Why your 200 Google reviews don't help ChatGPT

The review paradox

A dental practice in Basel has 200 Google Maps reviews and a 4.9-star rating. Their competitor has 80 reviews and a 4.6-star rating. The competitor gets cited by ChatGPT three times more often.

This is not a bug. It reflects a fundamental difference between how *humans* use reviews and how *AI engines* use them.

How AI engines read reviews

When a model synthesises an answer to "best dentist in Basel with good reviews", it doesn't count stars. It reads review *content* for:

• →Named procedures · Does this business do Bleaching? Implants? Invisalign? Reviews that name specific treatments tell the AI what the business actually specialises in.
• →Patient scenarios · Reviews mentioning "Angstpatienten", "Notfall", "Kinder", "Expats" signal that the practice serves specific high-intent queries.
• →Language-specific signals · A review written in French that mentions "patients francophones" makes the practice visible to AI engines answering French-language queries.
• →Sentiment specificity · "Excellent bleaching result, done in one session" gives the model something concrete to cite. "Sehr nett" does not.

In Klaréo's analysis of 500+ Swiss businesses, the ones with the highest AI citation rates have reviews that average ~40 words and mention at least one specific service or scenario. Generic review content · regardless of star rating · correlates weakly with AI citation.

The keyword signal problem

Most Swiss businesses have review content that looks like this:

"Sehr netter Arzt, sehr empfehlenswert. Freundliches Team."

"Super Praxis, alles bestens."

"Sehr professionell und nett."

These reviews are positive but signal-poor. An AI engine sees a well-liked business with no specific information to cite.

Compare to a review with keyword signal:

"Ich litt unter Zahnarztangst und wurde hier zum ersten Mal wirklich ernst genommen. Dr. Müller hat sich für mein Bleaching Zeit genommen und alles erklärt. Notfalltermin wurde noch am selben Tag eingeräumt. Neue Patienten werden herzlich empfangen."

This review mentions: Zahnarztangst, Bleaching, Notfalltermin, neue Patienten. It covers four high-intent query types in one review.

How to improve review signal without faking reviews

The fix is response templates, not fake reviews. Here's the approach:

Ask at the right moment. A satisfied patient leaving the practice is the best moment to ask for a review. The memory of the specific treatment is fresh and they'll naturally include it.

Give them a starting point. Not a script · a prompt. "If you'd like to mention what brought you in or what we helped you with, that's really helpful for other patients looking for the same care."

Respond to all reviews, mentioning the specific service. Your response to a review is also indexed. A response like "Thank you for your trust · we're glad the Bleaching session exceeded your expectations, and we look forward to seeing you for your next check-up" adds keyword signal even when the original review doesn't have it.

Diversify your review platforms. Google Maps is primary, but AI engines also index Medicosearch, Doctorfmh, and Jameda. A practice with reviews only on Google Maps has less cross-source signal than one present on three platforms.

The language distribution problem

In Switzerland, most practices get 90%+ of their reviews in German. If 12% of your city's population is English-speaking and 20% is French-speaking, you're invisible in AI answers to those communities.

You can't ask patients to write reviews in specific languages · that would be artificial. What you can do:

• →Make it easy for francophone/anglophone patients to leave reviews by providing the link in their language in follow-up communications.
• →Ensure your Google Maps description and website have FR/EN content (reviewed separately), so AI engines have something to cite even without FR/EN reviews.
• →Register on French-language directories (e.g. Doctorfmh, which has a FR interface) where your listing itself carries French-language signal.

What to measure

A practical metric: take your last 50 reviews and count how many mention a specific service, treatment, or patient scenario. If fewer than 30% do, you have a keyword signal problem worth addressing. The fix is conversation-level · it happens at the point of care, not at the keyboard.