The Pulse
DraftKings Used AI to Target Bettors Likely to Lose
A New York Times investigation says DraftKings used machine-learning models to direct betting promotions toward customers expected to lose more money. The company disputed that its marketing targets customers based on losses while declining
AI.info Team ·
“The core question was, ‘Is this person going to give us more than we’re giving them?’ And if the answer is yes, open the floodgates.”
Jayden Butts, former DraftKings data analyst
DraftKings built a machine-learning system to identify customers most likely to respond to betting promotions by gambling and losing more money, according to an investigation published September 19, 2026, by The New York Times.
The investigation, based on internal documents, betting records and interviews with more than 40 former DraftKings employees, describes a company using customer behavior data to decide who should receive free bets, deposit bonuses and other incentives. Former employees told The Times that DraftKings continued developing similar systems for sports and casino betting while shelving or declining to deploy predictive tools that might identify customers at risk of gambling problems.
DraftKings Called the Model “Elasticity”
DraftKings began developing the casino model in 2023. The system analyzed dozens of data points, including how frequently a customer played, daily account balances and the relationship between money bet and money lost. It also incorporated a separate model estimating how likely a customer was to stop gambling.
The resulting score was known internally as “elasticity.” Customers with below-average scores were marked for fewer incentives, while those judged more likely to respond profitably remained eligible for promotions. In September 2023, Butts tested the approach on about 5,000 casino players before expanding the experiments to a larger population.
Butts said he initially believed the company wanted to reduce spending on customers unlikely to generate a return. Instead, he said supervisors told him DraftKings wanted to “redeploy” the money, directing more offers toward the customers expected to lose the most.
The Times found that high-elasticity bettors gambled more on online slots, on average, than those with lower scores in internal casino data from early 2024. A DraftKings memo from 2023 described slots revenue as “more elastic” than revenue from other casino games, meaning promotions were particularly effective at driving users toward those games.
A $3 Billion Promotional Machine
Promotions are central to DraftKings’s business. The company generated about $8.7 billion in gross revenue from sports and casino gamblers last year and distributed roughly $3 billion in promotions, according to research by Citizens Bank cited by The Times.
A DraftKings executive told investors that data science and analytics improved margins on promotion-driven sports bets by 13 percent in 2025. The executive also said the company used artificial intelligence to personalize hundreds of millions of dollars in promotional spending.
Former employees described additional systems designed to identify customers at risk of leaving the platform, encourage them to return and increase promotions for people whose gambling had become more aggressive. Six former employees who worked on promotional targeting said they believed the company’s methods could direct incentives toward vulnerable gamblers.
One former analyst described the practice as “as predatory as it sounds,” saying, “If you lose more, we give you more, so you keep playing more.” The analyst requested anonymity because he feared retaliation.
DraftKings Rejects the Characterization
DraftKings said it “rejects any implication that its marketing practices are unfair or improperly targets customers.” The company said promotions are directed at customers who show “sustained, engaged use of our platform,” rather than customers selected because of their losses.
The company also disputed Butts’s account of his firing and said his promotion test appeared “preliminary and inconclusive.” DraftKings said it had not seen or verified some of the internal materials and analysis described by The Times.
Lori Kalani, DraftKings’s chief responsible gaming officer, said the company’s business depends on customers “who are betting within their means, are betting for entertainment and betting for fun.” She said DraftKings monitors potentially risky behavior and does not send promotions to customers who have been flagged or who have placed themselves on self-exclusion lists.
The Risk Model DraftKings Did Not Use
DraftKings employees also developed predictive systems aimed at identifying gambling problems. In mid-2024, data scientist Nestor Hernandez began building a model that analyzed deposits and withdrawals, attempts to chase losses, age and gender, among other factors, to estimate whether a customer might soon need intervention.
Hernandez said the goal was to predict trouble “a few days or a few weeks in advance” so the company could act earlier. After he left DraftKings in November 2024, another team continued the work. The project was later shut down, according to former employees.
Jake Shannin, who worked on the project, said the team concluded that even simple rules could provide “a better-than-guessing sense” of whether a customer would eventually need an intervention. DraftKings leaders made a “collective decision” not to use predictive technology for problem gambling, Kalani said, because the company determined it was not evidence-based.
DraftKings instead relies on triggers tied to behavior such as unusually large deposits or sharp increases in time spent gambling. The company says it monitors more than two dozen indicators nationwide. When a customer triggers one, DraftKings may send a responsible-gaming message, an educational video, a questionnaire or, in some cases, close the account.
Promotions and Addiction
The system places much of the responsibility on customers to recognize their own behavior and seek help. Bryan Biehl, who told The Times he lost nearly $70,000 gambling online, received 40 DraftKings promotions during the first two weeks of December 2024 while undergoing therapy for addiction. He eventually joined self-exclusion lists after gambling one last time on Christmas Day.
DraftKings declined to comment on Biehl’s account. Kalani described promotions as a marketing tool used across industries and said shopping can also become problematic.
The investigation leaves a specific contradiction at the center of DraftKings’s data strategy: the company used detailed behavioral signals to estimate which customers could be induced to gamble and lose more, while rejecting similar predictive methods for identifying customers who might be heading toward a gambling crisis. DraftKings does use large-language models to monitor messages sent to customer-service representatives for signs of distress, but that system cannot identify customers who never ask for help.