DraftKings used AI to target bonuses at players who bet more and lose more often, while dropping a separate harm-detection tool

The New York Times reviewed internal DraftKings documents and spoke with more than 40 former employees, and the picture is pretty clear: since 2023, the sportsbook has used an AI model to predict how a specific player will respond to a bonus and whether that promo will bring in more money than DraftKings spends on it. A separate AI system meant to flag risky play early was built, then shelved.

1. According to the NYT, the promotional model helped DraftKings raise the margin on bonus bets by 13% in 2025, and the company told investors that AI was used to personalize incentives on hundreds of millions of dollars.
2. Since 2023, another DraftKings team had been working on a protection system that would analyze player behavior such as deposits, withdrawals, attempts to chase losses, and other signals to estimate the risk of problematic gambling before it escalated.
3. That project’s presentation was canceled in early 2025, and development was later stopped entirely. DraftKings also walked away from two other similar initiatives, while FanDuel and Fanatics already use comparable systems.
4. DraftKings said it rejected the harm-detection model because there was a “lack of evidence” that it worked. Instead, the company said it manually tracks more than 20 signs of risky behavior.
5. The sportsbook also denies that it gives bonuses more often to customers with large losses, saying its promotions go to people who use the platform regularly and actively. Citizens Bank estimated that DraftKings spent about $3 billion on customer bonuses in 2025.