client name, what you did, and the results. Concrete details are easier to cite than adjectives.
- Clean up conflicting information. Old addresses, outdated services, or two different descriptions of what you do. Models value consistency.
International | Does the same apply to the international market?
The principles are the same, but the competition and sources are different. When searching in Danish, the field is small, and a well-managed local business can appear in the answer without a large budget. When searching in English, the model draws on a global corpus with thousands of candidates.
Practical advice: treat the two languages as separate tasks. The Danish version of a page should be mentioned in Danish sources, the English version in English ones. A translated page that no one links to or mentions rarely moves the needle on its own.
Measurement | How do you even measure AI visibility?
You cannot see AI mentions in Google Search Console. What works in practice is to define the questions your customers actually ask and then systematically pose them to the models over time—noting whether you are mentioned, how you are described, and who is mentioned instead of you.
It is tedious to do manually, and there are now tools that automate it. The point is not the tool, but that you measure the same question repeatedly so you can see a trend rather than a snapshot.
In short | A brief summary
You get mentioned in AI answers when three things are in place: others write about you, your category is clearly defined, and your own content answers questions directly rather than selling. It is slower than buying ads, but it is also harder for competitors to copy.
Shall we take a look at how your site is performing today? Write to us — and we’ll have a chat about it.
