OpenAI Partners with Fanatics: Rebuilding the Fan Shopping Experience with AI

From "Watching the Game" to "Wearing the Look": How AI Shortens the Fan's Decision Path

Sports e-commerce has long faced an awkward problem: the urge to buy that strikes fans in the moment of watching a game often fades because they can't find "the exact item in the picture." The partnership between Fanatics and OpenAI targets exactly this pain point—the Shop the Look feature the two launched lets fans, while browsing game footage or product content, directly get recommendations for complete outfits along with purchase entry points. According to official disclosures, in tests covering about 30,000 MLB products, the feature delivered a 10% increase in engagement rate. That number is not trivial in the e-commerce conversion chain—it means AI is dramatically reducing the friction between "recognizing interest" and "completing a purchase."

What Problem Does Shop the Look Actually Solve

The traditional jersey-buying process is search-based: fans need to know which team, which player, which size they want, and then actively search for it. But real shopping scenarios are often the opposite—users are moved by an overall outfit in a picture or a scene, yet can't articulate the specific product names.

Shop the Look's logic is image-driven shopping: the system identifies clothing elements in the frame and combines team jerseys, hats, accessories, and more into a list that can be added directly to the cart. For a platform like Fanatics with a massive catalog of licensed products, this is equivalent to turning the product library from a "search index" into a "visual shelf." The test covered about 30,000 MLB products, showing that the capability has been validated at real category scale rather than as a lab demo.

Multimodal visual recognition refers to a model's ability to understand both image and text information simultaneously. In e-commerce scenarios, the system first needs to detect clothing regions from game footage or outfit images, then perform fine-grained classification of jerseys, hats, and accessories, and finally map visual features to specific SKUs in the product library. This process typically combines object detection with image embedding retrieval: the former locates "what's being worn," while the latter finds the most similar products in vector space. The Fanatics–OpenAI collaboration leverages this kind of multimodal capability to transform unstructured visual content into structured product lists, enabling "image-driven shopping."

The Signal Behind the 10% Engagement Lift

A 10% engagement increase needs to be understood in the e-commerce context. Engagement rate usually refers to users' clicks, dwell time, or add-to-cart behavior on recommended content, and it sits upstream in the conversion funnel. Improvements in upstream metrics tend to cascade downstream—more clicks mean more product detail page visits, which in turn brings more transaction opportunities.

It's worth noting that the official wording is "engagement" rather than "sales," so it can't be directly equated with revenue growth. But the result at least proves two things: first, fans are highly receptive to AI-generated outfit recommendations; second, visual recognition has crossed the usability threshold in accuracy and practicality for highly structured categories like sports licensed merchandise.

Engagement rate is a broad upstream metric in e-commerce analytics, typically including click-through rate on recommendation modules, page dwell time, add-to-cart rate, and other behavioral signals. It differs from downstream metrics like conversion rate and average order value: engagement rate measures whether users "were attracted by the content," not whether they "paid." Therefore, a 10% lift shows that AI recommendations effectively reduced users' indifference to outfit content, but there are still several steps before confirming revenue contribution. In industry practice, the "click—detail page view—add to cart—payment" funnel model is commonly used to track whether upstream improvements cascade smoothly to transactions, which is also why the original article emphasizes the need to observe return rates and repurchase rates.

Why Sports E-commerce Is an Ideal Testing Ground for AI Shopping

Sports products have several inherent attributes that make them particularly well suited for visual recommendation technology:

• Clear product structure: Teams, players, positions, and colorways all have explicit tags, making it easy for models to establish mapping relationships.
• Highly concentrated scenarios: Game days, playoffs, transfer windows, and other moments create clear purchase motivations.
• Fan emotion-driven: Fans have a strong desire for "the same look," lowering the persuasion cost of recommendations.

As a major player in sports licensed merchandise, Fanatics' catalog covers multiple leagues and teams, providing ample training and testing samples for the model. OpenAI provides the underlying multimodal capabilities, while Fanatics provides product data and consumption scenarios—their division of labor is neatly complementary. According to the original information, the focus of this collaboration is "simplifying the fan experience" rather than simply stacking technology.

SKU (Stock Keeping Unit) is the smallest granularity of e-commerce product management; each combination of team, player, size, and colorway for a jersey corresponds to an independent SKU. Although sports licensed merchandise has a huge number of SKUs, its attribute dimensions are highly regular—league, team, player name, number, home/away colorways all have standard fields, making it easy for visual models to establish a stable mapping between "image features and product attributes." By contrast, fast fashion or home goods categories have vague style descriptions and attributes that are hard to enumerate, making visual recommendation harder to implement. As a result, the sports category is often seen as a "structured testing ground" for gauging the maturity of visual shopping technology.

Three Takeaways for the Industry

First, the value of an AI shopping assistant lies in shortening the path, not adding features. Shop the Look doesn't create a new shopping flow; it simply compresses "see—search—filter—order" into "see—order." Reducing steps is itself a friend to conversion rates.

Second, test scale determines credibility. A test volume of about 30,000 products is far more convincing than a small-sample demo. For e-commerce platforms planning to adopt similar capabilities, they should check whether vendors have validated results at real category scale.

Third, engagement rate is only the starting point. More worth watching going forward are metrics like return rate, average order value, and repurchase rate—whether AI recommendations can be "accurate and appropriate" ultimately depends on these downstream data.

Conclusion

The Fanatics–OpenAI collaboration presents a pragmatic example of AI implementation: it doesn't pursue a disruptive narrative but instead optimizes around a specific friction point and lets quantifiable test results speak. For sports e-commerce, Shop the Look may be just the beginning—as visual recognition, outfit recommendations, and inventory systems become more integrated, the distance for fans from "falling in love" to "wearing it" will continue to shrink.

One-line Summary

Fanatics and OpenAI use AI visual recognition to enable image-driven shopping, with tests showing a 10% engagement lift.

Paragraph Summary

Fanatics and OpenAI have partnered to launch Shop the Look, a feature that uses multimodal visual recognition to turn clothing elements in game footage or outfit images into product lists that can be added directly to the cart, transforming traditional "search-based shopping" into "image-driven shopping." In tests covering about 30,000 MLB products, the feature delivered a 10% increase in engagement rate, proving that fans are highly receptive to AI outfit recommendations and that visual recognition has crossed the usability threshold in structured categories like sports licensed merchandise. The article notes that sports products—with clear structure, concentrated scenarios, and strong emotional drivers—are an ideal setting for visual recommendation technology, while cautioning that engagement rate is only an upstream signal; real business value still needs to be assessed through downstream metrics like return rate, average order value, and repurchase rate.