DraftKings Accused of Aiming Bonus Bets at High-Loss Customers via AI
A new allegation claims DraftKings deployed predictive models to steer free-bet promotions toward gamblers its system flagged as most likely to lose money.
DraftKings is facing allegations that it used artificial intelligence to decide which customers received bonus bets, directing the promotions toward players its models predicted would generate the most losses. The claim, reported by Gadget Review, centers on the company's use of predictive analytics to rank bettors by expected profitability rather than by engagement or randomness. The story has drawn heavy search interest around "draftkings free bet," suggesting the accusation landed with the platform's own user base.
The mechanics described are familiar to anyone who has followed sportsbook data operations. Operators collect granular behavioral data — bet size, frequency, sport, in-game versus pre-game action, deposit and withdrawal patterns — and feed it into models that forecast future revenue per user. A bonus bet is not a giveaway in that framework; it is a marketing spend allocated where the expected return is highest. If the allegation holds, the targeting would have concentrated free bets on the accounts least likely to come out ahead.
DraftKings has not publicly confirmed the specific targeting practice described in the report, and the company has previously said its promotional programs comply with applicable gaming regulations. Regulators in several U.S. states, including Massachusetts and New Jersey, have scrutinized sportsbook bonus and VIP programs over the past two years, with some issuing fines tied to responsible-gambling failures. Those cases set a template: the legal question is usually not whether a company used data, but whether it did so in a way that exploited vulnerable customers.
The allegation also raises a technical point that will matter to any investigation. Predictive models that estimate customer value are standard across retail, streaming and financial services, and they are not illegal on their own. What turns them into a regulatory problem is the input variable — if a model is trained on projected losses rather than projected activity, the output is effectively a list of people to target with inducements. That distinction, not the presence of AI, is where the scrutiny will focus.
For the broader AI industry, the case is another example of a predictive system being judged by its objective function rather than its accuracy. A model can perform exactly as designed and still produce an outcome that regulators, and the public, find unacceptable. Gambling is an early test case because the harm is measurable and the data trail is unusually complete; the same argument is already being made about insurance pricing, lending and ad delivery.
What to watch next is whether state gaming authorities open a formal inquiry and whether DraftKings discloses how its promotional models are built and audited. If the company publishes that detail, it could set a benchmark for how sportsbooks explain algorithmic marketing. If it stays quiet, the allegation will keep circulating without a factual answer — which, for a business built on customer trust, may be the worse outcome.