Do you already know where you’re eating this summer holiday?

AI probably does.

We ran an analysis on one of Europe’s most famous tourist destinations to understand what AI is recommending people to eat, and the results were surprising. AI recommendations differ significantly from traditional platforms like Tripadvisor. But even more interesting: they also differ depending on which LLM you use. For the analysis we used a specific target group profile of young professionals with specific interests, so the answers are tailored and here are some findings:

• The restaurant most recommended by LLMs overall is only #8 on Tripadvisor — and it’s the only restaurant appearing in the Top 15 of both rankings. Everything else is completely different.

• OpenAI consistently recommends Periplo and Club Nàutic Portixol across multiple queries. These places barely appear in other LLM recommendations, and on Tripadvisor they hold ratings of 4.0 and 3.1.

• Gemini clearly has its own favorites: Ca n’Eduardo, Mar de Nudos, and El Bungalow. These are also mentioned by other models, but with much lower frequency. Their Tripadvisor ratings are 3.7, 3.5, and 3.7.

So where do LLMs get their recommendations from?

Surprisingly, the most cited source is still Tripadvisor itself, appearing 35 times across more than 500 source mentions.

But after that, the long tail becomes extremely fragmented, and local authority sources start playing a major role.

The most cited sources after Tripadvisor were:

• mallorcaguide
• abcMallorca
• mallorcard
• palmaweekly
• seemallorca
• El País
• estilo palma
• mallorca-magic

This raises an interesting new visibility approach

With AI becoming the new recommendation engine for travel and hospitality, restaurants shouldn’t just ask tourists to leave good Tripadvisor reviews.

They’ll also need to understand which local and niche publishers are shaping AI-generated recommendations — and build PR strategies around them.