This is a recurring, neutral study of a question nobody had measured: when someone in the Netherlands asks an AI assistant which bicycle or e-bike to buy, which brands does it name?
We ask five large language models the same Dutch-language questions real people type, repeat each one five times, and record how often, how early, and across how many assistants each entity appears. The result is an AI Visibility Score — a measure of findability, not of quality. A brand ranks highly here because these systems have information about it and surface it readily. That is all it means.
Why this series exists
Three facts sit next to each other, and the gap between them is the reason for this research.
1. This is a large, well-documented market — at the aggregate level. BOVAG and Koninklijke RAI Vereniging publish the one Dutch consumer-market series with a named research executor: 795,968 new bicycles sold in 2025, worth €1.53bn, research conducted by NielsenIQ (BOVAG, Fietsverkoop loopt terug in 2025, 9 February 2026). The prior year, on the same joint study executed by GfK, was 858,126 units and €1.55bn. Average selling price was €1,925 for a bicycle and €2,872 for an e-bike. The installed stock is roughly 24.4 million bicycles (BOVAG/RAI, Kerncijfers Tweewielers 2025, 24 April 2025) — an industry estimate, not a registration count, and the accessible page does not state its stock-estimation method.
Cycling here is infrastructure rather than a hobby. On CBS’s corrected ODiN 2024 figures — republished on 31 March 2026 after a method break, with 69,185 usable respondents — the bicycle carries 27.7% of all trips and 9.2% of passenger kilometres. Both percentages are calculated from CBS’s published counts rather than printed by CBS. The e-bike share of bicycle kilometres has risen from 24% in 2019 to 40% in 2024 (KiM, Addendum Mobiliteitsbeeld 2025, March 2026), a trend-model series that should not be mixed mechanically with raw ODiN totals.
Two credible e-bike sales shares disagree, and the difference is definitional. BOVAG/RAI put e-bikes at 49% of units and 73% of revenue in 2025, but their release does not publish a technical definition and treats fatbikes as a separate sales category. The Ministry of Infrastructure and Water Management’s Invloed van de elektrische fiets op het fietslandschap (8 September 2025) puts 61% of bicycles sold in 2024 as electric, counting compliant fatbikes as e-bikes and expressly excluding speed pedelecs, which are legally mopeds. We report both with their baskets and never blend them.
On trade, the only figures we can source to an exact code are for non-motorised cycles: HS 8712 exports of US$741.19m (869,048 units) in 2024 against US$690.17m in 2023, and imports of US$730.00m against US$778.06m (World Bank WITS / UN Comtrade, retrieved 7 September 2026). That basket excludes e-bikes, which sit in HS 8711, and excludes parts in HS 8714 — a serious limitation in a market where half of unit sales are electric. RVO’s Quickscan Circulair fiets (18 November 2024) describes the Netherlands as the world’s fifth-largest bicycle exporter, but that rank refers to 2022 and rests on the same non-motorised basket, so it is not an e-bike-inclusive claim.
Two things we could not establish and will not estimate: consolidated employment across bicycle manufacturing, wholesale, retail and repair — RVO’s own work notes that the relevant SBI codes contain invalid carriages and mopeds and reports employment in size classes — and 2025 production under an exact PRODCOM basket. There is also no registered geographical indication and no formally designated bicycle-manufacturing cluster in the Netherlands.
2. The brand layer is the missing layer — and the shape of the gap is unusual. In Turkey, the market this research covered before, brand-share data was simply absent from public view. Here, brand data exists in public. It just does not measure sales.
Multiscope publishes a genuine brand-share ranking with a disclosed method — 5,063 Dutch adults, representative, Gazelle 19%, Stella 13%, Sparta 11% for 2022 (Gazelle koploper Nederlandse e-bike markt, 21 March 2023). But it measures the installed stock of electric city bikes people already own, not bicycles sold in a year. Those are different markets, and in a category where the stock is 24.4 million and annual sales are 795,968, they are very different indeed.
