2
5 Comments

I tested 150 buyer prompts across ChatGPT, Gemini & Claude for 57 triathlon brands. The AI bias is wild

Nobody clicks through ten Google results anymore.

If someone is training for an Ironman and wants new shoes, a bike computer, or race nutrition, there’s a good chance they’ll just ask ChatGPT, Gemini, or Claude what to buy.

I wanted to see which triathlon brands actually show up in those answers, so I tested 150 buyer-style prompts across the three platforms and tracked mentions for 57 brands.

A few things surprised me.

Factor Bikes barely appeared in ChatGPT and Claude, but Gemini mentioned it 14 times. My guess is Gemini was picking up more recent wind-tunnel testing and reviews.

Saucony was another interesting one. Overall, it performs very well in shoe recommendations. But as soon as the prompt becomes specific - for example, asking about stability when your legs are tired late in a marathon - ASICS and HOKA dominate.

There were also categories where one brand seemed to own the conversation. Garmin accounted for 25.5% of the tech recommendations I tracked, while Maurten had 29.7% of nutrition mentions.

The takeaway for me is that being visible in AI answers is becoming a different game from traditional SEO. Reviews, comparison articles, structured product information, and niche sites seem to matter a lot.

If you work for an endurance or triathlon brand, leave the brand name below. I’m happy to check how often it appeared in the dataset.

on August 23, 2026
  1. 2

    The interesting part is showing that AI visibility isn't uniform across models or even buyer intent. A brand can dominate a broad category while disappearing when the prompt becomes more specific.

    1. 1

      Exactly this. AI search doesn't just rank brands. it filters them based on micro-intents and source bias.

      Gemini leans heavily into recent lab/wind-tunnel data, while ChatGPT leans into historical brand equity. That split alone creates completely different winner lists for the exact same query across platforms.

      If you have a specific brand in mind, drop it here and I can pull its specific intent-breakdown from the data :)

      1. 1

        That model-to-model difference is interesting. I’d be interested in continuing the conversation — would you be open to sharing the best email to reach you on?

        1. 1

          Thank you, you can drop me a message on rafal.arte@gmail.com

          1. 1

            Thanks! I’ve just sent it over.

            Looking forward to hearing your thoughts whenever you have a chance.