Case study
Reaching diners beyond ramen searches
Multi-location Copenhagen ramen restaurants
In one measured month, casual-dining visibility rose from 9.7% to 15.9% and visitor discovery from 12.2% to 17.6%. Ramen was already strong at 93% and closed at 96%.
Measured visibility in AI answers
9.7% → 15.9%Casual dining visibility+64% visibility for casual-dining questions14 August to 14 September 2026The restaurant group was already highly visible for ramen, but guests asking broader casual-dining and visitor questions had fewer reasons to encounter it. After a new 20-article blog and daily question tracking, the reported readings were higher on those discovery surfaces while the core category remained strong.
Measurement window
What exactly was measured?
- Baseline
- 14 August 2026. The report records category and visitor baselines before the month of publication activity.
- Current comparison
- Week ending 14 September 2026. The report tested 95 guest questions daily across ChatGPT, Gemini and Google AI answers in four markets. It does not establish that the prompt set was fixed throughout the month.
Work
What changed during the month?
- 01
A blog was built from no published articles to 20 articles during the first month. A production check on 19 August still found the blog route unavailable before that publishing sequence was complete.
- 02
The work focused on occasions and visitor situations, such as relaxed casual dining and first-trip discovery, rather than spending the whole month reinforcing the core ramen category.
- 03
The measurement ran continuously across guest-style questions, then separated category strength, broader discovery and tourism outcomes so a strong core category could not hide weak occasion coverage.
- 04
The restaurant lesson is practical: category dominance does not automatically create discovery for location, occasion or visitor intent. Coverage has to meet the question a guest actually asks.
Inside the programme
Being known for ramen was not the same as being discovered for an occasion
The starting point was not a weak restaurant brand. Ramen visibility was already 93%. The gap sat in the moments when a guest asks for a relaxed meal, somewhere after sightseeing, or an option without a reservation.
Those are illustrative question shapes, not verbatim tracked prompts. They show why a restaurant can dominate its food category and still be absent when a visitor asks the question that starts a meal plan.
Inside the programme
The first library answered the wider reasons to choose
Twenty articles were published in the first month, after a production check had still found no live blog on 19 August. The report records that 17 of those 20 articles were retrieved by assistants.
The library was designed for practical discovery decisions, including no-reservation guidance, visitor context and occasion-led choices. Retrieval is evidence that assistants found the material, not proof that any one article caused an answer.
Inside the programme
The result was stronger discovery, with a visible exception
The chart uses a common 0 to 100% scale. Casual dining and visitor discovery rose, while ramen stayed high. Japanese and wider Asian visibility moved the other way, from 66.7% to 52.8%, so the result is not a claim that every surface improved.
The report also records 2,374 owned-site citations and 1,933 tracked answers containing the site, equal to 19% of tracked answers. These are different counts. Third-largest source means third only inside the tracked Copenhagen sample, behind two official tourism sources.
Inside the programme
The operator takeaway is to map the question before writing
A restaurant should test more than “best [food]”. It should also test illustrative questions such as “where can we eat after visiting [landmark] without booking?” and the occasion, location and visitor variants that sit around a meal decision.
That is the practical distinction in this case: defend a category people already associate with you, then build helpful evidence for the occasions where they do not yet think to look.
Prompt set
Did the prompt set stay fixed?
The report documents 95 questions tested daily, but it does not establish a fixed prompt denominator for every day of the month. The before-and-after readings are reported as a measured programme result, not as a controlled causal experiment.
Engine coverage
Which engines were included?
The report names ChatGPT, Gemini and Google AI answers, tested across four markets. Google search traffic and AI-answer visibility are separate measures, and this study does not convert either into bookings or revenue.
Definitions
How are the metrics defined?
- AI visibility
- The share of tracked AI answers that name the restaurant group.
- Owned-site citation
- A reported citation of the restaurant group's website in a tracked AI answer.
- Answer retrieval
- A tracked answer in which the report recorded the restaurant group's site as present.
Interpretation
What should this result not be taken to mean?
- The study measures answer presence and citations, not restaurant bookings, covers or revenue.
- The cited-site count and the count of answers containing the site are different measures and are not presented as a single growth rate.
- Third-largest source means third within the tracked Copenhagen sample, behind two official tourism sources. It is not a city-wide ranking.
- Japanese and wider Asian visibility fell from 66.7% to 52.8%, so the result is not a claim that every discovery surface improved.
Next step
Find the occasions where discovery is thin
A free audit starts with the buyer and guest questions where a business is missing, then separates category recognition from the moments that create discovery.