We tracked how three AI systems, ChatGPT, Gemini and Google AI Overview, answer real questions about eating in Copenhagen. Across repeated daily runs we logged 1,328 usable answers to 190 questions grouped into eight topics: the best burger in the city, neighbourhood and location, quick casual and value, occasions and groups, quality and reputation, first-trip and tourist questions, the general question of where to eat, and dietary and sourcing questions. Every one of the 4,192 times an engine named a restaurant, food hall or market across those answers, we counted it against the same brand and the same denominator. The finding worth remembering: the brand AI names most often on its specialty question is named four times less often on the three question groups that actually decide where someone sits down for dinner.
The short version
- The most-named brand appears in 29.8 percent of all 1,328 answers, the highest of 27 brands that scored above zero.
- That same brand is named in 97.2 percent of the narrow question about the city's best burger. On the general "where to eat" question, first-trip and landmark questions, and neighbourhood questions combined, 579 answers, it is named in just 21.9 percent, a gap of 75 points.
- A second burger bar shows the identical shape at a smaller scale: strong on the specialist question, level with the leader on the general one.
- The city's own tourist board runs the answer: visitcopenhagen.com is opened in 30.2 percent of all 1,328 answers, more than any brand's own website, including the leader's, which is opened in 1.5 percent even though the brand itself is named in 29.8 percent.
- None of the 200 tracked questions returned a paid placement.
What we measured, and how
We tracked 29 restaurants, food halls and markets operating in Copenhagen today, each measured against the same 1,328-answer denominator. Two do not appear in the tables below. One never came up in a single one of the 1,328 answers, so there is nothing to publish. The other, a restaurant that closed in February 2024 and reopened in August 2026 under new ownership and without its previous stars, is excluded for a second reason on top of that timing problem: a share of the answers naming it were actually naming a burger bar it once ran as a side project, which we track separately under its own name. Counting both would have doubled part of one score. That leaves 27 published brands and 3,739 of the 4,192 raw brand names once the excluded restaurant's share is set aside.
Four numbers describe each brand:
- How often named: the share of the 1,328 answers that name the brand.
- Share of the names: the brand's share of every name handed out, since one answer can name more than one place.
- Tone: 0 to 100, how positively an engine writes about a brand when it appears.
- Average position: where in a list the brand lands when it is named. Lower is better.
The public measurement receipt carries the full 27-brand table, the exclusion reasoning, and every source and pillar figure in this report.
Who does AI actually name? The Copenhagen leaderboard
| Rank | Brand | How often named | Share of the names | Tone | Average position |
|---|---|---|---|---|---|
| 1 | Gasoline Grill | 29.8% | 18.1% | 72 | 2.0 |
| 2 | Reffen | 18.5% | 9.1% | 66 | 2.8 |
| 3 | Torvehallerne | 17.4% | 8.6% | 69 | 3.0 |
| 4 | Schønnemann | 14.6% | 6.5% | 69 | 1.9 |
| 5 | Sporvejen | 13.3% | 5.5% | 73 | 2.9 |
| 6 | POPL Burger | 10.6% | 5.7% | 67 | 2.9 |
| 7 | Høst | 8.9% | 4.1% | 77 | 2.2 |
| 8 | Bæst | 8.6% | 3.7% | 74 | 2.8 |
| 9 | DØP | 8.1% | 4.4% | 63 | 3.2 |
| 10 | Kødbyens Fiskebar | 7.9% | 3.0% | 73 | 2.5 |
| 11 | Tommi's Burger Joint | 7.2% | 3.3% | 70 | 4.1 |
| 12 | Sanchez | 6.9% | 3.7% | 62 | 3.6 |
The drop after that is not steep. The median across all 27 published brands is 5.5 percent. Every brand outside the top ten sits below 7.3 percent, a ceiling held by Tommi's Burger Joint, and at the bottom, three brands, Banana Joe, Grillen Burgerbar and Burgerklubben, are each named in exactly one of the 1,328 answers, 0.08 percent. The top two brands alone take 30.4 percent of every name handed out across the category; the other 25 split what is left.
The real finding: two different Copenhagens
The leaderboard above blends eight kinds of question into one number, and that blend hides the actual story. Split the same answers by what the question is really asking, and Copenhagen dining splits into two cities that barely touch.
One is the narrow specialist question: who makes the best burger in town. On that question alone, 145 answers, the market leader is named in 97.2 percent, close to a clean sweep.
The other is what a hungry person actually types before picking somewhere to eat: the plain "where should I eat in Copenhagen" question, the questions a first-time visitor and a landmark search generate, and the neighbourhood question. Combined, those three groups cover 579 answers, close to half the study. On that combined question, the same brand that wins the burger question is named in just 21.9 percent, a drop of 75 points and a fall to roughly a fifth of its specialist score.
| Question group | Answers | Gasoline Grill named |
|---|---|---|
| Best burger in Copenhagen | 145 | 97.2% |
| Where to eat in Copenhagen | 185 | 10.8% |
| Tourist and first-trip questions | 204 | 20.1% |
| Neighbourhood and location | 190 | 34.7% |
| Three decision groups, combined | 579 | 21.9% |
This is not one brand's fluke. Sporvejen, a second burger bar in the same city, shows the identical shape at a smaller scale: 31.0 percent on the burger question, then 10.8 percent on the general where-to-eat question, exactly level with the market leader on that measure. Two different brands, two different burger scores, and the same thin number once the question stops being about burgers.
| Best burger question | Where to eat question | |
|---|---|---|
| Gasoline Grill | 97.2% | 10.8% |
| Sporvejen | 31.0% | 10.8% |
The visitor gap gets sharper once you separate landmarks from guides
The first-trip and tourist group above, 204 answers, is itself two different questions. A general visitor question, "best cheap eats in Copenhagen," "best street food in Copenhagen" and similar, names the market leader in 33.3 percent of its 111 answers. A landmark-anchored question, someone asking what to eat near Nyhavn, Amalienborg, Tivoli Gardens or eleven other named sites, names it in just 4.3 percent of 93 answers. The same shape repeats by meal: a lunch question names the leader in 21.3 percent of 47 answers, a dinner question in 7.7 percent of 26. Read together, the closer a question sits to an actual moment of choosing where to sit down, an evening, a landmark, an occasion, the thinner the leader's showing gets.
