Can AI Help You Make Money Sports Betting? I Tried It for a Week to Find Out

If the scaremongers are to be believed, not only is AI turning the entire internet into machine-translated slop, it’s also coming for your job and your boyfriend.

That said, it’s also spawned a bunch of tools that can help you cheat at uni work and real work, which it’s hard to be too mad about. And given all the money being poured into it, there are surely some other useful applications. For example, can I get this technology to help me make an inconsequential bit of money by gambling on sports? There’s only one way to find out!

My plan is to start with £100 and see how I get on over the course of a week. I begin by asking ChatGPT to find some medium risk bets for the weekend’s Premier League games. “Given Manchester City’s strong home performance and Chelsea’s impressive away record, it’s plausible that both teams will find the net. Combining this with a City win offers attractive odds,” the chatbot tells me. 

I bet £5 on City to win and both teams to score, at odds of 3.5. Lo and behold, Man City wins 3-1, meaning I net £12 profit. ChatGPT also advises me to bet on Newcastle to beat Southampton, which seems as close to a done deal as you can get, given Southampton’s terrible run this season. I end my first day £20 up. 

So far, so good—but the next two bets recommended by ChatGPT fall short. I should have known better than to back a win for Ange Postecoglou’s Spurs. I also ask it to recommend a goal scorer in Portsmouth’s game against West Brom and it suggests I bet on Josh Koroma to find the net, which seems unlikely given he hasn’t played for the club since 2023. 

Artificial intelligence—from large language models like ChatGPT to more specialized tools that promise to predict the outcome of sports games—uses vast amounts of data to identify patterns, carry out tasks, and learn from experience in a way that makes it feel like you’re interacting with a human, not a computer. But as anyone who’s used these tools will know, they’re not immune to mistakes. 

“You’ve got to be really careful about [defining] what is an AI driven tip,” says Stuart Simms, CEO of FairPlay Sports media, the parent company of odds comparison website Oddschecker. Stuart explains that ChatGPT essentially crowdsources information by drawing from online sources and its training data, meaning it takes a consensus view of the betting market—which isn’t all that helpful if you’re trying to get an edge on the competition. He adds that successful betting syndicates and bookmakers use AI to generate raw, statistical outcomes, then apply a layer of human insight to refine those results. Easy!

Given ChatGPT’s patchy start, and Stuart’s advice, I move on to some AI tools designed specifically for sports betting. Google floods me with results claiming to offer an advantage over the bookies, and I settle on DeepBetting, a French AI start-up that recommends football games based on their betting value. It offers up a couple of free picks per day, or a monthly subscription that unlocks all of its recommendations for €29.99 per month. 

The site’s first suggestion is a bet on Genoa to beat Monza in Italy’s Serie A. The bookmaker I use offers odds of 1.49 for a “draw no bet” stake on Genoa to win. The implied probability of this outcome, based on the bookmakers’ odds, is 67.11%, but DeepBetting’s algorithm gives it a 75% chance, rating it a good value bet. I place £10 on it, cross my fingers, and it comes in.

At the same time, I sign up to two US-based AI betting apps—JuiceReel and Leans.ai—as a lot of these new tools appear to be focused on US sports. 

JuiceReel founder Ricky Gold describes his platform as like OnlyFans meets copy trading (where you replicate the trades of top investors). Users are encouraged to hook their betting accounts into JuiceReel when they sign up, and the tool runs an AI bot trained on this data to generate two picks per day that it believes to be good value. Users can also pay to follow the bets of the app’s most successful users, which is where Ricky’s comparison comes in. 

He says the bot has been profitable for about a year-and-a-half, with a win ratio of a little over 54%. “We get this really interesting data set that doesn’t exist anywhere else, because we have everyone’s track record,” he explains. “We know who’s a good better, we know who’s a medium better, and we know who’s a bad better, and we know what these good betters are doing and what these bad betters are doing.”