Commercial sales-share data plainly exists and cannot be inspected. NIQ/GfK’s E-Bike Monitor 2025 advertises “Brand shares” and states that, uniquely for the Netherlands, its 2,000-respondent consumer study can be linked to actual GfK Retail sales — but its public page does not disclose what denominator those brand shares use. Independent confirmation that a sell-out panel exists comes from Gazelle’s own slides at GfK’s Insight Summit on 27 June 2024, which report “YTD 24% marktaandeel in GfK vakhandel-panel.” One brand’s share, from a dealer panel, on a conference slide. The complete ranking, its denominator and its panel boundary are not public.
There is exactly one public brand-level sales ranking in this market, and it covers a legally distinct niche: BOVAG/RAI’s speed-pedelec table in Mobiliteit in Cijfers Tweewielers 2025–2026, sourced to RDC registrations — Stromer 1,808 units in 2024, Riese & Müller 390, Gazelle 374, Klever 290, from a total of 3,188. Speed pedelecs are mopeds, registered and plated, which is precisely why brand-level counts exist for them and not for bicycles.
Everything else public measures something other than sales. Multiscope’s own awareness figures (Stella 27% top-of-mind, Gazelle 22%; aided awareness Gazelle 87%, Batavus 84%) measure recall. BOVAG’s RODI Fiets 2025 ranks manufacturers on how 444 bicycle retailers rate them across 1,622 brand evaluations — Riese & Müller 8.5, Gazelle 8.2, Trek 8.1 — which is a supplier-relationship score, not consumer demand. ANWB’s E-biketest 2026 ranks twelve individual models on lab and road testing. Consumentenbond tests models independently. RAI’s Fiets Awards are jury judgements. All are legitimate; none is a market share.
CBS cannot fill the gap either. The CPI has a bicycle heading, 07.1.3 Fietsen, which now includes e-bikes and pedelecs — and also electric scooters. But that is a price basket, not a sales statistic. The public Household Budget Survey table stops at 070000 Vervoer (€3,955 per household in 2020, the entire Transport division). The retail turnover index exposes only 476 Winkels in recreatieartikelen; CBS has a bicycle-and-moped-shop code, 47641, but ceased publishing its turnover series separately in January 2009. The industrial production index goes no finer than division 30, other transport equipment. In e-commerce, CBS’s online-purchase category is “Voertuigen of onderdelen” — bicycles, mopeds, cars and parts together — and Thuiswinkel’s is Sport & Recreatie, which the published methodology defines as sporting goods, bicycles and accessories, and camping goods.
So a Dutch bicycle brand can look up national sales, the e-bike share, average price, the stock, modal share and even how dealers rate its account management. It cannot look up its own share of the market. Until now it could not look up its AI rank either.
3. The Netherlands uses AI assistants heavily, and concentrates on one of them more than the world does. Eurostat puts Dutch generative-AI use at 44.7% of individuals aged 16–74 in the last three months, against an EU average of 32.7% (2025 survey, released 16 December 2025) — one of the higher rates in the Union. CBS’s own national statistic is 39% of adults 18+ who used AI in the previous three months (AI in de samenleving, 25 February 2026), and GWI, via DataReportal, puts 31.6% of Dutch internet users aged 16+ on ChatGPT in the past month against 26.5% worldwide (Q2 2025 wave; a Q4 2025 wave exists with a worldwide figure of 31.2%, but its Netherlands row is not publicly exposed). These measure different products, ages and recall periods. They are not competing estimates and must not be averaged.
On shopping specifically, 25% of 2,000 Dutch consumers aged 16+ said they had used ChatGPT or another AI assistant while shopping in the past year (Censuswide for Adyen, fieldwork 10 February – 12 March 2025, published 15 May 2025).
