DraftKings Accused of Using AI to Target High-Loss Gamblers with Bonus Bets

A New York Times investigation published on September 19, drawing on interviews with more than 40 former employees and internal documents, reports that DKNG built a machine-learning elasticity score in 2023 that weighs a customer's play frequency, account balance and loss-to-wager ratio, and uses it to direct where promotional spending goes.

Two things in that account need to be kept apart. The existence and design of the score is documented, reported from internal material the Times reviewed. The characterisation of it as targeting problem gamblers is how former employees describe it, and DraftKings disputes that reading. The company's on-record response, which appears in consistent terms across outlets, is that it rejects any implication of unfair targeting and that promotions go to customers showing sustained, engaged use. It also called one analyst's supporting work preliminary and inconclusive.

The scale is not in dispute and comes from the company itself. DraftKings disclosed at its Investor Day that roughly $400 million of promotional spending was AI-automated in 2025, and executives have cited a 13% margin improvement from the approach. Whatever the intent, automated promo allocation is now a material part of how the business is run.

Former employees also describe a separate 2024 harm-detection model, developed by a researcher named Nestor Hernandez, that was shelved. DraftKings' explanation, per its chief responsible gaming officer, was an insufficient evidence base.

No regulator has opened an inquiry into this specific reporting, and no source ties a share price move to it. The existing pressure is structural rather than new: the federal SAFE Bet Act, various state bills, and a 2025 Baltimore lawsuit were already in motion. The question worth tracking is whether this reporting converts any of that into a formal state gaming commission review, which is the venue that could actually constrain promotional practice.