How DraftKings Uses AI to Target Gamblers Likely to Lose

DraftKings’ reported use of artificial intelligence raises a difficult question for the sports-betting industry: If a platform can identify customers most likely to lose money, can it also identify customers most likely to experience gambling-related harm?

Reporting summarized in the supplied sources says DraftKings uses artificial intelligence and data science to identify gamblers who may continue betting and lose money over time, then target selected customers with betting incentives. The same reporting describes allegations that a machine-learning model designed to identify gambling risk was sidelined rather than used as aggressively for customer protection.

Those claims require careful attribution. The available material does not provide DraftKings’ complete model documentation, accuracy data, internal policies, or a verified company response. A related lawsuit reportedly alleges that DraftKings “weaponized” artificial intelligence to encourage addictive behavior and increase betting activity. Those allegations have not been established in court. Source 8

The issue extends beyond one company. Sportsbooks collect detailed behavioral data, use predictive models to shape marketing, and increasingly personalize the betting experience. That technology can help identify valuable customers. It may also reveal patterns associated with financial distress, chasing losses, or compulsive gambling.

What the Reporting Says About DraftKings’ Use of AI

DraftKings reportedly identifies customers likely to lose

The central reported practice involves using artificial intelligence, machine learning, and data science to classify customers according to predicted betting behavior and financial outcomes.

The objective is not necessarily to predict the result of a particular game. A sportsbook does not need to know which team will win to determine whether a customer may be commercially valuable. Instead, a model could estimate whether a user will continue betting, make additional deposits, accept promotions, place frequent wagers, or generate net losses over time.

The New York Times reporting summarized in the supplied sources says DraftKings uses artificial intelligence to identify and target gamblers considered most likely to lose. Source 2

The phrase “likely to lose” can have several meanings:

• A customer expected to continue betting after losses.
• A customer predicted to maintain high betting volume.
• A customer unlikely to withdraw deposited funds.
• A customer likely to respond to a promotion.
• A customer whose future wagers may produce net revenue for the sportsbook.

The available source summaries do not establish which definition DraftKings used. They also do not disclose the model’s variables, accuracy, update frequency, or decision rules.

Data can make betting promotions highly personalized

A sportsbook could potentially analyze many forms of customer behavior, including:

• Deposit and withdrawal patterns.
• Betting frequency.
• Bet size.
• Sports and markets selected.
• Time between wagers.
• Responses to bonuses and free bets.
• Changes in betting activity.
• Whether a user returns quickly after a loss.

These examples describe common categories of behavioral signals that predictive systems may examine. They are not confirmed inputs for DraftKings’ models based on the supplied sources.

A model could divide users into groups for retention campaigns, personalized incentives, promotional offers, or increased engagement. Customers may receive different messages because the system predicts that some are more likely than others to keep betting or respond to a particular offer.

The technology changes marketing from broad advertising to individualized persuasion. Instead of offering the same promotion to every customer, a platform can theoretically direct an incentive toward users whose behavior suggests that the offer will produce more betting.

The reported strategy raises a conflict-of-interest question

A customer’s predicted losses may represent commercial value for a sportsbook. The same customer’s betting pattern may indicate financial pressure or gambling-related harm.

That overlap creates the central ethical concern. If a platform recognizes that a customer is likely to continue betting and lose money, should the system increase that customer’s exposure to promotions? Or should the prediction trigger a review, a limit, or a responsible-gambling message?

The supplied reporting describes criticism that DraftKings applied predictive technology to revenue generation while resisting comparable use of technology to identify and protect customers at risk of gambling addiction. Source 1 Source 4

That criticism does not by itself prove that DraftKings caused addiction or deliberately targeted people because they were vulnerable. It does show why the design and purpose of a predictive model matter.

How AI Targeting Could Increase Gambling Risk

Incentives can encourage more betting

Sportsbooks commonly use marketing tools such as:

• Free bets.
• Deposit matches.
• Odds boosts.
• Personalized promotions.
• Loyalty rewards.
• Time-sensitive offers.

