The RealReal's AI Overhaul Helps Drive 22% GMV Growth

The RealReal has spent the past year rebuilding its core operations around artificial intelligence, and growth has been accelerating alongside it. The company’s intake system, Athena, now automates authentication and cataloging steps that once required manual attribution, and a new pricing algorithm evaluates more than 100 data points per item to set and adjust prices without human override. Both systems exist to solve the same problem. CEO Rati Levesque in the second quarter earnings call described the shift as compounding rather than incremental. Every part of the platform, she said, from authentication to pricing to member experience, is making the others stronger. Pricing sits at the center of The RealReal’s AI investment. The company’s pricing algorithm now evaluates more than 100 data points per item, including page views and how often shoppers add an item to their favorites, to set an initial price and manage markdowns over time. That precision is showing up in unit economics. Sales of items above $1,000 rose 36% in the first half of the year, a mix shift CFO Ajay Gopal said is driving take rate down even as profit dollars per transaction rise. Higher-value items carry a lower percentage take rate but generate stronger absolute margins, and Levesque said the AI pricing system is built to protect sell-through rates as that mix continues shifting upward. On the discovery side, The RealReal began testing a conversational shopping agent built in partnership with Google. Levesque described a shopper searching for a dress for a fall wedding in upstate New York, an example she said the tool is built to interpret with specificity most keyword search cannot match. The company is also using AI to automatically enrich every listing with data on occasion, collection and current trends, information Levesque said used to require manual entry and now makes inventory more discoverable both on and off the platform. Intake Automation Cuts Costs as Volume Scales Athena, the company’s AI-powered intake system, is central to how The RealReal is managing rising volume without proportional headcount growth. The system, which automates authentication and cataloging steps that used to require manual attribution, is on pace to process close to half of all items by the end of the year, up from 35% at the end of 2025. The company has begun extending Athena into higher-value items that previously required manual handling, a shift Levesque called the next phase of the system’s rollout. Gopal said the technology is removing multiple dollars in processing cost per unit while also increasing speed to sell, a combination that let operating expenses leverage roughly 470 basis points year over year even as volume grew. What Else Stood Out The RealReal’s Real Partners referral program, which connects the company with stylists, real estate agents and other professionals, brought in consignors whose average consigned value is four times that of a typical new consignor. The company onboarded two large Japanese vendors into its international dropship program during the quarter, along with new partners in France and Italy, part of an asset-light approach to expanding supply outside the U.S. 44% of new consignors in the quarter came from The RealReal’s existing buyer base, up from 40% two quarters earlier, a sign the company’s buyer-to-seller flywheel is strengthening.