Fanatics and OpenAI Launch AI Recommendation Feature, Boosting Product Page Views by 10%

According to foreign media reports, global sports merchandise retail platform Fanatics has officially launched a shopping recommendation feature called 'Shop the Look,' developed entirely using OpenAI's Codex model capabilities. Its core purpose is to reduce the cost for users selecting from a vast product pool.

Currently, Fanatics offers over 1.5 million standalone SKUs, covering jerseys, headwear, team-themed streetwear, socks, and other full-category sports merchandise, with partnership agreements with nearly all major sports leagues and clubs worldwide. Its Fanatics Commerce division produces over 45 million items annually, selling to more than 180 countries and regions globally. Fanatics founder and CEO Michael Rubin disclosed earlier this month that the company's commerce division is expected to surpass $7 billion in revenue this year, with the collectibles division, including the Topps trading card business, generating approximately $5.5 billion. Team merchandise remains the company's largest revenue source.

As product scale continues to expand, platform operations have encountered clear bottlenecks. Justin Tsai, Chief Product Officer of Fanatics Commerce, discussed existing pain points during a product launch communication: the average single shopping session on the platform now exceeds 30 minutes, and with over a thousand new products added weekly, the time cost for users to filter products continues to rise. Traditional manual curation relies on offline photography and stylist pairing, with long preparation cycles that cannot accommodate products requiring rapid listing, such as championship commemorative shirts and hats. Completely random recommendation logic is also unsuitable for sports fan groups, where matching products for the same team and player is a core demand—fans of the Los Angeles Dodgers would not want to see recommendations for San Francisco Giants merchandise.

The implemented 'Shop the Look' feature is embedded in individual product detail pages, automatically matching in-stock items from the same series based on the product's team and style positioning, combining them into complete outfit suggestions. During operation, the Codex model automatically parses platform-wide product and catalog data, filters suitable items, and dynamically updates recommendations. When a user views a jersey, the page simultaneously presents matching hats, socks, and other accessories.

Initial operating data shows that since launch, page views for recommended products have increased by 10%, and the outfit displays help users more intuitively judge the compatibility between intended items and existing team merchandise, reinforcing purchase decisions.

This is also OpenAI's latest implementation in the retail shopping scenario. Previously, OpenAI tested a full-chain shopping feature within ChatGPT, launching a service last fall that allowed users to purchase merchant products directly in the chat interface. That instant checkout feature has now ceased testing, and the team has shifted to direct collaboration with retailers, building brand-specific app entry points within ChatGPT.

Rosemary DeAragon, OpenAI's Industry GTM Lead, shared her perspective on retail recommendation scenarios during partnership discussions: personalized matching in retail has long been an industry challenge. Past e-commerce recommendation systems built on user purchase history, visit geography, and account information still suffer from matching biases. AI technology can achieve more precise deterministic matching, pushing personalized recommendation experiences to a new stage. In future shopping scenarios, manual curation and in-store try-ons will still hold a stable share, but in scenarios of narrowing choices from vast product pools and helping users complete high-value product pairings, AI technology has clear room for experience optimization.

This article was first published on the Ebrun official website.