These incentives can influence when, how often, and how much a customer bets. For an occasional bettor, a promotion may function as ordinary advertising. For someone already betting persistently or escalating deposits, the same message may encourage further exposure to losses.

The supplied sources do not establish the precise content, timing, or frequency of DraftKings’ incentives. They also do not establish whether specific promotions caused specific customers to increase their losses.

The ethical distinction concerns targeting. Broad advertising treats customers similarly. Predictive targeting can identify people whose behavior suggests that they are especially likely to continue betting. If those customers also show warning signs of harm, personalized incentives may intensify an existing problem.

Predictive targeting can exploit behavioral patterns

Machine-learning systems can detect correlations that customers may not recognize themselves. A model may identify that a person who deposits more frequently, returns quickly after losing, and accepts certain offers is likely to place additional bets.

Potential warning signs in sports betting can include:

• Increasing deposit amounts.
• Longer or more frequent betting sessions.
• Rapid betting after a loss.
• Attempts to recover losses through larger wagers.
• Betting across multiple events in a short period.
• Repeated returns after an account balance falls.

These are general examples, not confirmed DraftKings model inputs. They also do not prove gambling addiction. High activity can reflect many circumstances, and a risk model can produce both false positives and false negatives.

Personalized marketing can still be more persuasive than general advertising because it responds to an individual’s observed behavior. The customer may see an offer when the system predicts the highest likelihood of another bet.

“Likely to lose” is not the same as “likely to win”

A common misunderstanding about sports-betting algorithms is that they must predict game outcomes. A sportsbook may use statistical models to price wagers, but customer-targeting systems serve a different purpose.

A customer can be commercially valuable even when the platform does not predict the outcome of a particular wager. The relevant prediction may involve:

• Continued betting.
• High turnover.
• Low withdrawal likelihood.
• Repeated promotion acceptance.
• Future net losses.
• Persistence after losing.

This distinction changes how the public should understand AI in sports betting. The model may not be gambling against the customer by forecasting every result. It may be optimizing the relationship around the bet: who receives an offer, when the offer appears, and how likely the user is to keep wagering.

The Alleged Risk-Scoring Model

What the reported risk scores were intended to do

One supplied account says DraftKings had a machine-learning model that assigned users “risk scores.” The account presents the model as a tool that could identify customers showing signs of gambling-related trouble. Source 7

A risk score could serve several purposes:

• Flagging accounts for review.
• Triggering responsible-gambling outreach.
• Pausing promotional messages.
• Offering limits or cooling-off periods.
• Supporting referrals to assistance.
• Identifying behavior that requires human attention.

The available summaries do not provide technical documentation, the score’s scale, the thresholds used, or evidence showing how accurately it identified harmful gambling. The model should not be described as diagnosing gambling addiction.

Reports that the model was sidelined

The reporting summarized in the supplied material alleges that DraftKings sidelined, or stopped using, a machine-learning model intended to identify users at risk of gambling-related harm. Source 7

That allegation matters because it suggests a possible asymmetry in the use of customer data. The company may have had access to behavioral signals relevant to both commercial targeting and customer protection. Critics argue that those signals were more useful to the business when they predicted future losses than when they indicated potential harm.

The available summaries do not independently establish why the model was sidelined, whether it was accurate, whether it was legally required, or whether another responsible-gambling system replaced it. They also do not establish that DraftKings ignored every indication of customer risk.

Why risk scores require safeguards

Automated risk scoring has serious limitations.

False positives can flag ordinary or high-volume bettors who are not experiencing harm. False negatives can miss customers whose behavior changes rapidly or does not match the model’s assumptions. A score may reflect financial activity without proving addiction, distress, or an inability to control betting.

A responsible system would require:

• Human review of significant alerts.
• Clearly documented intervention thresholds.
• Regular testing for accuracy and bias.
• Independent audits.
• Privacy and data-security protections.
• A process for customers to correct inaccurate information.
• Limits on how long risk classifications remain active.

