What Your Reviews Tell AI Assistants: How Review Content Shapes Local Recommendations

For years, the questions about reviews were mostly about numbers: how many you have, what your average is, and how you compare with the business next door. Those numbers still matter. But when a customer asks ChatGPT, Gemini, Copilot, Perplexity or Google’s AI Mode to recommend a local physio, plumber, salon or solicitor, the answer they get is not a number. It is a short written description, and much of that description tends to be built from what your customers have written about you.

That changes what a “good” review profile looks like. Two businesses with the same 4.7 rating can come out of an AI answer sounding very different, depending on what their reviews actually say, how recent they are, how the owner has replied and whether the picture is the same on every platform the assistant reads.

This guide looks specifically at review content: what the AI companies have publicly said about how their tools use reviews, what local search practitioners commonly observe beyond that, and what you can realistically influence. If you want to know how to collect more reviews or how to reply to a difficult one, we cover those in separate guides. Here the focus is on what your reviews say, and how that becomes what an AI assistant says about you.

What is documented, and what is only observed

It is worth separating two kinds of information before going further, because a lot of advice on this subject blurs them.

Documented means the company behind the tool has said it publicly, in a help page, an announcement or developer documentation. Observed means people who test these tools regularly, including local search agencies and researchers, have noticed a pattern, but the company has not confirmed it. Observed patterns can be useful, but they can also be wrong, out of date or specific to one country.

Here is roughly where things stand as at October 2026.

Tool	What the company has said publicly	What is commonly observed
Google Maps and Ask Maps	Google says its Gemini-powered Ask Maps feature draws on Maps information about places, including reviews from its community of contributors, to answer questions	Answers often paraphrase recurring review phrases, such as “friendly staff” or “short waits”
Google AI Overviews and AI Mode	Google says normal SEO best practices apply to these features, including keeping Business Profile information up to date	Local answers frequently cite or summarise Business Profile reviews
Gemini	Gemini can draw on Google Maps information for place questions	Descriptions often echo the same themes shown in Maps
ChatGPT	OpenAI says ChatGPT search may share queries with third-party search providers such as Bing	Descriptions often reflect reviews on Google, directories and industry sites
Microsoft Copilot	Microsoft says Copilot uses Bing to search the web	Bing’s local results and the review sites it surfaces appear to feed descriptions
Perplexity	Shows its sources alongside each answer	Cited sources often include review platforms and “best of” lists

None of these companies publishes a formula for how reviews are weighted, and features change frequently. The practical conclusion is modest but reliable: reviews are part of the material these tools read, and the content of those reviews, not just the score, ends up in the answer.

AI assistant

Which physio near me is good for running injuries and gets you in quickly?

Your Business is a local clinic whose recent reviews often mention running injuries, practical home exercise plans and short waits for appointments. One or two other clinics nearby are also well regarded for sports rehab.Google reviewsBusiness websiteLocal directories

Review summaries in Google Maps and local search

Google’s own products are the clearest place to see review content being turned into AI-written text, and they matter most for local businesses because so many customers start in Google Search or Maps.

Google has been adding Gemini to Maps in stages. Its Ask Maps feature, which lets people ask conversational questions about places, launched in the US and India in March 2026 and began rolling out in English to many more countries in August 2026. Google described that wider rollout as covering over 150 countries and territories, but it did not list every one, so the simplest way to check whether a customer in your area sees it is to open Google Maps on an up-to-date phone and look. Availability of individual features also varies by country, language and device.

Separately, Google made AI-powered review summaries available through its Maps Platform in September 2025, with UK English among the supported regions. That means apps and websites built on Google’s mapping data can show a short, Gemini-generated summary of what reviewers say about a UK place, even if the person never opens Google Maps itself.

physio near me

Your Business#1

★★★★★

Physiotherapist · Open until 7pm

CallDirectionsWebsite

Competitor#2

Physiotherapist

Competitor#3

Sports massage therapist

What does not change is the foundation underneath. Google’s guidance on local ranking says that review count and review score are part of how it judges a business’s prominence, alongside relevance and distance. So the reviews that help you appear in the map results are the same ones that AI features read when describing you. There is no separate set of reviews “for AI”.

A useful exercise is to look at your own Business Profile as a stranger would. Search your business name on your phone, open the reviews section and look at the topic chips Google sometimes shows above the reviews, such as “parking”, “waiting time” or “price”. These are a rough indication of the themes Google has picked out from your reviews, and they are a sensible proxy for what an AI summary might mention.

If you would rather have someone handle the profile, review and listings work alongside this kind of monitoring, local SEO and AI search services for small businesses are designed to cover both together.

Themes: how repeated phrases become a description

When an AI assistant summarises a business, it is effectively compressing dozens or hundreds of reviews into a sentence or two. The things that survive that compression are usually the things that come up again and again.