The distinctive Dutch fact is the assistant mix. On Statcounter’s August 2026 AI-referral data, ChatGPT accounts for 81.67% of Dutch AI-originated web referrals against 79.40% worldwide, while Gemini takes 8.73% in the Netherlands against 10.90% worldwide. Dutch Gemini share sat below the worldwide figure in every month we could verify from April to August 2026. This is the mirror image of the Turkish market, and it points at a different assistant.
It matters because of what that assistant does in our data. GPT-5.6 has the narrowest brand vocabulary of the five models tested: 62 distinct in-scope bicycle entities across its 223 answers, against Perplexity’s 113 and Grok’s 127, and 3.25 in-scope entities per answer against Grok’s 9.10. It is also the only model that puts Giant second, on 75 of its 223 responses. The assistant the Netherlands over-weights is the one that names the fewest brands.
On the Statcounter figures above
The gap. A market of 24.4 million bicycles where a quarter of all trips are made by bike, with national sales measured to the unit and brand shares measured only for a moped category. A population among the EU’s heaviest users of AI assistants, concentrating on the assistant with the narrowest brand vocabulary. And no public measurement of which bicycle brands those assistants actually recommend. That is what this series publishes.
What the first edition found
The September 2026 baseline produced five findings worth reading in full.
| Finding | The number |
|---|---|
| Gazelle leads the general ranking by a wide margin, named by all five assistants, and leads seven of the nine cuts. | 94.31 · named in 52.89% of 1,123 responses |
| The most-named entity in the study is not a bicycle brand. Bosch, a motor manufacturer, is named more often than any bicycle brand and is deliberately excluded from every leaderboard. | 626 of 1,123 responses · 55.74% |
| Six of the nineteen ranked entities were in or through insolvency when the questions were asked. Accell Nederland B.V. was declared bankrupt 23 days before collection. | 6 of 19 · Batavus 3rd, Koga 9th, Sparta 10th |
| Remove the origin clause and the Dutch lead nearly disappears. On the 34 questions that do not ask for a Dutch brand, Cube is named in more responses than Gazelle. | Cube 382 / Gazelle 381 of 848 · 40.0% Dutch top ten |
| Most of the tracked market is invisible. Of 222 entities identified, 19 clear the threshold in the general cut and 147 in-scope brands qualify nowhere in it. | 19 of 166 in-scope entities qualify |
Headline findings from the September 2026 baseline edition. Full method, all nine ranked cuts and every limitation are in the edition itself.
→ Read the September 2026 edition
What a high score does not mean
A high AI Visibility Score means one thing: these systems currently have information about a brand and surface it readily when asked. It is not a statement that the brand is better made, better value, more durable, more popular, better selling, larger in market share — or still trading.
That last one is not hypothetical in this market, and it is the clearest illustration this research has produced of what visibility does and does not measure.
- Quality is not measured. We record whether an entity is named and how early. We do not assess whether the recommendation was accurate, whether the description was positive, or whether the product performs. Frame quality, motor reliability, battery longevity, brake performance and service network are all real attributes, and none is in our data. ANWB and Consumentenbond both test bicycles properly; we do not.
- Sales are not measured, by us or in public by anyone. No publicly inspectable ranking of Dutch bicycle brands by annual sales share exists, as set out above. Nor does any published study estimate the incremental sales caused by a brand being named in an AI answer.
- Solvency is not measured. Accell Nederland B.V. was declared bankrupt on 13 August 2026, 23 days before this study’s collection window, leaving brand continuity unresolved for Batavus, Koga, Sparta and Carqon. Batavus still ranks third in the general cut, Koga ninth and Sparta tenth. VanMoof, ranked fifteenth, was acquired after its 2023 bankruptcy. QWIC, seventeenth, went bankrupt in November 2023 and was relaunched by Ecomotion. Stella, eighteenth, went bankrupt in November 2024 and relaunched in 2025 with a smaller store network. In the cargo segment, Babboe qualifies 32nd of 37 although Babboe cargo bikes are not currently sold in the Netherlands at all. Union, which ended new-bike production in 2025, is tracked and qualifies nowhere.
