Why ChatGPT never mentions my company even though my website ranks well ?

A company can hold the top spot on Google and still be completely absent from ChatGPT’s answers, because a generative engine does not rank pages, it cross-checks what the web says about a brand. Ranking proves that your website answers a query well. An assistant builds its answer differently: the model gathers several sources, mostly external ones, trade press, professional directories, review platforms, forums, videos, industry association pages, then names the brands those sources mention consistently. If nothing outside your own website talks about you, there is nothing to cross-check.

The study published by Ahrefs on 26 May 2025 , covering 75,000 brands, is clear on this: the number of times a brand is mentioned across the web is the factor most closely tied to its presence in Google’s generated answers, well ahead of inbound links. So your website is not useless, it is simply not enough on its own. It has to be accessible, structured and clear enough to be quoted, but what decides whether you get named is your presence across the wider ecosystem. Takaneo treats those two areas as one communication question, not as two separate services.

Why the question is coming up now, and especially in Luxembourg

The question rarely reaches a company through its marketing team. It arrives because a managing director asked an assistant something, saw three competitors named, and did not see their own company. In a few months the topic moved from curiosity to concern.

The GEM Luxembourg 2024/2025 report , presented on 16 October 2025 and relayed by STATEC, shows that 34% of early-stage Luxembourgish entrepreneurs consider artificial intelligence very important for their business, against a European average of 22.3%, which puts the country ahead of every other country surveyed. So the attention is there, but it almost always goes to internal use of the tools, and almost never to how those same tools describe the company to its customers.

Luxembourg adds a specific twist. The market is small, the volume of online content produced per sector is low, and a handful of local sources carries enormous weight in what a model can actually find. That is a problem for companies with no footprint, and a real advantage for those who move early.

A website only speaks for itself, and a model is looking for corroboration

A generative engine does not trust a page because it is well written. It looks for convergence: several independent sources saying the same thing about you, in the same words. An About page, however polished, provides one testimony only, and it is yours.

A mention in a trade article, an up-to-date listing in a professional directory, a detailed customer review, a quote picked up by an industry federation, a video in which your name is spoken: each of those traces counts for more, in the model’s eyes, than your own claim about yourself. The Ahrefs study from May 2025 indeed shows that the factors most closely tied to visibility in generated answers all sit outside the brand’s own website.

In other words, the question is not only whether your site is optimised, but what a credible third party has written about you in the past twelve months. For a great many Luxembourgish SMEs, the honest answer is nothing at all.

On-site optimisation is still the foundation, it is just not the deciding factor

It would be wrong to conclude that work on the website is pointless. Without a sound technical and editorial base, nothing surfaces. A site that AI crawlers cannot access, a page that takes eight seconds to load, contact details that contradict each other from one platform to the next, copy that circles a subject without ever answering a precise question, and you stay unusable even if twenty sources are talking about you.

Your website is what the model checks once it has picked up your name somewhere else. The two areas depend on each other, and handling them separately is expensive. According to the 2026 AI Visibility Index published by Semrush on 26 June 2026 , based on 126 million queries sent to AI engines, 81% of organisations running search and AI visibility in a single unified workflow report a rise in traffic or leads from AI platforms, against 36% of those managing the two in silos.

Your web agency cannot fix this alone

This is the uncomfortable part for an SME. The problem shows up on technical ground, but it is largely solved somewhere else. Press relations, speaking on sector topics, the accuracy and freshness of listings on third-party platforms, customer reviews, video content, consistency of the company name and positioning everywhere it appears. These are communication disciplines, not settings to adjust.

A web agency will handle the website side very well. A public relations agency will handle the mentions side very well. The trouble appears at the seam, when the two tell the same story in different words, because that consistency is exactly what a model is trying to detect.

Takaneo covers both sides, the website and what gets said elsewhere, with the same team. That is mostly what prevents the mismatch.

What this means in practice

The gap between visible companies and the rest is severe. The Ahrefs study from May 2025 found that brands in the top quarter for web mentions average 169 citations in Google’s generated answers, against 14 for the quartile immediately below, and that 26% of the brands analysed never appear at all. There is no gentle progression, there is a threshold.

In practice, the first step costs nothing. Ask three assistants the five questions your customers actually ask before buying, and note which companies get named and which sources get cited. In an hour you will have the exact list of places where you should exist. The second step is to pick three of those sources and work on them properly for a year, rather than spreading effort thinly everywhere.

Frequently asked questions

How can I find out what AI assistants currently say about my company?

Ask them yourself. It is the most reliable and the cheapest method available. Open three different assistants, ChatGPT, Gemini and Perplexity for instance, and ask the questions your customers ask before buying, without ever mentioning your own name. Note which companies appear, in what order, and above all which sources are cited beneath the answer. Repeat the exercise a few days later, because answers vary considerably. You end up with an honest snapshot of your visibility and, more usefully still, a list of the outlets models consult in your sector. That is where any serious diagnosis begins, well before paying for monitoring tools.

Do I need to rebuild my website to appear in AI answers?

Rarely. In most cases the existing site can be repaired: crawler access, loading speed, page structure, wording that answers real customer questions, consistency of identity details. A full rebuild makes sense when the technical base is too old to evolve, or when the positioning itself has changed. The real risk lies elsewhere. Rebuilding a beautiful website while assuming AI visibility will follow automatically does not work, because most of what models cross-check sits outside your own domain. The website is a necessary condition, not a complete strategy.

Does traditional SEO become pointless?

No, its role changes. Generative engines partly rely on existing search results to build their answers, so good rankings remain a useful signal. What has changed is that rankings alone are no longer enough: coming first on a query does not guarantee being named in a conversational answer on the same subject. The figures published by Semrush in June 2026 point the same way, since organisations handling search and AI visibility together clearly outperform those keeping them apart. The right conclusion is not to swap one discipline for the other, but to stop steering them as two separate worlds.

How long does it take to see results?

Expect six to twelve months for a noticeable change, longer if your company starts from zero external mentions. The reason is simple. Earning credible mentions means relationships, publications and public positions, and none of that can be bought in a week. Some effects come faster, particularly when a technical fix finally opens your site to crawlers, or when an incorrect listing is corrected on a widely consulted platform. Be wary of promises of results in thirty days: they usually rely on manipulations that work briefly, then vanish with the next model update.

Can a Luxembourgish SME compete with a large group on this?

Yes, more easily than in traditional search, and that is the good news. In a market the size of Luxembourg, the volume of available content per sector is low, which means a model has few sources to build an answer from. A handful of well placed mentions in the local business press, an industry federation, a respected sector directory and genuine customer reviews weigh proportionally far more here than in a saturated market. Takaneo sees that companies acting now are claiming ground that is still lightly contested, while the same effort will cost considerably more in two years.

What should I do if an AI states something false about my company?

Fix the source, not the answer. A model repeats what it finds, so outdated or wrong information almost always comes from an identifiable page: an old address in a directory, a dated press article, a listing nobody updated, a director who left years ago. Track down where the claim comes from, correct it at source, then publish clear, dated information on your own website to give the model something recent to cross-check against. Propagation takes time, usually several weeks. Takaneo recommends checking this twice a year rather than waiting for a customer to point out the mistake.

TAKANEO - Creative Communication Agency in Luxembourg