That is why one glowing review mentioning “the best sourdough in town” rarely shapes an answer on its own, while forty reviews that each mention “friendly”, “quick” and “parking” usually do. The pattern commonly observed is that assistants reach for the most frequent, most consistent themes, and often frame them cautiously: “reviewers often mention”, “many customers say”, “some reviews note”.

Think about what the recurring themes in your reviews are likely to be. For example:

• A plumbing firm might find its reviews keep returning to punctuality, tidiness and clear pricing.
• A dental practice might see “gentle”, “nervous patients” and “explained everything” again and again.
• A hair salon might see colour work, a particular stylist’s skill and being able to get a Saturday appointment.
• An accountancy practice might see responsiveness, plain-English explanations and fixed fees.
• A pub might see Sunday roasts, dog-friendliness and the beer garden.

Those themes are what an assistant is likely to repeat. If they match what you want to be known for, that is good news. If they do not, the most effective change is to the experience itself, not to the reviews.

Specific service mentions

Customers increasingly ask AI assistants detailed questions, not just “physio near me” but “a physio in Harrogate who is good with running injuries and can see me this week”. To answer that, an assistant needs evidence that a particular business does that particular thing well. Reviews that mention specific services are one of the richest sources of that evidence.

Compare two reviews of the same clinic:

1. “Great service, would recommend.”
2. “Saw Tom for a hamstring problem after a half marathon. Got an appointment within two days and the rehab plan had me running again in a month.”

Both are positive. Only the second gives an assistant anything to match against a question about running injuries or quick appointments.

You cannot write reviews for customers, and you should not tell them what to say. Google’s review policies say businesses should not request that specific content be included in a review, so the wording has to be left entirely to the customer. What you can do is make your own message specific: “Thanks for coming in about your hamstring this week. If you have a moment, we would welcome an honest review.” That names the job in your words and leaves the review itself to them. Customers who remember exactly what you did for them often mention it without being asked. It is also worth making sure your website and Business Profile list the specific services you offer, so that when reviews mention them, the assistant can connect the two.

Recency: describing the business as it is now

AI assistants try to answer as if describing the present. Several of them can see review dates, and Google’s own systems have long shown recent reviews prominently. It is commonly observed that descriptions lean on recent reviews where there are enough of them, although no company has published how much weight recency carries.

This matters in both directions. A business that has improved, perhaps under a new owner or with a new team, will want the recent reviews to tell that story, and older complaints will gradually matter less as newer reviews accumulate. A business that has gone quiet, with most of its reviews from three or four years ago, risks being described from an out-of-date picture, or being passed over in favour of competitors with a steady stream of fresh feedback.

The CMA’s guidance on reviews makes a related point from the publisher’s side: where a product or service has changed materially, older reviews that no longer reflect the experience may need to be handled with care. For a local business, the practical lesson is simply to keep the flow of reviews steady, so the most recent ones are always a fair reflection of how you work today.

Two businesses, similar ratings, different descriptions

The clearest way to see why review content matters is a side-by-side example. The two businesses below are invented for illustration, and the AI descriptions are the kind of wording assistants commonly produce, not quotes from any real tool.

Imagine two independent garages in the same town, both offering MOTs, servicing and repairs.

Garage A	Garage B
Google rating	4.7 from 210 reviews	4.7 from 190 reviews
Reviews in the last six months	About 40	About 8
Common themes	Honest about what does and does not need doing, MOT on the day, clear call before extra work	“Great service”, “friendly”, a few recent mentions of cars not being ready on time
Specific services mentioned	MOTs, clutch replacement, hybrid servicing, air conditioning re-gas	Rarely mentioned
Owner replies	Short, polite replies to most reviews, including the two negative ones	Occasional replies; one long argument with an unhappy customer
Presence elsewhere	Similar themes on a trade directory and Facebook	Few reviews outside Google

Asked “Which garage in town is honest about MOT repairs?”, an assistant might plausibly say something like: “Garage A is well reviewed, with customers often mentioning that it explains which repairs are necessary and calls before doing extra work. It also has positive feedback for hybrid servicing.” For Garage B, the answer might be: “Garage B has a good overall rating, although a few recent reviews mention delays in cars being ready.”

The ratings are identical. The difference comes from recency, specificity, consistency across sites and one visible pattern of complaints. None of it is a trick. Garage A simply has more evidence, in its customers’ own words, of the thing the question asks about.

Owner responses: what they add, and what they do not

Replies to reviews are part of the public record on your profile, and anything public can be read. Whether AI assistants give replies any particular weight has not been documented by the companies, so treat this as an observed pattern rather than a rule.

What is reasonable to say is that replies add context. A calm reply explaining that a delay was caused by a parts shortage, and that you have since changed supplier, gives a reader, human or machine, a more balanced picture. A pattern of replies showing that problems are acknowledged and fixed is itself a kind of evidence about how the business behaves.

