Google Business Profile for AI search

What it does and does not cover

Google Business Profile is the single most important local record you own — for two of the four engines.

Gemini and AI Overviews resolve local questions largely through Google’s own local data. If your profile is thin, wrong or unclaimed, you are difficult for either to name.

ChatGPT and Copilot resolve through Bing. A flawless Google profile does nothing for them, which is the most common reason a practice is strong on one engine and absent on two others. That asymmetry is covered in why Bing Places decides your ChatGPT visibility.

So: necessary, high-value, and roughly half the job.

The fields that carry weight

Not all of the profile matters equally. Ranked by what actually appears to influence selection:

Primary category. The highest-leverage field on the profile and the most commonly wrong. An engine matching “med spa in Scottsdale” is matching against a category value. If yours says “Skin Care Clinic” you are competing in a different set than you think. Pick the category that matches the words patients use, not the one that flatters the business.

Secondary categories. Add every category you genuinely operate in and none that you do not. This is where multi-service practices win qualified queries.

Services. Structured, itemised, individually named. If Botox is a meaningful share of revenue it should exist as a named service, not be implied by the category. This is the field most practices leave nearly empty.

Attributes. Machine-readable facts — accessibility, appointment requirements, who the business identifies as. Small individually, but they are exactly the qualifiers that decide “near me” variants.

Hours, including exceptions. Availability queries are a real category of local question. A profile that says you are open when you are closed produces a bad answer that gets attributed to you.

Q&A. Underused and unusually valuable, because the format is already question-and-answer. Seed the real questions your front desk answers daily and answer them properly from the business account.

Photos. They affect human conversion more than machine selection. Worth doing, not worth doing first.

The failure that costs most

Category mismatch, followed by an incomplete services list.

Both fail silently. Nothing is broken, no warning appears, and the profile looks finished. It simply never enters the candidate set for the queries you care about, and you have no way to notice from inside the dashboard.

The only way to detect it is to run the queries and see who comes back — which is what the prompt set builder generates the list for.

Consistency is the second job

Your name, address and phone must be character-identical across Google, Bing, Foursquare and Yelp.

“Suite 200” in one record and “Ste 200” in another is enough to stop an engine confidently resolving the two to a single business. When entity resolution fails, engines behave conservatively — and conservative means naming someone else.

This is unglamorous, takes an afternoon, and is the highest-certainty work available in the whole discipline. Everything else involves probability. This is arithmetic.

Posts, and honest expectations

Posts have modest direct effect on whether an engine names you. They signal an actively operating business, which matters at the margin, and they occasionally get surfaced.

They are not the lever. A practice posting weekly with eleven reviews and a wrong primary category is optimising the wrong thing entirely. Fix the structured fields first; post afterwards if there is time.

Where this sits in the order of work

1. Claim and verify every location.
2. Fix the primary category.
3. Fill services out properly.
4. Reconcile name, address and phone across every directory.
5. Complete attributes and hours.
6. Then Bing, or you have optimised for half the market — the reasoning is in AEO, GEO and SEO: what the terms actually mean.
7. Then reviews, against the local threshold rather than an arbitrary target — see how many reviews before AI recommends you.

For how these signals combine into a selection, how AI engines choose which business to name covers the mechanism. If you want the Google-specific surface in isolation, how to appear in AI Overviews goes deeper on that engine.

How to tell whether it worked

Directory changes propagate faster than review or editorial work, so this is one of the few interventions where movement inside three to four weeks is reasonable to expect.

Re-run the same prompts in a fresh, signed-out session about a month after making changes. Measure before you change anything, or you will have no way to attribute what happened.

The usual caveat

Google does not publish how its local data feeds AI Overviews or Gemini, or how the fields are weighted. The ranking above comes from measured output across repeated runs and from correlations documented across the industry — not from confirmed factors.

The safe part of the claim is much narrower and still worth acting on: engines cannot select a business from a record that is incomplete, miscategorised, or absent.