That is six of nineteen ranked entities in the general cut, and it is a finding about the assistants rather than a defect in the data. The models named these brands because the web is full of them. A bankruptcy three weeks before collection is exactly what a language model’s training data cannot contain and a web index absorbs slowly.
So when a brand ranks first here, the honest reading is narrow and specific: it is the most findable, not the most bought, and not necessarily the most available. Claims of market leadership in this category should be treated as vendor claims unless they arrive with a named research provider, a market definition, a period and a percentage.
What we measure, and how
- Five models, queried directly. Claude Haiku 4.5, Gemini 3.6 Flash, GPT-5.6 Luna Pro, Perplexity Sonar and Grok 4.3 — via API, no system prompt, provider-default temperature, no locale set. The questions are in Dutch; the models infer the market from language alone.
- Real questions, repeated, and never rewritten. Each edition uses Dutch-language questions of the kind people actually type — commuting distances, headwind, whether parts will still be available in five years — asked five times each to capture run-to-run variation. Ambiguous and ungrammatical questions are kept exactly as asked. Rewriting a question because its answers are inconvenient is selection on outcome.
- Entities discovered from the answers, then reviewed by hand. Candidates are extracted from the responses themselves, passed through a human review gate, and researched externally before any metric is computed. Origin is assigned by trademark ownership, so a brand made or sold in the Netherlands under a foreign mark stays foreign — and a brand most Dutch consumers would call Dutch can be classified foreign if its mark has moved.
- Component brands are tracked but not ranked. Motors, drivetrains, brakes and accessories are flagged
Adjacentand reported separately. This matters more here than in most verticals: Bosch is named more often than any bicycle brand in the study, and ranking it against bicycle manufacturers would have turned the leaderboard into a table about drivetrains. - A three-component score. AI Visibility Score = 0.45 × Mention + 0.30 × Position + 0.25 × Breadth — how often an entity is named, how early it appears, and how many of the five assistants know it.
- Nine cuts, each on its own published scale. Every edition is read across three product segments, origin scope (questions that demand Dutch brands versus questions that do not), and behavioural types (discovery / attribute / use-case). Each cut has its own denominator and its own scale, fixed once and never recomputed. Dutch-versus-foreign comparisons are made only on origin-neutral questions — the only fair basis.
What we do not measure. Quality, accuracy, sentiment, price, durability, safety or solvency. A score is not an endorsement.
Editions
| Edition | Published | Scale | Headline |
|---|---|---|---|
| September 2026 — baseline | 7 September 2026 | 1,123 responses analyzed · 45 questions · 5 models · 222 entities tracked | Gazelle leads at 94.31 and tops seven of nine cuts, but Bosch is named more often than any bicycle brand, six of the nineteen ranked entities were in or through insolvency, and on origin-neutral questions Cube is named in more responses than Gazelle. |
Published editions. The study is repeated quarterly; the second edition is due in December 2026, and trend analysis becomes possible from that point. Until then there is nothing in this series to compare against, and this hub makes no trend claims.
The wider context: is AI discovery real yet?
We think the honest answer is “real, measurable, growing from a small base — and smaller than the vendors say.”
The Dutch evidence specifically is thin, and we would rather say so than borrow a number. The best category-adjacent figures we can source are domain-level: Similarweb estimates 18.1 thousand visits from generative-AI platforms to coolblue.nl in February 2026, amounting to 0.25% of that site’s referral traffic, and 15.5 thousand to ah.nl at 0.11%. Those are single-domain vendor estimates of traffic rather than people, they are not bicycle retailers, and Similarweb sells AI-visibility products. We found no aggregate figure for the AI share of Dutch retail-site traffic, and no Dutch study of AI-assisted shopping for bicycles or durable goods at all.
