A few weeks ago we had a first call with a prospect who was looking for a software partner. Nothing unusual about that, except when we asked how they'd found us, the answer wasn't "Google" or "a referral." They'd asked an AI assistant: "developer gezocht in Vlaanderen" looking for a developer in Flanders. The AI mentioned us, they visited our site, they got in touch.
We wrote a short LinkedIn post about it at the time, because it was the first time we'd been directly told that a lead had come through an AI channel. Everyone in our industry has been talking about ChatGPT, Claude, Gemini and Perplexity as new discovery tools for a while. This was the moment where "everyone talking about it" became "we just experienced it."
That single conversation is not, on its own, a trend. But it's a preview of one, and any business owner reading this should be paying attention to what it means.
What GEO actually is (in plain language)
GEO stands for Generative Engine Optimization. It's the practice of making your website understandable, quotable, and citable by the AI assistants that increasingly sit between people and the open web.
If SEO is about being findable when someone searches Google, GEO is about being findable, and mentioned by name, when someone asks ChatGPT, Claude, or Gemini a question your business could answer. Same underlying idea. Very different technical execution.
The reason it matters is that the way people search is quietly shifting. Google's own search results now include AI-generated summaries. ChatGPT has hundreds of millions of weekly users. Claude and Perplexity are growing fast in professional and B2B contexts. When someone in your target market asks one of these tools "who builds custom software in Belgium?", the answer they get, or don't get, is now part of your marketing surface area.
You don't have to believe AI search will replace Google to take this seriously. You only have to accept that it will keep growing as one of the ways people find businesses. And unlike Google, where you can buy your way to the top with ads, LLMs don't currently sell placement. Visibility there depends entirely on whether your site is legible to the AI.
Why traditional SEO isn't quite the same thing
There's overlap between SEO and GEO, a well-optimised site helps with both, but the emphasis is different.
Google indexes your pages, ranks them by hundreds of signals, and shows the ranked list. It rewards backlinks, page speed, structured content, and dozens of other well-documented factors.
LLMs read the web differently. When an AI assistant answers a question, it's drawing on a mix of training data, live retrieval, and structured information about entities. It doesn't rank pages against each other in the same way. It synthesises an answer, and it decides, through a much less transparent process, which businesses to mention by name.
The practical implications:
Structured data matters more than ever. Schema.org markup tells the AI who you are, what you do, and where. Without it, you're just prose the AI has to interpret.
Being clear about what you are beats being clever. LLMs pick up straightforward, factual descriptions. Cute marketing language is often skipped or misinterpreted.
Consistency across the web strengthens signals. Your name, address, description, and specialisation need to look the same on your site, your Google Business Profile, industry directories, LinkedIn, and everywhere else you exist.
Content that answers real questions gets referenced. Blog posts that clearly address specific problems ("what's a proof of concept?", "how do you rescue a stalled project?") are the kind of material AIs pull from.
A /llms.txt file is emerging as a lightweight standard. It's a simple text file that summarises what your site is and where to find its most important pages. It's designed specifically to help AI systems understand your business quickly.
What we did (and what actually seems to work)
We started taking GEO seriously roughly six months ago, well before the prospect story above. Nothing exotic, just a series of deliberate, boring choices:
We rebuilt our schema.org markup from the ground up. Organization, LocalBusiness, Person entries, all interlinked through @id references so an AI parser can walk the graph and understand who we are, where we are, and who works here.
We rewrote our pages to be plainly descriptive. Less marketing language, more concrete factual statements about what we do, since when, on what stack, in what region. Not exciting to read as a human. Very legible to a machine.
We added a /llms.txt file. A curated summary of our site with pointers to the pages we most want AIs to reference. It doesn't guarantee citations, but it gives them a clean starting point.
We made sure our off-site presence matched. Google Business Profile, Sortlist, LinkedIn, industry directories, all with consistent name, address, description, and specialisation. Contradictions between platforms weaken every signal.
We monitored, but didn't obsess. Periodically we ask the major AI tools questions that a prospect might ask, and see what they surface. Not to game the results, but to notice when the answers change and whether our positioning is being picked up correctly.
Then, one day, a prospect told us they'd found us that way. The point isn't that any single thing on that list is what did it. The point is that a six months of quiet, structural work made the site legible in a way that eventually paid off, through a channel we couldn't have optimised for in the same way we optimise for Google.
The honest caveats (because there are several)
GEO is genuinely useful, but the industry around it is already producing bad advice. Three things worth knowing before you invest heavily:
1. Nobody can guarantee citation.
Any agency claiming they'll "get you cited in ChatGPT" is either misleading you or lucky. LLMs synthesise answers from vast training data and live retrieval, no external party controls what they surface. What you can do is prepare your site so that when the AI does look for information in your space, yours is clean, structured, and easy to include. That's readiness, not guarantee.
2. The rules are still forming.
GEO best practices in 2026 will look different from 2027. llms.txt is a proposed standard, not a settled one. Schema.org keeps evolving. Anyone who tells you they've cracked the code is oversimplifying, because the code changes.
3. GEO doesn't replace SEO.
People still Google. Google search still drives the majority of web discovery. GEO is a complement, not a replacement, and the strong news is that most GEO improvements also help traditional SEO, because clean structured data and clear factual content benefit both.
What this means for your website
If you run a business with a website, and if any of your prospects might ask an AI assistant a question your business could answer, GEO isn't a niche concern. It's an emerging channel that will only become more important.
The good news: most of the work is unglamorous and stable. Get your schema.org right. Make your homepage clearly describe what you do and for whom. Keep your off-site presence consistent. Consider a /llms.txt file. Write content that actually answers questions instead of chasing keywords.
None of it is magic. All of it compounds. And every so often, a prospect will tell you they found you through a channel that didn't exist a few years ago, which is when the quiet work suddenly feels very worthwhile.
Curious how your site looks to an AI?
We help clients audit their websites specifically for AI search readiness, how legible the structured data is, how consistent the off-site presence looks, and where the practical gaps are. It's not a magic wand. It's a clear picture of where you stand and what would actually help.
