Insights · AI visibility
Reviews and AI recommendations: why the wording matters
AI engines read and summarise your reviews. Why specific reviews get you named, how to ask for them fairly, and where to collect them on the Costa del Sol.
Reviews matter for AI recommendations because ChatGPT, Gemini and Perplexity read them and summarise what they say. A review that names the area, the service and the type of customer gives the AI a specific reason to recommend you. A review that just says “great service” gives it almost nothing.
So the star rating is only part of the story. The words matter more.
How do AI engines use reviews?
When someone asks “which dentist in Malaga is good for English speakers?”, the AI looks for evidence. Reviews are some of the best evidence it can find, because they are written by independent people describing real experiences.
The model does not count stars and pick the highest. It reads what reviewers say and looks for patterns. If ten reviews mention that the dentist speaks perfect English, the model can say so with confidence. If none do, it has to guess, and it may name another clinic instead.
AI engines often summarise reviews directly in their answers. You will see lines like “reviewers praise their clear communication in English” or “customers mention fast replies on WhatsApp”. Those phrases come from somewhere. They come from what your customers wrote.
What does a useful review look like?
Here is the difference, using a made-up estate agency.
Vague:
“Brilliant service, highly recommend! 5 stars.”
Specific:
“We are a Dutch couple and bought a townhouse in La Cala de Mijas from Rotterdam. Marta at Casa Clara Homes handled everything in English, found us a lawyer and sent video tours every week. We only flew out twice.”
Illustrative example. The business and people are made up.
Both are five stars. But the second one tells the AI five useful things: the buyer is Dutch, the area is La Cala de Mijas, the service was in English, the agency helps remote buyers, and it arranges legal help. That review can support an answer to “estate agent in Mijas for Dutch buyers” or “can I buy property in Mijas without visiting often?”.
The first review supports nothing in particular.
Which details help most?
The most useful details match the questions your buyers ask AI:
- Area. Marbella, Nueva Andalucía, Estepona, Nerja, Malaga old town.
- Type of customer. British retirees, Swedish families, Dutch second-home buyers, digital nomads.
- Service. The actual thing you did: implants, a conveyancing check, a wedding for 80 guests, a tax return.
- Language. “All in English”, “they spoke Norwegian”, “the receptionist speaks German”.
- Outcome. What changed for them, in their words.
You cannot write reviews for customers, and you should not tell them what to say. But you can remind them what would help other people like them.
How do you ask for better reviews, fairly?
- Ask everyone, at the right moment. Ask soon after the job is done, when the experience is fresh. Ask all customers, not only the ones you think are happy. Google’s rules say you should not selectively ask for positive reviews.
- Make it easy. Send a direct link to your Google review form by email or WhatsApp.
- Give a gentle prompt, not a script. Something like: “It really helps other buyers if you mention where you came from and what we helped with.” That is a nudge, not words in their mouth.
- Ask in their language. A Swedish customer may write a better review in Swedish. That is fine. AI engines read many languages, and a Swedish review helps you appear for Swedish questions.
- Never offer anything in return. No discounts, gifts or prize draws for reviews. Google’s policy does not allow incentivised reviews, and most review platforms have similar rules.
- Never write fake reviews, or ask staff, friends or family to post them. In the EU, fake reviews are an unfair commercial practice under consumer law. Platforms remove them, and they can take genuine reviews down with them.
Should you reply to reviews?
Yes. Replies are public, and AI engines can read them too. A good reply is short, personal and specific:
“Thank you, Pieter and Anna. It was a pleasure helping you find a home in La Cala. Enjoy the first summer.”
For a negative review, reply calmly, acknowledge the problem and say what you did about it. A fair reply to a bad review shows how you treat customers when things go wrong. That is useful information for a reader, human or AI.
Do not stuff replies with keywords. “Thank you for choosing the best estate agent in Mijas for Dutch buyers” reads as spam to people, and to the models.
Where should your reviews be?
Google is the most important for most local businesses. But AI engines look across many sources, and a spread of reviews looks more trustworthy than all of them in one place.
- Google Business Profile. The starting point for almost everyone.
- Trustpilot. Often cited for service businesses and agencies.
- Sector platforms. Doctoralia and Top Doctors for clinics. TripAdvisor and TheFork for restaurants. Booking.com for hotels. The major property portals for estate agents.
- Facebook and local community groups, where expats ask each other for recommendations.
You do not need to be everywhere. Find the two or three platforms that AI engines cite for your questions, and focus there. Our guide on how to check what ChatGPT says about your business shows how to find them.
Where reviews fit in the bigger picture
Reviews are one part of being recommended. They work best alongside a clear website, consistent listings and mentions in the local press. Our guide on how to get your business recommended by ChatGPT covers the rest, and how AI cites sources explains why independent sources carry so much weight.
See what AI says about you now
Our free AI visibility check tests ten buyer questions across ChatGPT, Gemini and Perplexity, compares you with three competitors, and shows the sources the engines rely on when they name someone else. You get the report within five working days.