AI puts my company in the wrong category. How do I fix it?

99Visibility Research · Updated October 21, 2026 · 6 min read

Short answer

Put your category in one plain sentence on a page AI crawlers can fetch, in the HTML before any JavaScript runs. Use that same sentence on your homepage, about page and third-party listings. Engines retrieve pages by topical fit, not keyword lists, so a clear and consistent definition gives them something accurate to quote. Nobody controls how fast a model updates, so expect no fixed timeline.

Key takeaways

A one-line category definition on a fetchable page gives AI engines a sentence they can quote instead of a guess.
Inconsistent descriptions across your own pages and third-party listings give engines conflicting evidence to choose from.
Research points to topical relevance, not keyword coverage, as one of the most reproducible drivers of what gets cited.
A wrong category can come from frozen training data or from live retrieval, and only the retrieval side responds to your edits.
Check that AI crawlers can reach your pages before you rewrite anything.

The short fix and why it works

AI engines describe your company using whatever text they can find and trust. If that text is vague, contradictory or missing, the model fills the gap with the closest category it knows. A project management tool gets called a CRM. A clinic gets called a spa.

The fix has three parts:

Write one plain sentence that states your category. For example: "Acme is invoicing software for freelance designers." Put it near the top of your homepage and your about page.
Make it fetchable. The sentence has to exist in the raw HTML, not only after JavaScript runs.
Repeat it everywhere you control. Use the same wording on directory profiles, marketplace listings, review sites and social bios.

This works because a specific, stated fact on a page an engine can fetch gives it a sentence it can quote. That is the only on-page change with a positive experimental signal in our evidence base. It is not a guarantee. It removes the reason for the guess.

For a broader plan covering every kind of brand error, see our fix plan for AI getting your brand wrong.

Where the wrong category comes from

A wrong answer can come from two very different places, and the fix depends on which one.

Trained knowledge (no web search)

When an engine answers without searching, it relies on what it learned before its knowledge cutoff. OpenAI's model docs list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path used in ChatGPT. If you repositioned after those dates, the model's frozen view can be out of date. Nothing you publish this quarter changes a retrieval-off answer.

Live retrieval (web search on)

When an engine searches, it pulls pages into its context and writes from them. This is where your one-line definition matters, because it can be fetched and quoted right now.

Retrieval is per query and at the model's discretion on most platforms, as of October 2026. The same question can get a searched answer one time and a memory-only answer the next. Our page on trained knowledge vs search errors in ChatGPT walks through how to tell them apart. If the category is simply old, not invented, the hallucination vs outdated answer comparison helps you diagnose it.

What the evidence says: relevance, not keywords

It's tempting to stuff every category term into your pages. The evidence points the other way.

A July 2026 critical survey of 45 GEO studies (arXiv 2607.14035) found that "topical relevance and context position are the most reproducible levers" and that generic heuristics transfer poorly. A 252,000-trial factorial study (arXiv 2605.25517) also found topical relevance and list position were the biggest drivers of being cited first. Both studies tested pages already in the model's context, so they say nothing about getting retrieved in the first place.

Google's AI optimization guide says you don't need to worry about having enough "long-tail" keywords or capturing every variation of how someone searches. The original GEO paper found keyword stuffing lowered its own visibility metric.

What this means in practice: describe what you actually are, clearly and consistently. Don't try to rank for every adjacent category. A page that plainly matches your real category is more relevant to the right queries than one that chases ten.

Two things don't help, despite common advice: adding schema markup and writing in a special "AI-friendly" style. Google says neither is required for its generative AI features.

Check access before you rewrite

A perfect definition does nothing if crawlers can't read it. Check these first:

JavaScript rendering. Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. If your category line only appears after the page hydrates, those crawlers may never see it. Googlebot renders JavaScript, so this mainly affects non-Google engines.
robots.txt per bot. OpenAI states each crawler setting is independent. OAI-SearchBot controls ChatGPT search. GPTBot controls training only. A blanket AI block can quietly remove you from ChatGPT search.
Firewall and CDN. A firewall can block a crawler that robots.txt allows, often with a 403 and no warning to you.

The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.

Edge cases

You genuinely span categories

Pick the primary one and lead with it. Mention secondary uses after. An engine given one clear primary category has less to reconcile than one given four equal claims.

You pivoted or rebranded

Old press coverage, directory listings and archived pages still describe your old category. Update the listings you control. For coverage you don't control, being named correctly on sites engines already cite is associated with appearing in answers, though no study promises an effect.

Third parties describe you wrong

When engines cite other sites instead of yours, those sites shape the answer. Ask listing owners to correct your category. Correcting the model inside a chat won't fix it for other users, as our page on correcting ChatGPT in a chat explains.

Someone is manipulating the answer

Hidden text or prompt injection on third-party pages can mislead AI search agents. Don't use these tactics yourself. They are illegitimate and risky. Watch for them as an inbound threat.

What to do next

Ask each major engine what category you're in, using the questions buyers actually ask. Note which engines get it wrong. The free checker on this site runs up to three buyer questions and shows how you're described.
Write your one-line category definition and place it in server-rendered HTML on your homepage and about page.
Audit every third-party listing and make the wording match.
Confirm crawler access per bot and at the firewall.
Re-check periodically. Retrieval-on answers can change as pages are recrawled. Retrieval-off answers change only when a vendor ships a new model.

If you'd rather escalate, see whether you can ask OpenAI or Google to correct wrong information. For ongoing errors, the hallucination triage playbook helps you rank which ones to fix first.

Frequently asked questions

How long until AI shows the right category?

Nobody can give you a reliable timeline. Searched answers can change after crawlers refetch your updated pages, but when that happens depends on each engine. Answers from trained knowledge only change when a vendor releases a model with a newer knowledge cutoff. Fix what is in your control, then re-check on a regular schedule rather than waiting for a date.

Will adding schema markup fix my category?

The evidence doesn't support that. Google says structured data isn't required for its generative AI search features. Ahrefs' matched-control study of 1,885 pages found adding schema produced no major uplift in AI citations on any platform. Existing markup does no harm, so there's no need to remove it, but it isn't a lever for fixing how AI categorizes you.

Why does ChatGPT say one category and Perplexity another?

Each engine uses its own index, its own models and its own retrieval decisions. One may have searched and found your current page while another answered from older trained knowledge. That's why you should check each engine separately instead of relying on a single blended score. A difference between engines is a clue about which stage the error comes from.

Should I create a separate page for each category I want to appear in?

Not if you aren't really in those categories. Research on AI citation points to topical relevance as a reproducible driver, and Google says you don't need to cover every keyword variation. Thin pages chasing adjacent categories add conflicting signals. Build pages for what you actually offer and describe each one plainly.

Does an llms.txt file help AI understand my category?

There's no evidence it does. Google says Google Search ignores llms.txt, and no other major engine documents reading it. It won't harm you, but it isn't a fix. A plain category sentence in your fetchable HTML is the more direct route, because that's the text engines actually retrieve and quote.

Sources

This page is AI-assisted. It was written from a verified evidence base and checked against our editorial policy. Engine behaviour changes often; facts are as of October 21, 2026.