Why ChatGPT Called Our AI Company a Marketing Agency, and the Google Setting That Fixed It

Here is the short version. EVOIX is an AI company, but ChatGPT, Google, and our own local search results kept calling us a marketing or advertising agency. The cause was not our website copy or our structured data, both were already clean. It was the category stack inside our Google Business Profile. We changed one primary category and pruned three secondaries. Below is exactly how we found it.

There is a particular flavor of embarrassing that comes from running an Answer Engine Optimization practice and discovering the answer engines cannot tell what your own business does. That was us, on June 27, 2026. We sell AEO as our flagship service. We had ourselves a positioning leak. So we treated EVOIX like a client, found the root cause, fixed it, and wrote down the method while the screenshots were still open.

What was actually broken?

EVOIX, an AI consulting company in DeLand, Florida, was being classified by AI engines and by Google as a marketing or advertising agency, and our flagship service, Answer Engine Optimization, was effectively invisible. The misclassification did not live on our website. It lived in an off-site signal we had set months earlier and forgotten.

The symptoms stacked up fast. When we asked ChatGPT for AEO agencies, it named competitors and skipped us. Google publicly displayed "Marketing agency" as our business category on Maps. Our canonical Answer Engine Optimization service page was sitting at position 60.7 for its own core term, "answer engine optimization." For a company whose entire pitch is "we get you recommended by AI," being filed under the wrong category is not a cosmetic problem. It is the whole problem. Google's own guidance on choosing a business category is blunt that the category shapes how and where you surface, and we had handed it the wrong one.

How did we find it? The Category Mirror.

The Category Mirror is a three-source check we run to find an entity leak. Ask the AI what you are, read the category Google publicly shows for you, and read the category your own schema declares. Where the three disagree is where AI search misreads your business.

We ran all three mirrors on EVOIX:

• The AI mirror. Ask ChatGPT, Claude, and Perplexity "what is [your business] and what do they do." We did this with live web search on. The engines leaned on third-party signals and described us in marketing-agency language, not as an AI company.
• The Maps mirror. Read the category label Google shows on your Business Profile, and check whether you appear for a wrong-category query. Google was displaying "Marketing agency." Worth noting, we did not actually rank organically for "advertising agency volusia county," the wrong-category footprint came through the local profile, not our pages.
• The schema mirror. Read the category your own JSON-LD declares. Ours was already correct, it described a technology company. That single disagreement, clean schema against a dirty Google profile, told us exactly where the leak was.

When two mirrors say "AI company" and one says "marketing agency," you do not have a content problem. You have one bad witness.

Why was a Google dropdown lying about us?

To an AI engine, your Google Business Profile is a witness, not a footnote. Large language models treat third-party profiles as testimony about what your business is, and they weight that testimony heavily because it is structured and trusted. Our profile testified "marketing agency," so the models repeated it, no matter what our own website said.

This is the part most businesses get backwards, so it is worth a name. Think of the AI as a judge and every third-party profile, Google, Yelp, LinkedIn, an old directory listing, as a witness. If the judge keeps calling you the wrong thing, you cannot win by shouting louder on your own website, that is just the defendant talking. You have to fix the witnesses the judge actually trusts. Google Business Profile is the loudest witness in local search, because it feeds Maps, the local pack, and the AI Overviews that read from them. We had let that witness say the wrong thing for months. The fix was not to argue with the judge. It was to correct the testimony. The same logic powers our local AEO work in Volusia County, where the local profile is often the single highest-leverage signal.

What did we actually change?

We made two edits inside the Google Business Profile category settings, nothing else, no code, no website change.

Category slot	Before	After
Primary	Marketing agency	Business management consultant
Secondary	Advertising agency	Removed
Secondary	Marketing consultant	Removed
Secondary	Software company	Kept
Secondary	Computer consultant	Kept
Secondary	Internet marketing service	Kept

Two honesty notes, because this is the part that bites people. Changing your primary category can trigger a Google re-verification of the listing, removing secondaries does not. We judged the re-verification risk worth it, because an accurate primary category is worth more than the inconvenience, and re-verification with independent corroborating sources rarely ends in a suspension. We submitted the change and it went into Google's review queue the same day.

What we did not touch, and why

We changed nothing on the website, because the website was not the problem. Our homepage already said EVOIX is a technology company, not an advertising agency, and our Organization structured data was already correct. Rewriting on-site copy would have cost a week and fixed nothing.

That is the lesson hiding in this whole exercise. When AI misreads your business, the instinct is to rewrite your homepage for the tenth time. Usually the leak is off your site, in a witness you forgot you set. We have watched this work in the other direction too: when the entity signals line up, AI engines respond fast. A local client of ours went from a documented baseline of zero AI citations to the number-one ChatGPT recommendation for a local buyer query eight days after we shipped the work. Entity alignment is not slow magic, it is plumbing, and plumbing responds when you fix the right pipe.

How do you run the Category Mirror on your own business?

You can run all three mirrors yourself in about fifteen minutes, no tools required beyond a browser.

1. Ask the AI. In ChatGPT, Claude, and Perplexity with web search on, ask "what is [your business name], and what do they do." Note the category language it uses.
2. Read the Maps mirror. Search your business on Google, open the Business Profile, and read the category shown under your name. That is the label Google broadcasts.
3. Read the schema mirror. View your homepage source and find your JSON-LD, or paste your URL into Google's Rich Results Test, and check the declared business type and category.
4. Find the disagreement. Line up the three answers. If they do not match, the one feeding the AI a wrong answer is your leak, and in local search that is almost always the Google profile.
5. Fix the loudest witness first. Correct the category that is wrong, starting with Google Business Profile, then move outward to other profiles that echo it.

If all three agree and AI still gets you wrong, the problem is upstream in your off-site mentions, and that is a different, longer job.

Does this matter for a local business in Volusia County?

For a local service business in DeLand or anywhere in Volusia County, the wrong category does more than confuse an AI, it pulls you out of the local pack and out of the "near me" AI answer entirely. A roofer filed as a "general contractor," or an HVAC company filed as an "appliance store," loses the exact query that was about to become a customer.

Local is where category precision pays the most, because local AI answers return one or two named businesses, not ten blue links. If Google has you in the wrong bucket for "best [service] in DeLand," you are not competing for that answer, you are not even in the room. We would rather a Volusia County business spend an afternoon on its category stack than a month on a blog nobody asked for. Fix the witness, then earn the citations.

What we are watching next

We submitted the category change on June 27, 2026, so two things are now in flight: whether Google flips the public label from "Marketing agency" to "Business management consultant," and whether our canonical AEO page climbs off position 60.7 once it is crawled. We will update this post with the after numbers rather than declare victory early, because a case study that skips the part where you wait is not a case study, it is an ad.

If you want EVOIX to run the Category Mirror on your business and fix the witnesses AI is reading, you can see exactly what that costs with no quote form, or start with what Answer Engine Optimization actually is.

---

Written by Stephane Morera, founder of EVOIX, an AI company in DeLand, Florida. Stephane holds an Elite-level AI and Machine Learning certification and a BrainStation diploma in full-stack software engineering. More on the team and method, or connect on LinkedIn.