Fine Dining · Helsinki

AI Visibility Report: Helsinki Fine Dining

40 brands · 3 AI models · 100 prompts · 600 responses · July 2026

The Thesis

A diner asking an assistant where to eat in Helsinki does not get ten links to weigh up. They get three or four names. There is no page two in an AI answer: a restaurant is named, or it is nowhere. We tested more than 100 dining questions in English and Finnish across ChatGPT, Google AI Mode and Gemini, and scored more than 600 answers to see which rooms the machine puts in front of a guest. This measures visibility only. It is not a judgement of anyone's food, service or hospitality.

The answer is far narrower than the city is. Six restaurants take 59.2% of every restaurant appearance we scored, and the top ten take 77.0%. The median venue in the study sits at roughly 1% visibility, which in practice means invisible. A Michelin star does not fix this: eight of Helsinki's thirty guide-listed restaurants appear in under 2% of responses. The guide recognition is real, but the digital footprint the models read from is thin.

Three findings sit underneath the ranking. The models disagree violently with each other: Boreal is named in 73.6% of ChatGPT answers and 12.7% of Google AI Mode answers, a 60.8-point spread hiding inside a single aggregate fourth place. English and Finnish prompts return different cities, so a room can lead the tourist conversation and be underweight with locals. And the models still recommend a restaurant that has closed, which tells you exactly how far behind the web an AI answer can run.

“A Michelin star does not buy a place in the machine's answer. Eight of Helsinki's thirty guide-listed rooms appear in under 2% of AI recommendations, and one restaurant the models still recommend has closed.”

34.6%#1 visibility scorePalace

~1%Median visibilityMost rooms are invisible to AI

77%Held by the top tenof all restaurant appearances

3AI surfaces testedChatGPT, AI Mode, Gemini

AI Visibility Rankings

Who AI recommends in helsinki fine dining

Ranked by weighted AI visibility across ChatGPT, Google AI Mode and Gemini, averaged over two collection days. These positions describe exposure in AI answers. They do not assess restaurant quality.

#BrandAI VisibilityModels

Palace

34.6

3/3

Grön

31.5

3/3

Olo

22.4

3/3

Boreal

16.6

3/3

Finnjävel Salonki

16.5

3/3

Demo

3/3

Nokka

4.7

3/3

Natura

4.2

3/3

Savoy

3.9

3/3

Plein

2.9

3/3

30 more venues in the full report, with per-model and per-language breakdowns.

What's in the Full Report

The full leaderboard

All 40 venues, including the second tier where visibility drops under 2.5% and the differences narrow to a point or two. Below sixth place these are better read as a close group than a strict order.

Three models, three cities

The aggregate hides the disagreement. Boreal is a ChatGPT restaurant at 73.6% and nearly absent from Google AI Mode at 12.7%. Ragu is the inverse: invisible on ChatGPT and Gemini, named in almost half of AI Mode answers.

The answer changes with the language

English and Finnish prompts do not produce the same dining map. Finnjävel Salonki swings 16.8 points between them. English visibility shapes tourists and international press; Finnish shapes local consideration.

The sources AI reads

Guides and food media drive 48% of all citations. timeout.com is cited 370 times, myhelsinki.fi 227. A restaurant's own site accounts for 28%, which makes it the cheapest lever any operator controls.

What the models get wrong

A restaurant that closed still appears in roughly 7% of answers, described as open and starred. Asked specifically for fine dining, the models name a bakery or a seafood counter 8% of the time.

Where different tables win

Palace takes 71.1% of occasion prompts and 87.5% of tourist prompts. Grön takes 88.1% of vegetarian and plant-based questions. Nokka takes 61.5% of cuisine-led ones. Owning an occasion beats ranking well in general.

Full Table of Contents