AI Visibility Index · 2026-q3
Which shops does AI recommend to Norwegian shoppers?
We asked four AI assistants 20 unbranded buying questions about sport and outdoor gear in Norway, 400 answers in total, and counted who they name. No shop is named in more than three answers in five, and the engines disagree with each other more than the shops differ from one another.
- Answers
- 400
- Engines
- 4
- Questions
- 20
- Captured
- 2026-08-16
Sport and outdoor, Norway
Ranked by presence rate: for each engine, the share of that engine's answers naming the company, then the mean across the four. Shops only. Manufacturers and review sites are left out, because a question about where to buy a tent names Helly Hansen and XXL for different reasons.
| # | Company | Named in | Spread across engines | ChatGPT | Gemini | Copilot | Claude |
|---|---|---|---|---|---|---|---|
| 1 | XXL | 59.2% | 62.0% | 75.0% | 50.0% | 50.0% | |
| 2 | Fjellsport | 40.2% | 36.0% | 57.0% | 44.0% | 24.0% | |
| 3 | Outnorth.no | 37.2% | 33.0% | 46.0% | 28.0% | 42.0% | |
| 4 | FINN.noMarketplace | 19.0% | 14.0% | 25.0% | 10.0% | 27.0% | |
| 5 | Prisjakt.noPrice comparison | 18.5% | 15.0% | 37.0% | 16.0% | 6.0% | |
| 6 | Naturkompaniet | 17.0% | 13.0% | 26.0% | 18.0% | 11.0% | |
| 7 | Hekta På Tur | 15.5% | 11.0% | 29.0% | 12.0% | 10.0% | |
| 8 | Intersport | 14.5% | 17.0% | 13.0% | 20.0% | 8.0% | |
| 9 | Sport 1 | 14.5% | 20.0% | 15.0% | 7.0% | 16.0% | |
| 10 | Bergans | 13.8% | 10.0% | 19.0% | 7.0% | 19.0% |
Each strip shows one company on a 0 to 100 per cent track: a light tick per engine, and the blue mark at the mean. The empty right half is the finding, not a layout choice.
Three things worth knowing
Nobody owns the category
XXL leads and is still named in only 59.2% of answers, so two in five answers about where to buy sport gear in Norway do not mention the market leader at all. Below third place, nothing on the table is named in even one answer in five. For a retailer outside the top three, this is a category where the ranking is still unsettled.
The engine matters more than the brand
Prisjakt is named in 37.0% of Gemini's answers and 6.0% of Claude's, a spread of 31 percentage points. That gap is wider than the whole distance between second and sixth place on this table. Which assistant a customer happens to open changes the answer they get more than which shop they were asking about.
One answer in seven recommends nobody
14.5% of the 400 answers named no shop at all: no retailer, no marketplace, no price comparison site. On ChatGPT and Copilot it is 19.0%, on Gemini 9.0%. Those answers stay in the denominator here, because an assistant that recommends nobody is a real miss for every shop in the category, and leaving them out would lift every number on this table.
Method
Everything below was fixed before the run and is published with the numbers. The questions, the engines, the sample count and the definition of the metric are all here, so the run can be repeated and the numbers compared.
Unbranded questions only
None of the 20 questions names a shop. A question that names one measures the question rather than the market. The wording was frozen before the run and is printed in full below.
Five samples per question, per engine
These models are not deterministic: ask twice and you can get two different lists. Each question was asked 5 times on each engine, so a single lucky answer cannot move a rank. That is 20 questions x 4 engines x 5 samples = 400 answers.
The mean across engines, not a pooled rate
For each engine we take the share of that engine's answers naming a company, then average the four. A talkative engine cannot carry a company up the ranking by answering more often.
Answers naming nobody count
Every answer is in its engine's denominator, including the ones that recommended no shop. This is deliberately different from how we report a customer's own share of voice, where the question is 'when AI names shops, how often is it you'. Here the misses are part of the finding.
Shops only
Companies are labelled by what they are, and the table keeps retailers, marketplaces and price comparison sites. Manufacturers and review sites are dropped: mixing them ranks two different questions at once.
The engines, and what was actually queried
- ChatGPTgpt-5.6-luna
- Geminigemini-3.5-flash
- Copilot (Bing-grounded)gpt-5.6-luna
- Claudeclaude-sonnet-4-6
Each assistant is reached through its provider's API as a stand-in for the consumer app. Copilot is GPT grounded in Bing web search, which is the index behind Copilot, not Copilot's shopping carousel.
The 20 questions, verbatim
Asked in Norwegian, because that is what a Norwegian shopper types. They are printed here in the original so anyone can repeat the run.
- 1Hvilken nettbutikk er best for sportsutstyr i Norge?
- 2Hvor bør jeg kjøpe turutstyr på nett?
- 3Hva er de beste nettbutikkene for løpesko?
- 4Hvem har best utvalg av friluftsutstyr på nett i Norge?
- 5Hvilke norske nettbutikker er billigst på treningsklær?
- 6Hvor får man kjøpt godt fjellutstyr på nett?
- 7Jeg skal kjøpe telt til fjelltur, hvilken nettbutikk bør jeg bruke?
- 8Trenger nye skisko før vinteren, hvor handler jeg dem på nett?
- 9Hvor kan jeg kjøpe sykkel på nett med levering i Norge?
- 10Jeg vil kjøpe sovepose til høstturer, hvilken butikk anbefaler du?
- 11Hvor bestiller jeg treningsutstyr til hjemmegym?
- 12Hva bør jeg se etter når jeg kjøper tursko på nett?
- 13Hvordan velger jeg riktig størrelse på skiutstyr når jeg handler på nett?
- 14Er det verdt å kjøpe dyrt regntøy til fjellturer?
- 15Hva slags utstyr trenger jeg for å begynne med langrenn?
- 16Hvordan finner jeg gode tilbud på sportsutstyr på nett?
- 17Jeg bor i Tromsø og trenger vinterutstyr levert raskt, hvem leverer dit?
- 18Hvor kan jeg kjøpe brukt eller rabattert friluftsutstyr i Norge?
- 19Trenger fotballsko til barn, hvilken nettbutikk har best utvalg i barnestørrelser?
- 20Hvor kjøper jeg utstyr til fisketur på nett?
What this does not tell you
One category, one market, one point in time. AI answers move, and a cycle is a snapshot rather than a live tracker. Ranks close together are not meaningfully apart: with 100 answers per engine, small gaps are inside the noise, and the interesting signal is the spread between engines rather than a one place difference between two shops.
See where you stand in your own category
This is the same measurement we run daily for customers, on the questions their own buyers ask. If you sell in a category we have not indexed yet, the fastest way to see your position is to track it directly.