Replies can also work against you. A defensive or sarcastic reply may become part of how the business is perceived, and some summaries mention how an owner engages with criticism. Copy-and-paste replies add nothing, and replies stuffed with keywords and town names read badly to customers and are unlikely to help with anything.

The most useful replies are short, specific and honest: thank the customer, refer to what they mentioned in natural words, and, if something went wrong, say what has changed.

Consistency across platforms

Different assistants read different sources. Google’s tools lean on Google’s own data, Copilot draws on Bing, and ChatGPT and Perplexity often cite a mix of directories, review platforms, local news and “best of” lists. That means the picture an assistant forms depends partly on which platforms it happens to read.

When the same themes appear everywhere, the description tends to be confident. When your Google reviews say one thing and an industry platform says another, or one platform has hundreds of reviews and the rest have none, the result can be patchy or cautious. It is commonly observed that businesses with broadly consistent feedback across two or three relevant platforms are described more clearly than those whose reviews sit almost entirely on one site.

Practical steps:

1. Choose your platforms deliberately. Google first for almost everyone, then one or two that matter in your sector, such as a trade directory for builders, Tripadvisor for hospitality or a specialist platform for legal or health services.
2. Keep your business details identical on each one: name, address, phone, services and areas.
3. Spread requests sensibly. You do not need to ask every customer to review you everywhere. Rotating which platform you suggest over time is enough to keep each one current.
4. Check older platforms. A forgotten profile with a handful of old, poor reviews can still be read and cited. Claim it, update it, and reply where appropriate.

Negative patterns and how they surface

A single bad review among hundreds of good ones is unlikely to appear in an AI description. A recurring complaint is different. If ten reviews over six months mention that the phone goes unanswered, it would not be surprising to see “some customers mention difficulty getting through” in an answer, however strong the rating.

Assistants commonly hedge these points rather than leading with them, but for a customer comparing three options, a single cautionary clause can be enough to tip the decision.

Typical negative patterns that tend to surface include:

• Communication: slow callbacks, unanswered emails, not being told about delays.
• Price surprises: final bills higher than quotes, unclear pricing.
• Timekeeping: late arrivals, appointments running behind, missed slots.
• Staff attitude: one recurring comment about rudeness can carry a lot of weight.
• Cleanliness or tidiness, especially for clinics, food businesses and trades working in homes.

The honest answer to a negative pattern is to fix the cause. As newer reviews stop mentioning the problem and start mentioning the improvement, the description has new material to draw on. There is no way to remove genuine criticism from what an assistant reads, and attempts to bury it, such as asking only happy customers for reviews or adding fake positive ones, breach Google’s policies and UK consumer law.

What you can influence, and what you cannot

It helps to be clear about where your effort is worth spending.

You can influence	You cannot control
The quality of the experience customers write about	What any individual customer chooses to write
Asking every customer, consistently, with an easy link	Which reviews an assistant reads in a given answer
Asking soon after the job, while it is fresh in the customer’s mind	How a company weights reviews against other information
Replying thoughtfully and fixing recurring problems	Whether a feature such as Ask Maps is available in your area
Keeping details consistent across your chosen platforms	Small day-to-day variations in what assistants say
Making your services clear on your website and profile	Reviews on platforms you have never heard of

Be wary of anyone who promises to change what AI assistants say about you through reviews. The legitimate route is slow and practical. The shortcuts, such as buying reviews, scripting them or swapping them with other businesses, are prohibited, and under UK law since April 2025 they can carry serious penalties.

A quarterly review-content check

A short, regular check of what your reviews actually say gives you a far better sense of your AI description than counting stars. It takes an hour or so.

1. Export or scroll through your last three months of reviews on Google and your main secondary platform.
2. List the themes, positive and negative, and count how often each appears.
3. List the specific services and situations customers mention, and note which important ones are missing.
4. Compare with your website and Business Profile. Are the services customers praise clearly described there?
5. Ask two or three AI assistants a realistic question about your business and your type of service, and note whether the description matches the themes you found.
6. Pick one operational fix and one review habit to work on before the next check, such as answering the phone more reliably, or asking for reviews after a particular kind of job.

Over time, you should see the themes in your reviews and the descriptions in AI answers move closer to the business you are trying to run.

Frequently asked questions

Final thoughts

AI assistants describe local businesses largely in their customers’ words. That makes the content of your reviews, what they mention, how recent they are and whether they agree across platforms, at least as important as your average rating. It also means there is no clever workaround: the descriptions improve when the service improves and when happy customers are given an easy, honest way to say what they valued.

Keep asking every customer, leave the wording to them, reply with care, fix the problems that keep coming up, and check what your reviews actually say every few months. Results build gradually, usually over several months rather than weeks, but the same work strengthens your Google Maps visibility and your conversion rate at the same time.