What that leaves is a genuine asymmetry, and it is the reason a baseline is worth publishing now. Adoption of the assistants is measurably high — 44.7% of Dutch 16-to-74-year-olds used generative AI in a three-month window, well above the EU average, and a quarter of Dutch consumers say they have used one while shopping. The commercial consequence of that adoption is not yet measurable in this category by anyone. The measurement is most valuable before the behaviour becomes mainstream. If AI-assisted bicycle shopping in the Netherlands grows the way general AI adoption has, the brands visible when it happens will not be the ones that start optimising afterwards.
We do not claim to explain why a model names a brand. There is no established public model of how commercial assistants decide which brands to surface, and we are not going to invent one. This series does the one thing that can be done rigorously: measure the output. We record which entities are named, how often, and how early — and we publish the method and the data pack so anyone can check us.
Data, caveats and sources
- Point-in-time. Each edition is a snapshot from a single collection window — 5 September 2026 for the baseline. Model versions and web indexes shift; trend claims become possible only from the second edition onward.
- Visibility is not quality, sales, market share or solvency. Repeated because it matters, and because in this market the third cannot be checked against public data and the fourth changed three weeks before collection.
- Market figures in this hub carry different confidence levels. Modal share, theft, price, classification, tax and regulatory figures are official (CBS, KiM, Eurostat, Belastingdienst, RVO, Rijksoverheid). Sales, stock and dealer-satisfaction figures are industry-association research (BOVAG/RAI, executed by NielsenIQ or GfK) and are labelled as such. AI referral and traffic statistics are commercial or vendor-published, and where the vendor sells AI-visibility products we say so.
- Different bases are not interchangeable. The 49% e-bike unit share is BOVAG/RAI’s market category, which treats fatbikes separately and publishes no technical definition; the 61% figure is the Ministry’s, which counts compliant fatbikes as e-bikes and excludes speed pedelecs. The HS 8712 trade figures cover non-motorised cycles only and therefore miss roughly half the market by units. The “fifth-largest exporter” rank rests on that same non-motorised basket and refers to 2022. We name the basket every time.
- Some things are simply not public, and we say so. There is no consolidated employment figure for the Dutch bicycle sector under a coherent SBI basket; no CBS bicycle-only retail turnover series since January 2009; no bicycle-only household spending figure; no bicycle-only e-commerce statistic; and no national figure for the second-hand or refurbished bicycle market. We flag these gaps rather than filling them with estimates.
- The two shop counts are not the same number. 2,980 establishments is the KvK Handelsregister count under SBI 47641, which also contains mopeds, mobility scooters and brommobielen and excludes secondary-activity locations (Marktdata.nl, 3 September 2025). 2,528 is the narrower Locatus count of physical bicycle shops at end-2024 (via WielerFlits, 6 February 2025). They measure different things and neither is wrong.
- The bicycle stock figure is an industry estimate. 24.4 million is BOVAG/RAI’s number, not a registration count, and the accessible source does not disclose how it was estimated. KiM’s 1.3 bicycles per inhabitant is internally inconsistent about its own reference year, so we quote it without attaching one.
- The 2024 mobility figures must be sourced to the correction. CBS identified a method break in ODiN 2024 and republished corrected results on 31 March 2026. The 27.7% and 9.2% shares above are calculated from those corrected tables, not from the original release, and not printed as percentages by CBS.
- Regulation in this market is moving, and proposals are not law. As of September 2026 a compliant fatbike is legally an electric bicycle with no minimum age, no helmet requirement and no registration. The Cabinet announced an intention to develop a fatbike minimum age on 24 April 2026 without specifying a level, and a helmet requirement for under-18s on e-bikes and fatbikes is in preparation with a target of 2027. Neither is in force. Speed pedelecs are mopeds: minimum age 16, AM licence, approved helmet, insurance and a plate.
- Theft context, with its own caveat. Police registered 86,950 bicycle thefts in 2025 (provisional). The Veiligheidsmonitor 2025 found 4.9% of people aged 15+ were victims and that only 38.7% of experienced incidents resulted in a formal report — so 61.3% were not. Multiplying one by the other would mix different universes and periods, and we have not done it.
