SEO for AI: how to get found in ChatGPT and AI Overviews

You see the number of visitors in your stats slowly dropping, while your phone keeps ringing as usual. Strange? Not really. More and more people nowadays ask their question to ChatGPT, Gemini or Perplexity instead of to Google. They get a ready-made answer there and only click through when they really want to go further.

That changes the question you need to ask yourself. Not just: do I rank high in the search results? But also: am I mentioned when an AI tool puts together an answer? That calls for a different approach, with a name of its own: SEO for AI, also known as Generative Engine Optimization.

In this article, you will learn why AI search engines choose different sources, what you can arrange technically and in terms of content, and how to measure whether it works.

What is SEO for AI?

SEO for AI is optimising your website and your brand so that AI tools use your information and mention you as a source. Think of Google AI Overviews and AI Mode, ChatGPT, Gemini, Perplexity and Bing Copilot. So you are not only optimising for a spot in a list of links, but for a spot in the answer itself.

The abbreviations fly at you: GEO, AEO, AIO and SXO. We explain what they mean exactly and how they relate to each other in our article on SEO vs GEO. Here it is purely about the practice: what do you do on Monday morning?

Why AI tools choose different content

• Training data or live search results: roughly speaking, AI tools work in two ways. Some answers come from the model’s training data: everything it has seen of the internet during training, without any up-to-date check. Other tools combine the language model with live search results and then refer to concrete sources.

  That difference determines your approach. For the first group, what counts is whether your brand appears often and in the same way across the internet. For the second group, what counts is whether your page is the clearest and most readable source at the moment the question is asked.

• Repetition and recognition: AI models choose sources based on repetition and recognition. A company that is described in a similar way in ten different places gets picked up sooner than a company that only appears on its own website. See it as a job application where the references have already been called: the model wants confirmation from outside before it mentions your name.

Technical SEO: make sure AI tools can reach you

Before an AI tool can cite your content, it first has to be able to reach it. Two things determine that: do the crawlers fetch your page, and do they understand what it says?

Let the crawlers in

This sounds like an open door, but surprisingly often things go wrong here. AI crawlers are a lot less patient than Googlebot. They struggle with slow pages and with websites where the content only appears after JavaScript has loaded. So go through this list:

• Your robots.txt: are you blocking AI crawlers such as GPTBot, PerplexityBot and ClaudeBot? Sometimes that is still in there from an earlier decision. Fine if that was deliberate, annoying if nobody remembers anymore.
• Your loading time: a slow page gets skipped more often. This is exactly the same win you also get for regular visitors.
• Your HTML: is your text simply in the source code, or does a script have to run first? If you work with a headless setup, make sure you use server-side rendering.

Structured data and a logical structure

With structured data you tell machines who you are, what you offer and how the parts of a page relate to each other. Structured data is a piece of code that helps search engines and AI systems understand the context of your page precisely, without having to guess it from the running text.

The same goes for your heading structure. AI systems look for hierarchy and for context in bite-sized pieces. A page with one H1, logical H2s and short paragraphs under each heading is simply easier to process than three screens of continuous text.

Content that AI tools love to cite

The technique is half the job. The other half lies in how you write your text. Three things make the difference between a page that gets cited and a page that gets skipped.

Give the answer first

The structure that slowly takes a reader towards a conclusion works against you. Put the answer in the first two sentences under the heading and only explain it afterwards. That way, a model can literally take over that piece without having to summarise your whole article first.

So formulate your headings as the question someone really types in. Not “Our approach”, but “How long does a migration take?”.

Be specific, with figures and sources

Pages with concrete figures are more often chosen as a source by generative search engines. Vague marketing language works against you: there is nothing to cite from it. So mention amounts, turnaround times, percentages and years, and state where they come from. Factual accuracy and up-to-date information make your content more trustworthy in the eyes of a model.

Add your own experience to that. That is exactly what AI models themselves cannot access, and it is the reason why E-E-A-T carries so much weight.

Build clusters, not loose articles

One article about a topic does not make you an authority yet. A group of articles that together cover a topic and link to each other does. Such content clusters increase the chance that an AI system sees your site as a reliable source on that theme. Internal links are not an SEO trick, but the way you show how your knowledge hangs together. You can read more about that in our article on semantic SEO.

Make sure your brand gets mentioned outside your own site

This is the part most companies skip, while it may well weigh the heaviest. You can set up your own website perfectly, but a model that mainly draws on training data never even comes by. It knows you from what others write about you.

So work on mentions on external platforms:

• Trade media and industry associations: a guest article or an interview in a publication that is read in your industry.
• Review platforms: real customer reviews on platforms that AI tools use as a measure of trust.
• Forums and communities: platforms like Reddit are often cited in AI answers, because that is where people share experiences in plain language.
• Podcasts and events: less measurable, but it does deliver the repetition that models pay attention to.

And no, link building is not dead. Backlinks remain an important signal for the authority of your domain. The difference is that a mention without a link also counts nowadays, as long as your brand name is clearly stated.

How do you know if it works?

This is where things go wrong for many companies. Your trusted reporting does not show this. You have to measure visibility in AI separately from your regular positions. Three ways that do work:

1. A prompt tracker. With it, you track per prompt whether your brand name appears in the answer, in which model, and how you compare to your competitors. Our own label Upendo does exactly this.
2. Your server logs. See which AI crawlers visit your site and which pages they fetch. That tells you directly whether you are technically reachable.
3. Webmaster tools. Bing Webmaster Tools has an AI report in preview, which shows how you appear in AI answers within Bing and Copilot.

Combine that data with your regular SEO reporting. Only then do you see the full picture: where you are losing traffic, and where you are gaining visibility without a click in return.

Checklist: is your site ready for AI search engines?

Go through these points. If you can tick them all off, the basics are in order and you can work on the rest.

Technical

• Your robots.txt does not block AI crawlers, or does so deliberately.
• Your most important text is in the HTML and not only after a script.
• Your pages load fast, also on mobile.
• Structured data is on your organisation, your services and your articles.

Content

• Your headings are formulated as the question someone really asks.
• Under each heading, the answer is in the first two sentences.
• There are concrete figures, amounts or turnaround times in it, with source references.
• Your articles on the same topic link to each other.
• There is your own experience in it that a model cannot get anywhere else.

Brand and authority

• Your company is described in the same way on multiple external platforms.
• There are up-to-date reviews on the platforms that count in your industry.
• You collect mentions in trade media, not just backlinks.

Measuring

• You track a fixed set of prompts you want to be mentioned for.
• You know which AI crawlers visit your site and which pages they fetch.
• Your reporting shows AI visibility and regular positions side by side.

The bottom line

SEO for AI is not a completely new discipline. It builds on exactly the same foundations: a technically healthy site, a clear structure, strong content and a brand that others talk about. What changes is the balance. Clear answers, concrete figures and mentions outside your own domain weigh heavier than ever.

The good news? Most companies are not working on this yet. Whoever starts cleaning up, structuring and measuring now builds a lead that is hard to catch up with.

Curious whether your brand is already mentioned by AI tools? We would love to take a look with you and show you where the opportunities lie.