HomeInsights › Invisible or miscategorised?
Diagnosis & Sequencing

Invisible but understood, or visible but miscategorised? Why miscategorisation is the harder problem to fix

By Neil Harte, Founder · Genivista · July 2026 · 7 min read

The short answer

  • Two very different AI visibility problems: invisible but accurate (AI rarely mentions you, but gets you right) and visible but miscategorised (AI talks about you constantly, in the wrong box).
  • Both cost you deals. But miscategorisation is the harder, more expensive problem — and the one to catch early.
  • The reason is simple: it is two jobs, not one. You have to un-teach the wrong story and teach the right one. Being invisible is one job — build presence on a blank page.
  • The wrong story is also sticky. A confident belief persists even after correction (the continued influence effect), and AI systems only fully relearn a category on slow retraining cycles.
  • And you are already in the wrong conversation, which the model reinforces every time it repeats it. So: if miscategorised, fix accuracy before you amplify — never the other way round.

Ask an AI what your company does and you will get one of three answers: nothing, a clear and accurate description, or a confident answer that puts you squarely in the wrong category. Those last two — invisible-but-right and visible-but-wrong — are different problems with different fixes. Faced with a straight choice, most people assume invisibility is the worse place to be. In our experience it is the other way round: the wrong story is harder to fix than no story at all.

Two failure modes that look nothing alike

A brand can be barely present in AI answers yet described perfectly whenever it does appear. Its problem is discovery: the story is right, it just is not being told often enough. Another brand can be everywhere in AI answers and consistently mis-filed — grouped with the wrong peers, credited with the wrong offering. Its problem is accuracy: no shortage of story, but the wrong one. Same category of symptom, opposite diagnosis — and, crucially, opposite difficulty.

Why miscategorisation is two problems, not one

Start with the mechanics. To fix an invisible brand you build presence: get mentioned, accurately, in the sources AI trusts, and the entity forms on what is effectively a blank page. To fix a miscategorised brand you have to do that and first overturn a category the model already holds — with confidence. A blank page takes one pass. A page with the wrong thing written on it in ink takes two: erase, then rewrite. That second job, the un-teaching, is the hard one, and the invisible brand never has to do it.

Invisibility is a blank page. Miscategorisation is a page with the wrong thing already written on it — in ink. One takes a pass; the other takes two.

The wrong story is sticky

There is a well-documented reason corrections underperform. In human cognition it is called the continued influence effect: once a false belief is formed, it keeps shaping judgement even after it has been explicitly corrected. Simply stating the truth rarely erases the earlier impression. Miscategorisation behaves the same way — a confident wrong association is not neutralised by a single accurate mention; it has to be outweighed.

And the model has to be re-taught — slowly

The AI mechanics compound it. Large language models learn brand-to-category associations from repeated co-occurrence across their training sources: when your name appears consistently alongside a category, the model learns to file you there. Once that association is consistent it becomes a confident representation — and confident representations only fully shift when the model runs a retraining or fine-tuning cycle, which happens on the order of months, not days. Inconsistent signals in the meantime produce a confused entity rather than a corrected one. Building a new, accurate association is slow; overwriting an entrenched wrong one is slower.

You are already in the wrong conversation

There is a commercial dimension on top of the technical one. A miscategorised brand is not absent from the buyer’s consideration — it is present, confidently, as the wrong thing. It is compared against the wrong competitors, judged on the wrong criteria, and ruled in or out for reasons that do not apply. And every fresh mention that does not correct the frame quietly reinforces it. You are not starting the conversation; you are trying to change one already in progress, against a model that keeps restating its prior.

So what do you fix first?

This is where the judgement lives, and why we diagnose before we prescribe. If AI already understands you and the gap is discovery, the priority is presence, and unbranded, buyer-style questions are the sharpest way to measure it. If AI miscategorises you, the priority is accuracy — and the first move is branded questions that surface exactly what the model believes, so you can see the wrong story in full before you start overturning it. In both cases the rule is the same: correct before you amplify. Drive visibility while the story is still wrong and you do not fix the problem — you scale it.

Why an independent read catches it early

Because miscategorisation hardens with repetition, the cheapest time to fix it is before it sets — which means the value is in catching it early, not in a bigger project later. An independent diagnosis tells you which problem you actually have, how entrenched it is, and the order to address it in, with no fix to sell and no reason to inflate the work. The goal is not simply to be visible, and not merely to be accurate, but to be present and correctly understood — in that order.

Frequently asked questions

Is it worse to be invisible to AI or miscategorised by it?
Both are problems, but miscategorisation is usually the harder and more expensive one to fix. Being invisible is a single job: build presence. Being miscategorised is two jobs — un-teach the wrong story and teach the right one — and the wrong story is sticky. You are also already in the wrong conversation, which the model reinforces every time it repeats it.
Why is fixing miscategorisation two problems instead of one?
Because before you can establish the correct category, you have to overturn the incorrect one the model already holds with confidence. An invisible brand is a blank page; a miscategorised brand is a page with the wrong thing written on it in ink. Erasing and rewriting is more work than writing on a blank page.
Why do AI systems hold on to a wrong brand category?
LLMs learn brand-to-category associations from repeated co-occurrence across their training sources. Once an association is consistent it becomes a confident representation, and confident representations only fully change on slow retraining or fine-tuning cycles. This mirrors the continued influence effect in human cognition — corrected beliefs keep exerting influence long after the correction.
If I am miscategorised, what should I fix first?
Accuracy before amplification. Correct the sources and signals that define your category before you drive more visibility — otherwise you scale the wrong story. In measurement terms that means leaning on branded questions first to surface exactly what AI believes, then unbranded discovery questions once the story is right.
Is being invisible to AI harmless, then?
No. Absence keeps you off the shortlist entirely, which costs real deals. The point is not that invisibility is fine — it is that if you are choosing what to catch early, catch miscategorisation, because it hardens with every repetition and gets more expensive to unwind the longer it stands.

See what AI says about your business

A no-charge 30-minute AI Visibility Initial Review. Live on screen. Unedited. No preparation needed — just your company name.

Book an Initial Review →