When AI recommends a bike, what does it choose? Our research on the Italian cycling market seen through AI
A growing number of people, before buying a bike, a helmet or a piece of kit, no longer open Google alone: they ask ChatGPT, Gemini, Google's AI. And the AI answers with names. But where do those names come from? Who gets recommended, who stays invisible, and which sources sit behind the answer? We asked ourselves the question, and rather than guess we measured it. That is how "The Italian cycling market seen through AI" was born, the research we carried out together with GeoSnap. Nearly 600 AI responses, 8 product categories and more than 10,406 brand mentions put under the microscope, to understand how artificial intelligence now steers whoever is about to buy a bike in Italy.
Why it matters who the AI names
When an AI recommends a product, it builds the answer from what it finds written and cited online, and hands it back as a firm suggestion, often with no alternatives. That recommendation grows from how, and how much, a brand is written about and cited online, and no advertising budget can stand in for it. For a brand this changes the rules. Being named by a model means entering the shortlist of someone about to buy; being left out means disappearing from a slice of the market that grows every month, before the prospective customer even reaches a website or a shop. Visibility inside generative engines is becoming a marketing asset in its own right, and like any asset it has to be measured first.
How we ran the research
We queried three AI assistants in Italian – ChatGPT, Gemini and Google AI Mode – through their public interfaces rather than their APIs, because the answers a real person gets differ from the ones the APIs return. Each question was asked three times, to reduce model variability, and every prompt is classified by category, sub-category, intent and budget range, the way a cyclist about to buy would frame it. That produced nearly 600 responses, analysed across eight product categories, from which we extracted more than 10,406 distinct brand and product mentions: the raw material behind the rankings.
What the report reveals
The full 55-page report sets out, for each of the eight categories, who gets named and how much, measuring mentions, share of voice, question reach, average position in the answers and consistency across repeated queries. Beyond the rankings, it tackles the questions that really matter to a brand:
- which brands the AI puts at the top, and how often depending on the model;
- how much Italian brands are recommended compared with international competitors;
- where the three models contradict one another, suggesting different products to the same question;
- which sources the AI draws on for what it says: publications, e-commerce or official brand sites;
- how the recommendations shift with sub-category, price and type of use.
Together, this data tells a simple and uncomfortable truth: AI recommendations are anything but neutral, and understanding the logic behind them is the first step to not being at their mercy.
Download the full research
"The Italian cycling market seen through AI" is available in full. You can request the complete report here.
It is the work we do every day: how a cycling brand is told, and now also how it gets named by the new discovery interfaces. If you want to understand what AI says about your brand, we have written about it here too.
Vitesse is an Italian Media PR and communications agency specialised in cycling, sport, outdoor and active tourism. We have worked since 1991, alongside events, brands and destinations. Explore our services or get in touch. Coffee's on us.