Want the underlying data?
Frequently asked questions
Gazelle leads the general ranking at 94.31, named in 52.89% of all 1,123 analyzed responses by all five assistants. Cube (73.04) is second, then Batavus (69.27), Urban Arrow (61.06) and Trek (58.10). Leadership changes by product segment: Gazelle leads everyday and commuting questions, Cube leads sport and recreational, and Urban Arrow leads cargo, family and special formats.
Not by sales. Multiscope publishes a genuine brand-share ranking with a disclosed method, but it measures the installed stock of electric city bikes people already own rather than annual sales. NIQ/GfK advertises brand shares and demonstrably operates a Dutch dealer sell-out panel, but the ranking, denominator and panel boundary are not public. The one public brand-level sales table covers speed pedelecs, which are registered mopeds. For ordinary bicycles and e-bikes, no inspectable annual sales-share ranking was found.
Bosch is a motor manufacturer, classified Adjacent. It appears in 626 of 1,123 responses, more than any bicycle brand, and is reported separately rather than ranked. Scoring every Dutch bicycle manufacturer against a German component supplier would answer a question nobody asked. Shimano, at 460 responses, is in the same position.
Substantially, once you remove the origin clause. On the 34 questions that do not ask for a Dutch brand, 7 of 20 qualifying entities are Dutch and Dutch brands take 40.0% of the top ten — and Cube is named in 382 of those 848 responses against Gazelle's 381. On the 11 questions that explicitly ask for a Dutch brand, 15 of 19 qualifiers are Dutch, but that is what those questions were built to return and cannot be read as evidence of visibility.
Not quickly. Accell Nederland B.V. was declared bankrupt on 13 August 2026, 23 days before this study's collection window, leaving brand continuity unresolved for Batavus, Koga and Sparta — which rank third, ninth and tenth. Babboe qualifies in the cargo segment although its cargo bikes are not currently sold in the Netherlands. Six of the nineteen ranked entities were in or through insolvency at collection. Visibility is not viability.
No. Claude, Perplexity and Grok grounded 100% of their answers and GPT-5.6 grounded 86.2%, but Gemini grounded only 41.8%, against an overall rate of 85.6%. For its ungrounded answers Gemini is describing what its training data recorded rather than what the Dutch web says now — which is exactly the condition under which a recent bankruptcy or product withdrawal would be missed.
No. The score measures only how often and how prominently an entity is named across the assistants. It does not assess quality, durability, battery range, service network, price, accuracy or sentiment, and it does not establish that a brand is still trading. A score is a measure of visibility, not an endorsement.
About Herm.io & disclosure
Herm.io is a consumer behaviour and marketing data company. We study how people discover and choose brands to help businesses reach the right customers. This report forms part of our public research and is conducted quarterly.
Disclosure & neutrality
The rankings and metrics in this report are entirely impartial and are based on the methodology described within the report. No brand can pay or provide sponsorship to be included in the report, improve its ranking or influence how it is described in the content.
A brand’s score in the report is not an endorsement or assessment of quality; it is solely a measure of its current visibility within AI models. No service or product offered by Herm.io is sold for the purpose of directly changing the results of this report or guaranteeing a ranking. To preserve objectivity, the publisher’s own domain is checked for in the cited-source data of every edition. In the September 2026 edition it does not appear anywhere in that data, so no exclusion was applied and no citations were removed. This ensures that the company does not appear in, measure itself through or benefit from its own study.
Brands seeking to understand their position in the data can schedule a consultation for an impartial assessment of the findings. The consultation is advisory and free of charge.
Quarterly series · Baseline edition published 7 September 2026
Mert Can Elkaya
Contributor
I'm a product builder working at the intersection of product, fintech, and growth. From martech and venture capital to leading product at a proptech platform and co-founding a fintech startup, I help teams—and shoppers—make smarter, more confident decisions.
More from Mert