Guide · 9 min read

Reviews and AI recommendations: how ratings shape assistant answers

Ask an assistant for a good dentist, hotel or accounting tool and the answer often repeats what customers wrote somewhere else. This guide explains the link between reviews and AI recommendations: what the vendors themselves say, which platforms matter for each kind of business, how to earn more genuine reviews without breaking platform rules or the law, and what you can and cannot put in your structured data.

Review platforms and outside presence are part of our monthly work. Plans start at $449 a month with a three-month minimum.

Reviews and AI recommendations: how ratings shape assistant answers

Why assistants lean on reviews

When someone asks for a recommendation, an assistant has to choose between businesses it cannot visit or test. It falls back on what it can read: your own site, articles and lists that mention you, and what customers say on review platforms.

Reviews offer things your own pages cannot. They come from third parties, they are numerous, they are dated, and they describe specific experiences. A rating of 4.7 from 600 people and comments that keep mentioning short waiting times give a model concrete material to summarize.

That is why reviews sit alongside the brand mentions that AI relies on. Both are signals about you that you do not fully control, and both are read in bulk.

What the vendors actually say

Most AI companies do not publish ranking formulas, so it is worth separating documented statements from guesswork. Here is what we could confirm at the time of writing.

  • Google says local results are based mainly on relevance, distance and prominence. In its own words, prominence reflects how well known a business is, and review count and ratings factor into local ranking: more reviews and positive ratings can help. Source: how Google determines local ranking.
  • Google Maps added Gemini features in October 2024, including review summaries on place listings and the option to ask questions about a place. Google describes them as grounded in its places data and insights from the Maps community.
  • OpenAI's help pages on shopping in ChatGPT describe review summaries in product results as model-generated from reviews on public websites, and note that OpenAI does not verify those reviews or ratings.

None of these statements says a higher rating guarantees a mention. They do confirm that review data is part of what these systems read, summarize and show to people deciding where to buy.

Which review platforms matter by business type

The useful question is where customers in your sector leave reviews, and therefore where an assistant will find them. A rough map:

  • Local businesses (clinics, restaurants, trades, shops): Google Business Profile first, because it feeds Maps, local results and AI features built on them. Then the directory your sector uses, such as a doctor-booking platform, Yelp in some markets or a trade directory.
  • Hotels and travel: Google, Tripadvisor and the booking platforms where guests leave verified reviews after a stay.
  • Restaurants: Google and Tripadvisor, plus regional guides and delivery apps that show ratings.
  • Ecommerce: reviews on your product pages, marketplace reviews, and company-level platforms such as Trustpilot.
  • Software and B2B services: G2, Capterra and similar comparison sites, which also feed the best-of lists assistants quote.
  • Professional services (law, finance, consulting): Google plus sector directories and professional associations.

Spreading effort across ten platforms usually achieves less than doing two or three well. Pick the ones your customers already use and that appear when you ask assistants about your category.

Reviews and AI recommendations: what the text adds

The star average is only part of the picture. Assistants work with language, so the words in reviews matter as much as the score. A review that says "they fixed my boiler the same afternoon" gives a model something to repeat when someone asks for an emergency plumber.

This is where reviews and AI recommendations connect most directly. Questions are specific ("quiet hotel near the station", "dentist who treats anxious patients") and the reviews that mention those details help match you to them.

A fictional example: a clinic with 4.9 stars and 40 reviews that say only "great service" gives an assistant little to work with. A clinic with 4.6 stars and 300 reviews describing treatments, staff and waiting times gives it far more. You cannot write those reviews, but you can make it easy for customers to describe what actually happened.

How to get more genuine reviews within the rules

The basic method is simple: ask every customer, at the right moment, with as little friction as possible.

  1. Ask everyone, not only the customers you expect to be happy. Several platforms prohibit selective asking.
  2. Choose the moment: just after the service is delivered, the product arrives or the problem is solved.
  3. Make it one tap. Google lets you share a direct review link or a QR code from your Business Profile.
  4. Use a neutral request. Ask for an honest review and, if you like, suggest what to mention ("what we did and how it went"), never what rating to give.
  5. Build it into the process: a line in the follow-up email, a card at the counter, a reminder in the invoice.
  6. Send one reminder at most. Repeated chasing annoys customers and can look like pressure.

Review velocity matters too. A steady flow over months looks natural; fifty reviews in a week after two years of silence can trigger platform filters.

What each platform forbids

Platform rules are stricter than many businesses assume, and they change. Check the current version before you design a campaign.

