Where AI Assistants Actually Get Their Local Business Recommendations

Ask ChatGPT for a good dentist near you, then ask Google's AI Mode the same thing. You will often get different names.

That is not randomness. The two systems are drawing on materially different sources, and knowing which sources feed which answer is the difference between working on your visibility and guessing at it.

A striking illustration surfaced this summer. PPC Land reported that Yelp — a site whose entire business is local recommendations — sees only around 5% of its AI citations coming from ChatGPT, against roughly 66% from Google AI Mode. Same directory, same data, wildly different weight depending on which assistant is asking.

Why the same business gets different answers

Google AI Mode sits on top of Google's local index. It has direct access to Google Business Profile data, the map pack, Google reviews, and everything Google already knows about proximity and category. When it recommends a local business, it is largely reformatting an answer Google could already produce.

ChatGPT is assembling from a different pile. Its live web access historically leans on Bing, and its picture of your business comes from whatever it can retrieve and whatever was in its training data. Bing Places matters here in a way it does not for Google. So do the large directories, your own website, and any editorial coverage naming you.

The practical consequence: optimizing only for Google gets you into one of these answers, not both. A business with an immaculate Google Business Profile and a neglected Bing Places listing is well positioned in AI Mode and close to invisible in ChatGPT.

The trust gap that undercuts everyone's assumptions

The same PPC Land reporting noted two figures worth holding together.

Around 65% of consumers want to take action on a platform they already trust when using AI for local discovery. They will accept the AI's suggestion, then go and book through a name they recognize.

But only about 15% of Americans say they highly trust AI search results.

Those two facts explain a pattern that confuses a lot of business owners. AI recommendations are influential without being decisive. People use them to build the shortlist, then verify through a channel they already believe: your Google profile, your reviews, your actual website.

So the AI answer is rarely the final step. It is the step that decides whether you are in the running at all. Which means being named matters enormously, and being named is not sufficient — whatever the person finds when they check you afterwards still has to hold up.

What this means practically

Claim Bing Places. Today, if you have not. It is the single highest-leverage listing most local businesses have never touched, because its value is now wildly out of proportion to Bing's actual search share. It feeds assistants that Google's ecosystem does not.

Do not treat directory listings as a box-ticking exercise. Yelp, Healthgrades, Avvo, Houzz and their equivalents are not merely citation sources. They are heavily weighted inputs to the assistants that do not have Google's index, and they rank in their own right.

Get your data identical everywhere. This matters more for the non-Google assistants, because they have less context to resolve a contradiction. Google can tell that "Ste 4" and "Suite 4" are the same place. A system reconciling four directories with three different formats is less certain, and uncertainty costs you the mention.

Then make the verification step land. Because most people will check you after the recommendation, the assistant's answer only earns you the click. Your review count, your rating, your recency and your actual site have to do the rest. A recommendation that leads to a thin, stale profile converts worse than no recommendation.

Test it rather than assume it

This is measurable in about ten minutes and almost nobody does it.

1. Write down the five questions a good customer would actually ask. Not keywords — questions. "Who should I call for an emergency boiler repair in [town]?"
2. Ask each one in ChatGPT, in Google AI Mode, and in Perplexity.
3. Record who gets named, in what order, and what each assistant says about you if it names you at all.
4. Where a competitor appears and you do not, go and look at what they have that you do not. Usually it is review volume, a directory listing you skipped, or a page that answers the question directly while yours circles it.

Repeat quarterly. The answers move.

The mistake to avoid

Do not build separate strategies per assistant. There is a small industry forming around "ChatGPT optimization" as a distinct discipline, and it is mostly repackaging.

The inputs overlap heavily. Accurate structured business data, consistent details, strong recent reviews and specific quotable content feed all of them. What differs is the weighting, not the fundamentals. Fix the fundamentals once, make sure you are present on the non-Google surfaces too, and you have covered most of the ground that matters.

The one genuinely assistant-specific action on this list is claiming Bing Places, and that takes ten minutes.

What to do this week

1. Claim and complete Bing Places if you have not.
2. Run the five-question test across three assistants and write down the results.
3. Audit your top five directory listings for identical name, address and phone formatting.
4. Check your review recency. If your most recent review is more than a month old, the habit has lapsed.

The bottom line

Different assistants read different sources, and a business optimized only for Google is competing in one of them. The Yelp figures make the gap concrete: 5% versus 66%, for the same underlying data.

Most people will still verify you after the recommendation, through a channel they already trust. So the goal is not simply to be named. It is to be named, and then to survive the check.

Want to know which assistants name you and which name a competitor? Request a visibility check.