Centium AI partnered with Long Run Labs on its largest study yet, 2,200 AI prompts across five models and 22 running categories, on how AI recommends running brands.
Centium AI partnered with Long Run Labs, the endurance-sports podcast and newsletter from operator Jonathan Levitt, to run the largest study the platform has produced and publish it as a joint report on how AI recommends running brands. The two ran 2,200 unaided prompts across the five major AI models, ChatGPT, Claude, Gemini, Perplexity, and Grok, covering 22 running categories, to measure which brands the models recommend on their own.
Levitt and Centium AI Founder Michael Rueckert published their findings as "How AI Recommends Running Brands in 2026: The Largest Study Yet," a written report on Long Run Labs paired with a 90-minute podcast conversation, and the two plan to repeat the analysis roughly every six months. Centium AI prompts at the category level rather than naming brands, because AI models anchor to any brand mentioned in a question. To keep the results unbiased, the prompts went through the model APIs with no chat history and no location, asking open questions like the best running shoes or the best running headphones, then counting how often each brand surfaced. The run produced 16,990 citations from 2,558 domains and named 2,052 distinct brands.
The picture that came back was concentrated. Nine brands accounted for 22 percent of all mentions study-wide, and 56 percent of the brands that surfaced appeared exactly once. The average AI answer named just 6.5 brands, a narrow window that separates the names models reach for from the ones they rarely surface. Study-wide, Nike led at 23 percent, followed by Brooks at 20 percent, HOKA at 19 percent, and Adidas and Saucony tied at 16 percent.
The study's central finding, as Levitt framed it, is that authority does not transfer across categories. Winning best running shoe does not win best running headphones. HOKA led trail and gravel shoes, Brooks dominated road shoes and sports bras, Salomon owned hydration, and Science in Sport led energy gels at 66 percent. In running headphones, the most-cited source was not a running publication at all but a specialist ratings site that appeared in 81 percent of answers. The model cannot try on the shoes or taste the gels, so for each category it goes to that category's specialists. The takeaway for smaller brands is that beating Nike is not the assignment. Owning a category is.
Where AI models looked mattered as much as who they named. Fewer than 5 percent of citations went to brand-owned websites. The rest went to third parties: media outlets, Reddit, YouTube, and independent gear reviewers. Runner's World was cited in 40.5 percent of answers, with iRunFar and OutdoorGearLab close behind in their lanes. "Best of" roundups made up 53 percent of all citations, and the models favored fresh content, with a median cited page 4.1 months old and 48 percent of citations coming from work published within the prior six months. Rueckert's read is that AI evaluates a brand the way a researcher would, not the way a shopper does.
I like to describe AI as an evidence-based thinker and an evidence-based decision maker. It's not making emotional decisions like humans might. We've been in this world of trying to inspire customers, think like Nike's just do it campaigns, where you're trying to show your product and inspire somebody, make them feel something. AI is not like that. AI is logical. AI just wants the data, it wants the information.
Emotional brand campaigns still move people, but the models are pulling specs, rankings, and third-party reviews to assemble an answer, which is why ranked lists in trusted publications carry so much weight. The pattern in the running data matches what Centium AI sees across the more than 200 brands it tracks. The brands gaining AI visibility fastest are the ones good at getting other people to talk about them, through PR, reviews, and a wide footprint of content, rather than the ones optimizing their own websites in isolation. Levitt's read in the report was blunt: living on your own site is not a strategy, because the models are reading a dozen magazines, and a brand absent from those magazines is absent from the answer.
Pick your category, get into the roundups, stay fresh, and know that it takes more content than you realize to move AI models meaningfully.
For endurance and outdoor brands, the study is a map of where the AI answer is decided. It names the categories still open to own, the publications the models lean on, and the content formats that earn citations, and Centium AI and Long Run Labs plan to re-run it every six months to track how those patterns move. Centium AI measures how the five major AI models recommend brands and surfaces the exact sources and signals each one weighs most, so a brand can see where it stands and what to do about it. Built by an outdoor marketer, Centium AI tracks more than 150 outdoor brands across running, cycling, hydration, and tourism, and is the outdoor industry's go-to platform for AI visibility. See how it works at Centium for Outdoor.
Long Run Labs, the endurance-sports newsletter from Jonathan Levitt, features Centium AI in an essay arguing AI chat replies are the new top of the marketing funnel.
Second Nature Podcast features Centium AI Founder Michael Rueckert on why PR now outperforms paid search for AI-driven brand discovery in the outdoor industry.
The ROI Masterclass Podcast features Centium AI Founder Michael Rueckert on AI visibility for hospitality and tourism brands in the AI era.
Pick a plan and see your first dashboard today, or try our free tools to explore the platform in action. You’re just a few clicks away from building your customized methodology.