The organic and sustainable question favours nobody near the top
One question group breaks the leaderboard entirely: eight questions about organic and sustainably sourced food, 52 answers. The brand with the highest score across the whole study is not the leader here.
| Brand | Named on the organic and sustainable questions |
|---|---|
| Bæst | 46.2% |
| DØP | 26.9% |
| Beyla | 19.2% |
| POPL Burger | 11.5% |
| Torvehallerne | 11.5% |
| Høst | 9.6% |
| Jagger | 7.7% |
| Dandelion Burger | 5.8% |
| Gasoline Grill | 5.8% |
| Schønnemann | 3.8% |
| Halifax | 3.8% |
Bæst, a pizza restaurant, leads with 46.2 percent, ahead of DØP at 26.9 and Beyla at 19.2. The overall market leader from the main table appears in just 5.8 percent of these answers, tied with Dandelion Burger and behind three brands with far lower scores everywhere else. A strong showing in the main leaderboard does not travel to every question, and this group is the clearest proof of that in the dataset.
Which sources decide the answer
| Source | Type | Opened in share of the 1,328 answers |
|---|---|---|
| visitcopenhagen.com | Institutional | 30.2% |
| visitcopenhagen.dk | Institutional | 16.8% |
| copenhagentourism.org | Institutional | 8.6% |
| tripadvisor.com | Corporate | 7.0% |
| reddit.com | Community | 7.0% |
| copenhagen-trip.com | Editorial | 5.1% |
| cphhygge.dk | Editorial | 4.9% |
| visitdenmark.com | Institutional | 4.9% |
| tripadvisor.dk | Community | 4.8% |
| tripadvisor.co.uk | Community | 4.5% |
| kobstaden.dk | Editorial | 4.0% |
| restaurantsforkings.com | Editorial | 3.7% |
| facebook.com | Community | 3.5% |
| falstaff.com | Editorial | 3.4% |
| theinfatuation.com | Editorial | 3.2% |
Three things stand out. First, a Copenhagen dining answer is written almost entirely by the tourist board: visitcopenhagen.com and its Danish sibling visitcopenhagen.dk are opened in 30.2 and 16.8 percent of all 1,328 answers, more than every other source on this list combined. Second, the market leader's own site is barely part of the conversation. It is opened in 1.5 percent of answers even though the brand itself is named in 29.8 percent of them, and every other tracked brand's own site does even less, averaging 0.24 percent. That is not one site losing to a better-built rival page. Nobody's own website is where this argument gets won. Third, no paid placement appeared anywhere across the 200 tracked questions. Whoever gets named earns it through a source, not a media budget.
Question coverage tells the same story from another angle: at least one of the 27 published brands appeared in 108 of the 190 tracked questions, 56.8 percent. The other 82 questions, 43.2 percent, returned none of the tracked brands at all, which points to a wider bench of Copenhagen restaurants an engine will reach for once a question moves away from the handful of names that already dominate the general conversation.
What this index can and cannot tell you
Treat this as a bounded read of one city in one measurement window, not a permanent ranking or a verdict on food quality. It can tell you which of 27 tracked Copenhagen brands AI systems named most often, how sharply that changes once a question moves from a specialist topic to a general one, and which sources carried the answer. It cannot tell you whether a name led to a booking, whether a restaurant outside the tracked cohort would have scored higher (this is a measured sample, not a census of every kitchen in the city), or whether the same gap holds in a different week. It also does not settle culinary quality: a kitchen can be better and still lose the citation.
For the same shape of measurement in a different city, see our UK Restaurant AI Visibility Index, a separate market and a separate window, useful for scale rather than direct comparison.
What this means if you run a restaurant
If your place wins its specialist question and disappears on the general one, that is not a reputation problem, it is a sourcing problem. AI systems have learned exactly what you are known for and answer accordingly. They will not promote you into a broader question your own pages, reviews and press coverage never argued for. Look at what actually gets retrieved for the general question in your city, usually a tourist board, a listings page or a community thread, and check honestly whether your restaurant can appear on it at all.
Landmark and neighbourhood questions reward different evidence than a specialist one does. A guide built around a place a visitor already knows, a street, a station, a park, needs a reason to include you that has nothing to do with your specialty. That reason has to exist somewhere public before an engine can repeat it back.
If you want your own figure rather than a borrowed one, the free AI search audit shows how often the major AI engines name your restaurant for the questions diners actually ask, and which sources are deciding that answer instead of you.
Sources
- Schmitdy, Copenhagen Dining AI Visibility Index measurement receipt: full methodology, exclusions, pillar figures and source table behind this report.
- Schmitdy, UK Restaurant AI Visibility Index: the same measurement shape applied to London, a separate market and window.
Squarespace
Webflow
Sanity
Shopify
HubSpot