  • Google's contribution policy for Maps forbids offering incentives such as payment, discounts or free goods for reviews, selectively asking for positive reviews, discouraging or prohibiting negative ones, and reviews based on a conflict of interest, such as employees reviewing their employer.
  • Trustpilot's guidelines for businesses ask you to invite customers consistently and fairly, ban incentives connected with writing or editing reviews, and exclude owners, their families, staff and competitors from reviewing.
  • Yelp goes further: its support pages tell businesses not to ask anyone for reviews, and say prompted reviews may not be recommended by its software.

Breaking these rules can lead to removed reviews, warnings on your profile or a suspended listing. Any of those is worse for your visibility than a slightly lower rating.

Responding to reviews

Google says replying shows customers you value their feedback, and recommends answering both positive and negative reviews. Replies are public text on the same page, so assistants that read reviews may read them too.

  • Be specific. Thank people for the detail they mentioned instead of pasting the same line everywhere.
  • Stay factual with criticism. Acknowledge what went wrong, explain what you changed and offer a direct channel to continue.
  • Never share a customer's personal data, argue, or use a reply as an advert.
  • Report reviews only when they break the platform's rules. Disliking a rating is not a reason, and Trustpilot says it monitors misuse of flagging.

Review schema: what you can and cannot mark up

Structured data helps machines read your pages, as we explain in our guide to schema markup for AI search. Reviews are a special case with strict rules.

Google's review snippet guidelines say that if the entity being reviewed controls the reviews about itself, its pages using LocalBusiness or Organization markup are not eligible for star review features. In plain terms, you cannot mark up testimonials about your own company to get stars. Source: Google review snippet documentation.

  • Product reviews written by real buyers on your own product pages can be marked up, if the reviews and ratings are visible on the page.
  • Never mark up ratings that are not shown, copied from another platform, or invented.
  • Keep your LocalBusiness or Organization markup for facts: name, address, hours, services and links to your profiles.

This section is general information, not legal advice. If you run review campaigns at scale, check the rules in your market with a lawyer.

  • United States: the FTC's rule on consumer reviews and testimonials took effect on 21 October 2024. It prohibits fake reviews (including AI-generated ones), buying positive or negative reviews, undisclosed insider reviews, company-controlled review sites posing as independent and certain review suppression tactics, and lets the FTC seek civil penalties. Source: FTC announcement.
  • European Union: the Omnibus Directive (EU) 2019/2161 amended the Unfair Commercial Practices Directive. Traders that give access to consumer reviews must say whether and how they check that reviews come from people who used or bought the product. Submitting or commissioning fake reviews, misrepresenting reviews, and claiming reviews are verified without reasonable checks are listed as unfair practices.

The practical rule is the same everywhere: ask real customers, publish what they say, and do not pretend your reviews are something they are not.

How to measure the effect

Track two things together. First, your review profile on each platform: count, average, monthly volume and the themes customers mention. Second, what assistants say when asked the questions your customers ask.

We explain the second part in how to measure AI visibility. Look for changes in whether you are mentioned, how you are described and which strengths the assistant repeats. If answers start quoting themes from your recent reviews, you can see reviews and AI recommendations working together.

Where RecommendedGEO fits

We do not write or buy reviews, and we do not run incentive schemes. What we do is find the platforms assistants rely on for your category, check that your profiles are complete and consistent, help you build a compliant review request process, and measure the answers month by month.

For local businesses this is a central part of our local GEO service. A free GEO audit shows where you stand today.

Frequently asked questions

Do more Google reviews make ChatGPT recommend me?

No vendor promises that. Google documents that review count and ratings help local ranking, and assistants often draw on local results and review platforms. More genuine reviews improve your odds; they do not guarantee a mention.

Can I offer a discount for leaving a review?

On Google and Trustpilot, no: both ban incentives connected with reviews. In the United States the FTC rule also bans paying for reviews that express a particular sentiment. Ask without offering anything in return.

Can I add star ratings from my testimonials to my site's schema?

Not for your own business. Google says pages where the reviewed entity controls its own reviews are not eligible for star features with LocalBusiness or Organization markup. Genuine product reviews shown on product pages are a different case.

Should I answer negative reviews?

Yes, calmly and specifically. Google recommends replying to both positive and negative reviews. A clear, factual reply shows how you handle problems, and it is public text that people and assistants can read.

How do reviews and AI recommendations relate for B2B companies?

The mechanism is the same, but the platforms differ. Software and B2B services are reviewed on comparison sites such as G2 or Capterra, and those sites often appear as sources when assistants answer questions about a category.

The four workstreams of GEO: measuring AI answers, citable content, outside presence and technical GEO
AI assistants tracked every month: ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI Overviews
Monthly GEO process: measure, fix, publish and report

Pay with Stripe

On the live site this button takes you straight to secure card payment (Stripe). This prototype doesn't charge yet.

Scroll to Top