Live shopping, the internet’s reinvention of television home shopping, has demonstrated immense potential, yet its growth in North America has lagged significantly behind its Asian counterparts. While China’s live shopping market generated an estimated $900 billion in sales in 2025, a figure nearly equivalent to the entire U.S. eCommerce market, 68% of North American consumers report never having purchased anything through social media. This stark disparity is not attributed to content quality but rather to a fundamental challenge: discovery. Addressing this, Whatnot, a live-shopping marketplace valued at over $11 billion, has made a strategic move, acquiring machine learning startup Shaped to infuse real-time artificial intelligence into its platform, aiming to bridge the gap between content and conversion.
The Discovery Dilemma in Live Commerce
The core premise of live shopping is compelling: a host demonstrates a product live, engages with viewers through comments, offers limited-time deals, and enables purchases with a single tap, all within the stream. This format dramatically shortens the path from product discovery to transaction. However, the sheer volume and dynamic nature of live commerce present a unique hurdle. Thousands of live streams can run simultaneously, inventory changes by the minute, and auctions conclude in seconds. A viewer joining a stream even slightly late might miss the very product that would have prompted a purchase, leading to a suboptimal experience and potential disengagement.
Traditional, static recommendation systems are ill-equipped to handle this rapid flux. Matching the right buyer to the right live stream at the precise moment requires a level of personalization and speed that conventional algorithms cannot deliver. This is precisely the problem Whatnot, which processes more than 500,000 hours of live video and millions of user interactions each week, is determined to solve.
Whatnot’s AI-Powered Acquisition of Shaped
In a significant development reported by TechCrunch on Wednesday (July 15), Whatnot acquired Shaped, a machine learning startup specializing in real-time recommendation and search infrastructure. While financial terms of the acquisition were not disclosed, the move signals Whatnot’s deep commitment to leveraging advanced AI. Shaped’s founder and CEO, Tullie Murrell, who previously developed recommendation systems at Meta, will join Whatnot along with nearly a dozen engineers and researchers. This team will spearhead a new applied AI research group within the live-shopping giant.
Emmanuel Fuentes, Vice President of Data and AI at Whatnot, emphasized the transformative potential of this integration. “By combining Shaped’s technology with Whatnot’s existing systems, we can make recommendations faster, more responsive and more personalized,” Fuentes stated. Shaped’s proprietary systems are designed to integrate customer data with sophisticated AI models and machine learning, enabling highly personalized search and discovery functionalities. This synergy is expected to empower Whatnot to suggest the most relevant live streams to users, dynamically adapting based on their past purchases and real-time behavioral signals within a session.
AI Personalization: The Key to Mainstream Adoption
The challenge for Shaped is inherently specific to live commerce. Unlike stable eCommerce product catalogs, Whatnot’s inventory is in constant flux, and a buyer’s interests can shift rapidly even during a single broadcast. Whatnot’s own growth trajectory underscores the increasing complexity of this discovery problem; the platform launched more than 35 new product categories in 2025 and over 45 in the first half of 2026, with sellers collectively surpassing 1 billion orders. Each new category, while expanding choice, simultaneously amplifies the difficulty of connecting shoppers with desired items efficiently.
The strategic integration of real-time AI is designed to surface the ideal stream for the right buyer at the opportune moment, continuously updating as user behavior evolves within the session. This capability is critical for preventing shopper frustration and ensuring repeat engagement. A shopper who consistently lands on irrelevant streams or misses desired products is unlikely to return.
Building Consumer Trust in AI-Mediated Commerce
The broader consumer landscape indicates a growing readiness for AI-driven shopping experiences. According to the PYMNTS Intelligence report, “The Agentic Commerce Deep Dive: Payment Infrastructure and the Path to Trust for Merchants,” a collaboration with Visa Acceptance Solutions, nearly half (48%) of online shoppers utilized AI to research their most recent purchase. Furthermore, AI-referred traffic to U.S. retail sites experienced a remarkable 393% year-over-year growth in the first quarter of 2026, signaling a robust and accelerating consumer appetite for AI-mediated interactions in commerce.
By acquiring Shaped, Whatnot is making a significant bet that solving the intricate discovery problem through advanced AI is the pivotal step required to transition live commerce from a niche format into a widely adopted, mainstream retail channel. This strategic investment underscores the belief that intelligent, real-time personalization is not just an enhancement but a necessity for the future of interactive online shopping.
