Branded queries are not the same as visibility

If someone tells you they “show up in ChatGPT,” ask what they typed.

Most of the time it is a branded prompt: the business name is already in the question. The model has an easy job. Naming the firm back is not the same as being the answer when a shopper does not already have that name.

What a branded prompt actually tests

“What is Harbor Street Dental?” and “Tell me about Harbor Street Dental in Riverton” test recognition. A clear website, a complete Google Business Profile, and consistent third-party listings can make the business easier for search-enabled systems to identify. That is useful. It is not the same test as discovery by category.

A branded miss is still a finding — it often means the entity is thin, confused with a similarly named shop, or missing from the sources the model is willing to cite. Treat it as an identity problem, not as proof you “rank in AI.”

The questions that decide the next job

Shoppers who do not know you yet ask different things:

• Category — “estate planning attorney in Riverton,” “emergency plumber near downtown”
• Local — “dentist open Saturday,” “best coffee near the courthouse”
• Comparison — “Harbor Street Dental vs Maple & Pine,” “which HVAC company is better in [city]”
• Hire / buy / visit intent — “who should I hire for a kitchen remodel,” “where to buy hearing aids in [city]”

Those prompts do not hand the system the answer. Depending on the product, it may draw from a search index, web results, maps, reviews, third-party articles, or model knowledge. The exact source mix is not public or consistent across products.

A business that is featured on its own name and absent on category questions is not “good at AI search.” It is findable to people who already know it. That is the same split local SEO people have lived with for years — branded vs non-branded — now showing up in generated answers.

Do not average the story

One “AI visibility score” that blends branded and category prompts will lie to a client. Branded hits inflate the average. Category misses are where the work is.

When you brief this, keep the clusters separate. Report branded recognition on its own line. Then show category, local, comparison, and hire-intent separately, and by engine if you have more than one. The interesting sentence is usually: “They are named when you ask for them. They are missing when you ask for the service.”

What to do with a branded-only result

You do not need a new channel strategy on day one. Check the boring inventory:

1. Is the name, address, and phone consistent across the site and the listings the models actually cite?
2. Do service pages say the category in plain language, or only in a slogan?
3. Is there a Google Business Profile with categories that match how people ask?
4. Does robots.txt block the search crawler for the product you are testing? Training crawlers and search crawlers are not interchangeable—for example, OpenAI documents GPTBot and OAI-SearchBot as separate controls. (OpenAI crawler documentation)

If those are messy, generated answers will stay messy. If they are clean and you still only win branded prompts, the next sprint is category proof: pages, listings, and third-party mentions that would justify naming you when the prompt does not.

Visibility is what happens when the shopper did not start with your name.