Do Guest Reviews Affect AI Hotel Recommendations? What New Research Shows (2026)

Yes. In a 2026 study of 12 AI models, a top guest rating raised a hotel's chance of being recommended by 31.6 percentage points. Review volume and recency mattered too, but much less. Replies to reviews had no direct effect. Price was the only factor as strong as the rating.

In short:

• A 4.7 rating instead of 3.9 raised a hotel's chance of being recommended by 31.6 percentage points.
• 2,100 reviews instead of 45 added 8.3 points. A recent review instead of an 11-month-old one added 1.6.
• Replying to reviews had no measurable direct effect on AI's choice.
• Price had about as much weight as the rating, in the other direction.
• The study tested how AI weighs a hotel's reputation, not how it finds hotels online.

{{cta-vt="/features/product-updates"}}

Do guest reviews affect AI hotel recommendations?

They do, mainly through your rating. In the study "Whose hotel does the AI recommend?" by Baig, Gillani and Ali, published in Information Technology & Tourism, a hotel rated 4.7 was 31.6 percentage points more likely to be recommended than an otherwise similar hotel rated 3.9. Even a step from 3.9 to 4.3 added 10.1 points. This held across all twelve models tested.

How did the study test AI hotel recommendations?

The researchers gave AI models five hotels at a time and asked them to pick one for a traveller. The hotels were fictional profiles. For each profile, seven signals were set at random: guest rating, number of reviews, how recent the latest review was, whether the hotel replies to reviews, chain or independent, price and eco certification. The order of the hotels in the list varied too.

Each model made 3,024 choices, across three types of traveller and nine ways of asking. The twelve models included versions of ChatGPT (GPT-4o mini), Gemini and Claude, plus four open models. Because every signal varied independently, the study can show how much each one changed the AI's choice.

Which review signals matter most to AI?

• Rating 4.7 instead of 3.9: +31.6 percentage points
• Rating 4.3 instead of 3.9: +10.1 points
• 2,100 reviews instead of 45: +8.3 points
• 420 reviews instead of 45: +2.3 points
• Latest review recent instead of 11 months old: +1.6 points
• Hotel replies to reviews: +0.1 points (no measurable effect)

The rating is by far the strongest review signal. Review volume helps, but mostly at high numbers. Recency matters a little.

What else influences which hotel AI picks?

• Price: $249 instead of $129 per night lowered the chance by 30.0 points, about as much as a top rating raised it.
• Eco certification: a Green Key label added 11.6 points, with big differences between models.
• Chain or independent: chains were 1.8 points less likely to be picked, a small difference.
• Position in the list: being listed fifth instead of first cost 3.7 points on average. One model showed a much stronger first-place bias.

Do review replies help your AI visibility?

Not directly. In the study, whether a hotel replied to reviews made no measurable difference to which hotel AI picked.

Our take: that doesn't mean replies don't matter for AI visibility. Research on hotel reviews shows that replying affects the reviews that follow. In a study of hotels on Tripadvisor, Proserpio and Zervas found that hotels that started replying gained 0.12 stars, received 12% more reviews and got fewer negative ones. Replies may work through your rating: they improve your reviews, and your rating then drives AI recommendations. We see two plausible ways this happens:

1. Positive feedback gets reinforced. When guests see that every review receives a personal reply, they feel heard. That motivates happy guests to leave a review too.
2. Fewer public complaints. Most people avoid public conflict. If a hotel visibly replies to every review, unhappy guests may be more likely to contact the hotel directly instead of posting a negative review.

A study that tests replies, ratings and AI recommendations together would be the next step, and one we'd be keen to see.

What are the limits of the study?

• Fictional hotels, no web search. The hotel profiles were written into the question. In real life, AI first has to find your hotel online and read your rating somewhere. The study shows how AI weighs a reputation once it has the facts, not how it finds them.
• No photos, review texts or amenities. Real decisions draw on more, so the exact numbers may differ in practice.
• English, US-style prompts. Results may vary in other languages and markets.
• Models change. The weights can shift with every model update.

A study of real AI answers points the same way. In Stiplo's analysis of 27,360 AI travel answers in Barcelona, the number of Google reviews was the factor most strongly linked to being recommended. Our guide to GEO for hotels covers how AI platforms find and read hotel information.

What should hotels do?

1. Treat your rating as an AI visibility lever. Fix the issues that come up again and again in your reviews; that's what moves the rating.
2. Keep new reviews coming in, especially on Google. Volume and recency both count.
3. Reply to every review. Not because AI reads your replies, but because replies can shape the reviews that follow.
4. Check what AI says about you. Ask the questions your guests ask, and look at which hotels are recommended instead and why. Our article on what AI visibility tracking is explains how.

MARA's reputation management software helps you reply to every review and see which topics move your rating. The AI Visibility Tracker shows how often AI platforms recommend your hotel.

{{cta-vt="/features/product-updates"}}