How sportsbooks use AI for pricing, risk limits & marketing

A sourced guide to where machine learning sits in sports betting, what a report on DraftKings’ ‘elasticity’ model alleged & what regulators in Massachusetts & Great Britain have said about AI.

Where AI sits in a sportsbook

The sources describe artificial intelligence (AI) & machine learning at work in several parts of the sports betting business, from Genius Sports’ GeniusIQ platform to DraftKings’ promotions & an AI-generated character in a Midnite advert.

The sections below cover pricing, the risk that skilled bettors pose to bookmakers, promotions, safer gambling interactions with customers & adverts, in that order. Notes on oversight follow.

The next section also draws on a 2004 paper by Steven Levitt in The Economic Journal, which describes bookmakers as ‘more skilled at predicting the outcomes of games than bettors’.

Pricing & trading

Genius Sports describes GeniusIQ as its ‘next-generation data and artificial intelligence platform’ that ‘ingests billions of data points from multiple sources’ to ‘decipher gameplay, identify patterns and forecast probabilities’. The company says GeniusIQ powers solutions across ‘performance analysis, fan engagement, advertising and sports betting’.

Among the applications listed on the GeniusIQ page, Genius Sports includes ‘Sportsbook Optimisation’, which the page sets out as ‘Maximising in-game handle and engagement’.

According to Steven Levitt’s 2004 paper in The Economic Journal, bookmakers ‘announce a price, after which adjustments are small and infrequent’, & they ‘take large positions’ on the outcome of games rather than matching buyers & sellers.

Using a data set of approximately 20,000 wagers on the National Football League, placed by 285 bettors at an online sports book as part of a high-stakes handicapping contest, Levitt concluded that bookmakers ‘are more skilled at predicting the outcomes of games than bettors and systematically exploit bettor biases by choosing prices that deviate from the market clearing price’. By contrast, he writes that in most financial markets ‘The prevailing price is that which equilibrates supply and demand.’

Levitt gave an example: if bookmakers know local bettors prefer local teams, they can skew the odds against the local team. He added that the distortion has limits, since bettors who know the ‘correct’ price can profit if the posted price strays too far from the true odds.

Risk & winning customers

Pricing errors carry real risk for a bookmaker. Levitt wrote that if a bookmaker sets the wrong line it ‘can lose money, even in the long run’, & that a small number of skilled bettors ‘could prove financially disastrous’.

The authors of a research paper titled Beating the bookies with their own numbers say they designed a strategy to beat football bookmakers with their own numbers. They used the probability information in publicly available odds to find mispriced bets, rather than building their own forecasting model.

The strategy was profitable in a 10-year historical simulation using closing odds, a 6-month simulation using minute-to-minute odds & a 5-month period of real-money betting, according to the paper. The authors concluded that ‘the football betting market is inefficient’ & that the industry compensates with ‘discriminatory practices against successful clients’.

The abstract also says the industry ’employs teams of data analysts to build forecast models’. The sources do not describe in detail how operators use AI to flag or restrict winning accounts.

Marketing: the ‘elasticity’ controversy

One case in the sources concerns promotions at DraftKings. According to a New York Times investigation, as summarised by Gizmodo, DraftKings built a model in 2023 designed to score users on ‘elasticity’.

Gizmodo describes elasticity as an economic concept that is ‘more like a measure of responsiveness to incentives’. Here it described how much a user responded to free bets & promotions, with more elastic users generating more revenue, according to the report. Covers says DraftKings reportedly used the metric to estimate how much more a gambler might wager after receiving a promotional offer.

The Times said a 2023 memo it reviewed found slots players were more elastic, & that early 2024 data showed highly elastic slots players lost more money than less elastic ones, according to Gizmodo. The investigation drew on interviews with more than 40 former employees.

One former employee, Jayden Butts, told the Times his task was to apply a model built around the question: ‘Is this person going to give us more than we’re giving them?’ If the answer was yes, he said, the company would ‘open the floodgates’.

The report also described former employees trying to build a model to predict when users were heading towards a crisis. One of them, Jake Shanin, told the Times an internal crisis prediction model was showing promise. The Times reported that in early 2025 the team prepared to share it with company officials, but ‘the day of the presentation, the meeting was canceled’, & two other similar attempts were shelved, according to two former employees.

DraftKings told the Times it ‘rejects any implication that its marketing practices are unfair or improperly targets customers’, Gizmodo reports. It said promotions are ‘directed toward customers who demonstrate sustained, engaged use of our platform, not toward customers based on their losses.’

Lori Kalani, DraftKings’ chief responsible gaming officer, told the Times the company monitors for ‘potentially risky behaviors’, though she said evidence had shown risk modelling technology was not helpful, according to Gizmodo.

What regulators have said

According to Covers, the Massachusetts Gaming Commission (MGC) said it would evaluate sportsbooks’ use of AI following the Times report. Covers says it will look at all its licensed sports betting operators, not only DraftKings.