Risk scores should support assistance, not permanently label customers. Their purpose should be harm reduction, not more aggressive marketing.

Commercial AI Versus Responsible-Gambling AI

The supplied reporting presents two contrasting applications of predictive technology:

Commercial model: Identify users likely to continue betting and lose money, then direct incentives or engagement efforts toward them.

Protective model: Identify users whose activity may indicate gambling-related risk, then reduce exposure and offer support.

Critics view this as an asymmetry because both systems may rely on overlapping data. Deposit frequency, betting persistence, promotion response, and post-loss behavior can help predict revenue. They may also help identify vulnerability.

This criticism remains a reported ethical argument, not an established legal conclusion. The question is whether a company that uses data aggressively to increase customer value should apply comparable resources to preventing foreseeable harm.

Responsible-gambling measures can include:

• Deposit limits.
• Self-exclusion.
• Cooling-off periods.
• Account monitoring.
• Reality checks.
• Spending information.
• Referrals to support services.

These tools differ from proactive behavioral intervention. A generic warning may satisfy a formal requirement, while a personalized system could recognize that a particular customer’s behavior has changed and intervene at the relevant moment.

The supplied sources do not establish the full scope or effectiveness of DraftKings’ responsible-gambling policies. They also do not show whether the company’s protective tools were integrated with its marketing systems.

A platform can present responsible-gambling information while simultaneously delivering personalized offers designed to encourage more betting. The practical effect depends on which system receives greater access to behavioral data and greater authority over the customer experience.

Legal and Public Allegations Against DraftKings

What the lawsuit allegations claim

A reported lawsuit alleges that DraftKings “weaponized” artificial intelligence to target gamblers, encourage addictive behavior, and increase betting activity. Source 8

Those are allegations, not proven facts. The supplied source summary does not establish that a court accepted the claims or found DraftKings liable.

A court evaluating the case could examine:

• Internal model objectives.
• Marketing records.
• Customer communications.
• Promotion timing.
• Account histories.
• Evidence of company knowledge.
• The relationship between predictive scores and customer activity.
• Applicable gambling, consumer-protection, and advertising laws.

The legal outcome would depend on the evidence and the specific claims pleaded.

Why legal claims and investigative reporting differ

Investigative reporting may describe internal practices, interviews, documents, or allegations. A lawsuit must establish specific legal elements. Depending on the claim, those elements could involve a duty, unfair conduct, misrepresentation, causation, consumer harm, and damages.

A report can raise serious questions without resolving legal liability. A complaint can describe alleged conduct without proving it. Readers should consult court filings, docket entries, and official rulings for procedural updates.

The role of social media posts in the information chain

Several supplied sources are social media posts linking to or summarizing reporting. These posts can identify a story, but they may omit qualifications, dates, responses from DraftKings, methodological details, or legal context. Source 10

The New York Times report and court documents are stronger foundations where available. Entries containing only “200+,” “psg vs,” or “bandara” provide no usable factual context and should not support claims in this article.

What Remains Unknown

Important unanswered questions include:

• Which customer data was collected and used?
• How did DraftKings define “likely to lose”?
• Did the model predict net losses, future activity, or customer value?
• How frequently were predictions updated?
• What accuracy standards applied?
• Were different models used for marketing and responsible gambling?
• Who approved model changes?

These details determine whether the system was merely a marketing-segmentation tool or a more consequential decision-making system affecting vulnerable customers.

High betting activity does not automatically prove gambling addiction. A profitable customer and a customer at risk may display overlapping behaviors, but the concepts have different definitions.

More evidence would be needed to determine:

• Whether targeted users showed recognized harm indicators.
• Whether incentives increased their betting.
• Whether DraftKings knew about individual risk.
• Whether protective interventions were available.
• Whether customers could refuse personalized marketing.
• Whether risk scores influenced account decisions.

