Because an AI system is not comparing your stars with your neighbour’s. It is answering a different question: who does this work in this city. Your 300 reviews answer that question only if they are attached to the place you work from, and only if they say what you actually did. The neighbour with 4.3 is not winning on rating — he fits the question and you do not.

Reviews belong to the profile, not to the company

I will start with how this is built, because nothing below makes sense without it. Google reviews do not belong to a company. They belong to its Business Profile, and that profile is tied to the place the company works from. If you serve customers at your premises, that place is an address. If you travel to the customer, no address is needed and Google explicitly requires you to remove it: instead you set a service area of up to twenty cities or postal codes.

That area has a limit. Google caps it at roughly two hours of driving time from where the business is based. One profile therefore cannot cover a whole country: however many cities you list, the area will not stretch beyond that boundary.

Which leads to something owners rarely hold in mind. Every review from every customer accumulates on one profile, and that profile works inside its own area. A person never notices: they open your site, see the page for their city and the reviews underneath it. An AI system sees something else — it looks for companies that work in the city it was asked about.

What three measurements in one city showed

We asked four AI systems — ChatGPT, Gemini, Perplexity and Google AI — the same questions: who to choose for this service in one specific city. Three measurements between 10 July and 8 September 2026, 27 companies appearing in the answers, 200 answers in which the systems named a specific company.

The picture looked like this. Company A: 516 Google reviews, rating 5.0, profile in this city — named in 80 answers out of 200. Company B: 133 reviews, rating 5.0, profile here — 63 answers out of 200. Company C: 90 reviews, rating 5.0, profile here — 48 answers out of 200. And company D: 2,419 reviews, rating 4.9, but its profile in another city — 25 answers out of 200.

Two things follow. First, the rating explains nothing here. The three local companies all have the same rating, 5.0, yet they are named 48, 63 and 80 times out of 200 — almost a twofold spread on an identical rating. The rating cannot be the cause of the difference, because there is no difference in rating.

Second, reviews do work, but inside their own area. Among the three local companies the order holds: 90 reviews give 48 mentions, 133 reviews give 63, 516 reviews give 80. More reviews, more mentions.

Company D falls out of that line entirely. It has 2,419 reviews — more than the three local companies combined — and a rating of 4.9. It was named 25 times out of 200, and 20 of those 25 fell in the first measurement: 20 mentions in July, one in August, four in September. The company works in several cities, but it has a single Google profile, tied to a city several hours’ drive away. Under Google’s two-hour rule our city cannot be inside its service area, and it has no profile of its own here. Its website does have a page for our city, with reviews shown underneath — but those are the same reviews from that one distant profile.

Hence the answer to the title question: your reviews work where your profile works. Outside its area they do not exist for an AI system.

AI does not count stars, it reads text

The second thing we see in every measurement: AI systems quote the wording of reviews, not their arithmetic mean. A review saying “everything was great, highly recommend” tells a machine nothing about what you do. A review saying someone arrived the same day and fixed a specific thing in two hours is a ready-made sentence an AI can lift and repeat when asked about that service.

So 300 reviews with no substance and 30 reviews describing the work done are not the same thing to a machine, even though both numbers look equally good on the profile.

What the research says about reviews

The largest open study of AI answers about local businesses is BrightLocal, 16 September 2026: 200,085 localised searches, 1.9 million citations, 1,342 locations. Google Business Profile accounted for 28.63% of all citations, more than any other source, which matches what we see in our own measurements. Google Maps shows a tracked business in 66% of searches, while AI systems name it in only 32–38%.

And the part that matters most for our question: neither rating nor review count appears as a factor in that study at all. It found no link between them and being recommended by AI.

The second source is Whitespark’s annual survey of local ranking factors, 47 specialists, November 2025. Reviews are rated there, but as one factor among a hundred: high Google ratings rank sixth, quantity of reviews ninth, recency of reviews eleventh. These are expert opinions rather than a measurement, and none of the three makes the top five.

When this does not apply to you

If you work in one city and your profile is here, everything described above is already in order for you, and working on reviews will not add recommendations — your bottleneck is elsewhere. If you serve several cities but are not prepared to set up a separate profile in each, this will show you the cause but not the solution. And if you expect AI systems to start naming you within two weeks, they will not: weeks pass between changes to a profile and changes in AI answers.

How to check yourself in ten minutes

First look at which city your Google profile is tied to, and whether that is the city where you look for customers. If you work in several cities and have one profile, you already know where all your reviews are sitting. Google sets out the service-area rules in its own help pages.

Then open your last ten reviews and answer three questions: when was the most recent one left, how many of the ten say what you actually did and where, and do any reviews of you exist anywhere other than Google.

If none of the three answers satisfied you, the problem is not your 4.9.

So when did you get your last review, and what did it say?

Sources: BrightLocal, Local AI Visibility Study (200,085 searches, September 2026); Whitespark, Local Search Ranking Factors 2026 (47 specialists, November 2025); Google Business Profile Help — service areas.

To see whether your business already appears in AI answers, see the AI Visibility Audit

Want to check whether AI already recommends your business?

We start with a diagnosis of the company’s current visibility and real customer decision scenarios.

VIEW THE AUDIT