Covers reports that the MGC has not determined that DraftKings committed any wrongdoing. Chairman Jordan Maynard said: ‘These AI technologies are evolving rapidly across our society and we share the concerns over their application.’

According to Covers, executive director Dean Serpa & staff will speak with DraftKings about its use of AI & machine learning, including promotions. Maynard said: ‘Any further action or policy-making can then be determined by the Commission when or if appropriate.’

The review builds on earlier work. Covers reports that the MGC commissioned a 2025 study, titled ‘AI and Player Risk Identification and Response’, through the University of Nevada Las Vegas’ International Gaming Institute, which found a ‘governance gap’ between technological advancements & existing company practices & regulatory guidelines. It recommended the AI task force that Serpa now leads.

Separately, in Great Britain, the published text of a speech by chief executive Andrew Rhodes at the International Association of Gaming Regulators (IAGR) Conference on 20 October 2025 says ‘artificial intelligence is starting to make itself felt’. The page notes that it is ‘the speech as drafted and may slightly differ from the delivered version’.

The drafted text says safer gambling interactions with customers had been happening but that their quality was often ‘really inconsistent’, & that operators are ‘increasingly now using generative AI’ to try to improve the consistency of customer interactions. It also flags ‘the general trend towards hyper personalisation’, saying risks must be balanced so that people do not start to ‘feel an increased intensity that nobody necessarily intended’.

AI-made ads

The ASA has ruled on a gambling advert featuring an AI-generated person. According to the ruling, it concerned a paid TikTok ad for Midnite, seen on 3 July 2026, showing a man using a gambling app.

Dribble Media Ltd t/a Midnite said it had not sanctioned the ad, according to the ruling. It said an affiliate, Limay Media, had made it, & that it had since ended that contract. Limay Media Ltd said the man was an AI-generated character & that the prompt used to create him did not specify his age.

According to the ruling, the CAP Code states that no-one who is, or seems to be, under 25 may be featured gambling or playing a significant role in gambling ads. The ASA judged that the character’s youthful facial features, plain grey hoodie & excited manner made him seem under 25.

It found the ad breached CAP Code (Edition 12) rules 16.3 & 16.3.14, & said the ad ‘must not appear again in the form complained of’. The ASA added that the ad was identified through its Active Ad Monitoring system, which itself uses AI to search for online ads that might break the rules.

What good governance looks like

The sources do not set out an industry standard for governing AI in betting. They do point to recurring themes.

Human review is one. Limay Media told the ASA its usual process required managerial approval before publication, but that this was not followed. It said it had since added safeguards, including age-related prompts, reviews of perceived age & keeping records of the AI generation & approval process.

Oversight is another. The UNLV study described a ‘governance gap’, & Covers reports that the MGC formed its AI task force shortly after the findings were published. At DraftKings, according to the Times as quoted by Gizmodo, a crisis prediction model was due to be shown to company officials, but the meeting was cancelled.

What bettors should know

Offers differ between customers. Covers notes that promotions ‘can vary based on a customer’s betting activity, loyalty, account history, and previous promo usage’.

Prices are not always set to balance money on each side, according to Levitt’s 2004 research. Winning customers may face different treatment, according to the authors of the arXiv paper, who describe ‘discriminatory practices against successful clients’.

Gizmodo says DraftKings is ‘prone to emailing users’ & sending push notifications to app users, which Redditors have described as difficult to turn off.

Questions

Do sportsbooks use AI to set odds?

The sources do not say directly how operators use AI to set odds. Genius Sports says its GeniusIQ platform applies AI & machine learning to ‘forecast probabilities’ & lists ‘Sportsbook Optimisation’ among its applications.

What was the DraftKings ‘elasticity’ model?

According to the New York Times, as summarised by Gizmodo, it was a 2023 model scoring how strongly users responded to free bets & promotions. Gizmodo reports that DraftKings told the Times it rejects any implication that its marketing practices are unfair.

Has Massachusetts found DraftKings broke any rules?

No. Covers reports that the Massachusetts Gaming Commission has not determined any wrongdoing & is reviewing AI use across all its licensed sports betting operators.

Did the under-25 rule apply to an AI-generated character in the Midnite ad?

Yes, in that case. The ASA’s Midnite ruling found the ad breached the CAP Code because, although the video was AI-generated, ‘the overall impression created by the individual depicted was that he seemed to be under 25’.

Do sportsbooks restrict winning customers?

The authors of an arXiv paper say their strategy was profitable during a 5-month period of real-money betting with football bookmakers & describe ‘discriminatory practices against successful clients’. The sources do not detail how AI is used in such decisions.

Sources

• covers.com
• gizmodo.com
• gamblingcommission.gov.uk
• arxiv.org
• pricetheory.uchicago.edu
• geniussports.com
• asa.org.uk

Sources checked 11 October 2026. Information only. 18+.