Leans.ai is a more premium service, with prices to match; a subscription costs $299 a month. The tool’s founder says he doesn’t want to reveal his identity lest angry punters find out who he is or where he lives, in the event the algorithm “goes south”, so he goes by the alias Steve Westfield. 

Steve says his bot looks at millions of data points to work out good value bets and puts out five picks per day based on this analysis. I haven’t got a clue about American college basketball, so I take his word for it and chuck some money on Stanford University to beat the snappily-named Southern Methodist University Cardinals. 

The odds for the picks by both JuiceReel and Leans are a little higher than DeepBetting’s recommendations, and the results are patchy. Not all of the college basketball games they suggest to bet on are available on the UK bookies I use, so it’s also a smaller sample size. “When you’re analysing our product, just know that it’s built on long-term analytical trends,” says Steve. “It might have fairly good days, it might have terrible days, but the way that it’s designed is built to hold up over the course of a season.”

Although the results of my US sports bets are mixed, I decide not to hunt Steve down, partly because my feelings are tempered by DeepBetting’s picks consistently coming in. I’m up on my initial £100 and it feels pretty good. Who knew winning free money could be so addictive? I email my editor—maybe another week of this would increase the sample size (and my winnings). “You don’t feel out of control in any way, right?” he fires back. “I feel like I have to ask this stuff as an editor.” It’s pretty sobering. We agree I’ll go all in for the final weekend and call it quits wherever I end up. 

I don’t go all in on one game, but spread my bankroll across a few matches on the final Sunday, upping both the stakes and the potential payout. The results are good: all wins, bar one loss and one voided bet. By the end of the week I’m £80 up, with DeepBetting doing much of the heavy lifting to deliver my AI-gotten gains.

Overall, the experiment was a success, but there’s one question I can’t shake—and it could be asked of anyone offering up online cheat codes to untold riches, from crypto bros and Forex traders to Moneyball-esque AI bettors. If your system is so successful, why are you selling it to no-marks like me? 

The team behind DeepBetting didn’t respond to requests for comment for this story. Although I hit an incredibly lucky streak using their tool—all 13 of their suggested bets came in—it obviously doesn’t always have a 100% win rate. For a recent game between Fiorentina and Inter Milan in Serie A that I didn’t bet on, the AI suggested a “draw no bet” wager on Inter, but Fiorentina won 3-0. You can see more previous results from its recommendations here.

When I put the question to the US guys, they have similar answers. “People are looking for these quick win, unicorn tools that don’t exist in reality,” says Steve. “I hate to say this, because I know you’ll publish it, but [Leans.ai is] kind of a boring product, in the sense that you’re just barely above that 52.4% [the ratio needed to break even vs the bookies]. But over time that adds up.” 

Ricky agrees: “That mild edge is a little less exciting. There are strategies that earn you a couple percent ROI [return on investment] that are pretty good. It’s not as exciting to make 4% over a year to a lot of amateur bettors that are looking to turn 10 bucks into a million.”

While it does feel good to turn the tables on the bookies, if only for a week, they are—of course—already using AI to their own advantage. Big US bookmaker DraftKings has said it uses AI to price its odds, and Stuart from Oddschecker says this is standard across the industry,  though the level of technological sophistication varies.

AI is also seeping into the industry in more nefarious ways. A Deloitte report from last year highlighted how generative AI—the tech behind chatbots and image generators like Dall-E—is being used by betting companies. The report said AI could help create “individually themed online slot games that can respond to a player’s voice and even generate novel content in response to a player’s behavior and game history”. Case in point: UK start-up Future Anthem is using AI to keep users engaged on casino websites through “real-time personalisation”. 

Experts have raised concerns about the gambling industry embracing AI to maximise its profits, which are of course still as healthy as ever. Last year, the gross gambling yield for UK bookies—that’s the amount wagered by customers minus the winnings paid to them—stood at £15.6 billion. 

So, despite my profitable week, the old cliche rings true: the house always wins. 

You can follow Hayden on X @haydenvernon