GEO for service providers: staying visible in AI

A commercial director at a medium-sized B2B company opens ChatGPT and types: “Which Dutch marketing agencies help technical service providers become visible in AI answers and search engines?” The model replies with three names, two short descriptions, and one link. Your agency is not among them. The client clicks through and schedules a meeting. Not with you.

Until three years ago, this conversation still happened on Google. If you had the right meta description, a few inbound links, and a blog post ranking for “AI visibility for technical service providers”, you had a chance of winning the click. That mechanism still works, but the volume is shifting. Gartner predicts that traditional search volume will drop by around 25 percent in 2026, replaced by traffic from AI chatbots and other virtual assistants. At the same time, 89 percent of B2B buyers say they use generative AI during their buying journey. For business services, where every tender starts with a research phase, our observations suggest that percentage is even higher.

What that means for your services and product: search engine optimisation is not going away, but a new layer is being added on top: Generative Engine Optimization, or GEO. The goal is the same: being found. The mechanism, however, differs. AI assistants do not read your content as a Google bot does. They interpret, cite, and synthesise. And if your site is not built for that, you become invisible to half the market.

How AI assistants actually read your site

When an AI assistant is asked a question about Dutch service providers, it roughly does the following. First, it has a training corpus it can cite directly. Your site is included in that corpus if many other sites on your topic link to your content. For more recent questions, it also has a retrieval layer, similar to a specialised search engine, that retrieves content and uses it to answer with source references.

Both channels differ from traditional SEO. For the training corpus, authority and citation density matter most. For the retrieval layer, what matters is whether your content is structured so a model can quickly extract an answer. And in both cases, something that was secondary in traditional SEO matters: whether your content answers natural questions rather than merely saturating it with keywords.

What GEO involves in practice

Five steps you can take this quarter as a service provider with forty to a hundred FTEs, in order of impact per hour spent.

Schema.org markup for your services and service delivery. Service, Organization, and LocalBusiness schemas on the relevant pages give AI assistants structured facts about what you do, for whom, and in which country. We used this as the foundation for our own sector pages: each page has a Service block with areaServed and priceRange. For your services, the minimum is: areas of expertise, target groups, region, opening hours, and contact options.

An llms.txt file at the root. Similar to robots.txt but aimed at AI assistants, this file provides a structured index of your most important content, with links to the canonical versions. AI models that respect the file get a clean entry point instead of having to laboriously parse your site. It is not yet universally supported, but the direction is clear and it takes twenty minutes.

Clear, semantic HTML. No more <div> soup. An <article> for a blog post, a <section> for substantive segments, an <h2> for specific questions you answer. This is not a GEO trick; it is simply good HTML, but AI assistants use structure to assess which parts can be cited as facts and which are marketing.

Content structured as answers to real questions. An article titled “GEO for B2B marketing” is not the same as an article titled “How does a technical B2B company become visible in AI answers from ChatGPT and Perplexity?” The second directly answers a question someone might literally ask an AI assistant. Write for that question, not for the keyword.

Make authorship and authority visible. AI assistants value who writes something. A blog post without an author is less citable than a post by a specialist with ten years of demonstrable expertise. LinkedIn profiles, mentions in industry commentary, professional literature: it all counts. For your services: make sure specialists publish under their own names and that their profile pages connect to the rest of the web.

Where services companies get GEO wrong

Two patterns we often see among service providers looking to take on GEO work.

The first is that marketing and knowledge management remain disconnected. Marketing writes short, SEO-optimised texts. Specialists write substantive memos and opinions that never reach the site, or are only turned into generic web copy that loses their precision. For GEO, that is suboptimal: AI assistants are looking for substantive authority, not a bland summary. The best thing you can do is let your specialists write, with light editing, rather than paraphrasing their work.

The second is that service providers’ websites are often built around service overviews, while clients want AI assistants to address their concrete situations. “What to do if…” and “When is…” perform considerably better than “Our service X”. A sector-based structure containing concrete question-and-answer content keeps pace with how people search today.

What it delivers

GEO is not a magic growth button. It is an investment that will matter more each year, and one you can start small with now. For a service provider with forty to a hundred FTEs, we typically mean twenty to forty hours of internal work plus two to three weeks of external support to establish the foundation. After that, an editorial rhythm of one to two posts a month keeps it going.

Measuring results is harder than in traditional SEO because attributing click behaviour to AI assistants is still immature. What you can measure: whether your content appears as a citation in AI answers about your field. Tools such as Profound, Similarweb, AthenaHQ, Scrunch AI, and Peec AI are moving in that direction, but the category is young. The most important signal remains examining your own website performance with platforms such as Google Analytics and Bing Webmaster Tools, and continuing to optimise it.

A first step that works

If you do one thing this quarter, I would recommend this: take your five most requested services and rewrite the page titles from “Our service Y” to “What should you do if…” or “When is…” questions. Make sure the body answers that question with a concrete framework, an example, and the implications for your clients. Add schema.org markup. Link to the specialist who wrote it.

That is not a complete GEO strategy, but it is a test of whether the approach works for your services. If you are then ready for a structured, continuous approach, that is the moment to continue.

For the broader context: Marketing & Branding is the area of expertise in which we handle this work, and Business Services is the sector where this dynamic is most pronounced. The AI Readiness Scan maps where your site stands now and what would deliver the most value to tackle first for your specific business.

A note on how this was made: our content is AI-assisted. We use AI for research and drafting a first concept. A human reviews and edits the text and takes responsibility for everything we publish. Read how we make our content.