Without that evidence, it is not possible to conclude that every targeted customer was addicted or that every promotion caused harm.

The supplied summaries do not provide a verified DraftKings response, denial, explanation, or complete description of its responsible-gambling safeguards. No response should be invented.

What Responsible Use of AI in Sports Betting Could Require

Use risk signals to reduce harm, not increase exposure

When behavior suggests escalating or harmful gambling, the default response should be protection rather than additional incentives.

Potential safeguards include:

• Pausing promotional targeting.
• Applying cooling-off periods.
• Showing personalized spending and loss information.
• Offering voluntary deposit or wagering limits.
• Escalating cases for trained human review.
• Providing independent support resources.
• Allowing customers to self-exclude easily.

A risk signal should not automatically trigger punishment or account closure. It should trigger a proportionate, transparent, assistance-focused response.

Make automated decisions transparent

Customers should be told when their behavior is analyzed for personalization or risk monitoring. They should receive plain-language explanations of:

• What data may influence a promotion.
• Why an intervention or restriction occurred.
• What controls are available.
• How to request a review.
• How to correct inaccurate information.
• How to opt out of certain marketing uses.

Transparency cannot guarantee that a model is fair, but secrecy makes meaningful accountability difficult.

Require independent oversight

Sportsbook AI systems should receive independent oversight through:

• External audits.
• Regulatory reporting.
• Accuracy and bias testing.
• Documented model changes.
• Retention limits for sensitive data.
• Public responsible-gambling metrics.
• Reviews of customer outcomes.

Oversight should evaluate more than revenue performance. It should examine whether models increase deposits, prolong betting sessions, or affect customers showing signs of harm.

Conclusion: The Central Question Is How Predictive Power Is Used

DraftKings’ reported use of AI illustrates how predictive technology can distinguish customers by expected betting behavior and financial outcome. A system may identify people likely to continue betting and lose money without predicting the winner of any particular game.

The ethical concern is the possible overlap between commercial value and vulnerability. The same data that identifies a profitable customer may reveal escalating deposits, persistent betting, or attempts to chase losses. If a platform uses those signals to increase exposure while failing to deploy comparable protections, the imbalance deserves regulatory and public scrutiny.

The lawsuit claims remain allegations unless established by a court. The available summaries also leave important questions unanswered about DraftKings’ models, safeguards, internal decisions, and customer outcomes.

The broader question is straightforward: If a sportsbook can predict who is likely to lose, should it also be required to identify and protect customers most likely to suffer harm?

Frequently Asked Questions

How does DraftKings reportedly use AI to target gamblers?

DraftKings is reported to use artificial intelligence and data science to identify customers whose behavior suggests they may continue betting and lose money over time. The available summaries do not disclose the complete model design, data inputs, or accuracy. Source 2

What does “gamblers likely to lose” mean?

The phrase generally refers to customers predicted to generate future losses for the sportsbook. It does not necessarily mean an algorithm predicts a particular game. It may estimate future betting activity, spending, persistence, promotion response, or net losses.

Did DraftKings use AI to identify gambling addiction?

The supplied sources report allegations involving a machine-learning model that assigned users “risk scores” and could identify possible gambling-related risk. They do not establish that the model diagnosed addiction, how accurately it worked, or whether it was fully deployed. Source 7

What are the lawsuit allegations against DraftKings?

A lawsuit reportedly alleges that DraftKings used artificial intelligence to target gamblers, encourage addictive behavior, and increase betting activity. These claims have not been established in court based on the provided source summary. Source 8

Can sports-betting algorithms predict who will lose money?

Algorithms can estimate patterns associated with future betting activity or losses, but predictions are not certain. Reliability depends on the data, model design, validation methods, and customer behavior. A prediction does not prove that a customer has a gambling disorder.

What protections should sportsbooks provide?

Potential protections include proactive risk monitoring, limits on promotional targeting, cooling-off periods, self-exclusion, personalized spending information, human review, transparent explanations, independent audits, and accessible support resources.