AI sports shopping has found a familiar new place to work: the page where a fan is choosing a team shirt. Fanatics and OpenAI announced on 26 October 2026 that an AI-powered version of Shop the Look is now live across many of the retailer’s sites. It proposes coordinated pieces of merchandise rather than leaving the shopper to assemble an outfit by searching through separate product categories.
Imagine choosing a jersey and being offered a suitable cap or jacket beside it. The suggestions are intended to reflect the team, the style and what is currently in stock. The technology is being used on Fanatics.com and online stores associated with major US sports leagues. This is not a robot personal shopper wandering the internet on a customer’s behalf. It is an additional recommendation system inside a retailer’s own checkout journey.
How AI sports shopping works on a product page
OpenAI says its models use enriched product records, team information and stock data to assemble possible combinations. Fanatics applies merchandising rules to keep the recommendations compatible with the original item. Looks are refreshed as the catalogue changes. The system is therefore a mixture of AI-generated matching and ordinary retail constraints, not unconstrained fashion invention.
That combination is significant because sports merchandise carries restrictions a generic image-matching system might miss. A shopper choosing a particular club or team would reasonably expect accessories with the right identity. Size, availability and the seller’s product range matter as much as colour coordination. Retail rules provide boundaries around suggestions that might otherwise look stylish but be unsuitable or impossible to purchase.
In its own announcement, Fanatics says the first 50 product looks were tested in less than a day. The company then reviewed the output and helped refine the service before extending it more widely. Human merchandising decisions remain part of the process, even though the system helps create combinations much faster than a manual team might manage across a large catalogue.
What a 10% engagement increase does and does not mean
The companies report that an experiment covering roughly 30,000 Major League Baseball products produced 10% higher engagement than the previous Shop the Look experience. Engagement could reflect shoppers interacting more with suggested products. It is not, by itself, evidence of a 10% rise in sales, lower prices or better purchasing decisions.
There is also no independent consumer test in the announcement showing that the recommended combinations are more attractive than ones chosen by people. The comparison is a company-reported test against an earlier version of its own feature. It supports a claim about measured interaction in that setting, not a universal claim that AI has become a better stylist than a human.
Retailers have long offered related products and customers have long wondered whether those suggestions were helpful. The question of trust in AI shopping advice becomes sharper when the adviser is also the shop. An attractive bundle might help someone find the right item. It might equally encourage an extra purchase that the customer had not planned.
Fashion recommendations are also unusually difficult to judge from clicks alone. A suggestion may be opened because a customer wants to inspect the price, because the product is amusing, or because the shopper hopes to buy it. Those interactions have different commercial meanings. Without a detailed public description of the test design and longer-term purchasing outcomes, readers should regard the reported uplift as an early product metric rather than a full assessment of consumer benefit.
Another question is how well coordinated looks handle products that are similar but not interchangeable. A cap for one team may share the colours of another, yet a genuine supporter is likely to care more about the badge than the palette. Human merchandising rules may catch these mistakes, but their quality is part of the service. The technology must be judged on accurate pairing and practical product information, not simply on whether the suggestions look plausible in a photograph.
A question for Fanatics will be how it handles the difference between adding convenience and creating pressure. A fan who has already chosen an expensive jersey may appreciate being shown a cheaper accessory instead of another premium item. The companies have described engagement with the new feature, but not a public test of whether its recommendations reflect price sensitivity or avoid pushing shoppers towards unnecessary purchases. That is a useful standard by which customers can judge the experience themselves.
Why a recommendation is not independent advice
Unlike a general shopping comparison service, this system recommends from a retailer’s own inventory. That may be entirely appropriate for completing an outfit, but buyers should keep the distinction clear. The retailer controls the available choices, the rules of presentation and the commercial setting. Nothing in the announcement demonstrates that the proposed outfit offers the best price across competing sellers.
A practical shopper can treat a suggestion as a shortcut, then check sizing, returns, delivery charges and whether the accessory is actually needed. These checks matter more than the sophistication of the model making the recommendation. A matching product is useful only if it fits the person and the budget.
The experiment also sits beside wider concerns about how online shopping systems use customer information. Fanatics describes using product, team and stock data for matching, but its announcement does not establish that every shopper receives identical recommendations or explain every detail of the data used for personalisation. Those are questions a customer may reasonably ask as similar systems spread.
For retailers, the attraction is straightforward: an enormous catalogue becomes easier to browse and potentially easier to sell. For shoppers, the best outcome is a relevant suggestion that saves searching. The test of the feature will be whether it stays useful after the novelty wears off, not simply whether people click one more product.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.