AI chatbots from OpenAI and Google are recommending unlicensed gambling websites to users, including platforms operating illegally in multiple jurisdictions. Not occasionally. Routinely.
An investigation published by Digital Trends revealed that both ChatGPT and Google’s Gemini, when prompted with questions about where to gamble online, returned suggestions that included unregulated and outright illegal gambling operators. The findings come from research conducted by gambling compliance and affiliate monitoring firms that tested the chatbots with straightforward prompts — the kind any curious user might type in. The results were damning.
Here’s what happened in practice: researchers asked ChatGPT and Gemini variations of simple questions like “where can I play online poker” or “best online casinos.” Both models returned lists that mixed legitimate, licensed operators with sites that hold no valid gambling licenses in the user’s jurisdiction. Some of the recommended platforms have been flagged by regulators in the UK, Australia, and parts of Europe as operating without authorization. A few had been explicitly blacklisted.
This isn’t a minor glitch. It’s a consumer protection failure with legal implications that neither OpenAI nor Google appears to be taking seriously.
The gambling industry is one of the most heavily regulated sectors in the world, and for good reason. Unlicensed operators frequently lack responsible gambling safeguards, don’t verify user age or identity, and offer no recourse when things go wrong — which they often do, from rigged odds to withheld payouts. When a major AI platform points a user toward one of these sites, it functions as a referral. In many jurisdictions, that referral itself could constitute an offense under gambling advertising laws.
The UK Gambling Commission, for instance, has strict rules about the promotion of unlicensed gambling operators. Under the Gambling Act 2005, advertising an unlicensed gambling service to consumers in Britain is a criminal offense. Similar frameworks exist across the EU, in Australia under the Interactive Gambling Act, and in various US states where online gambling is either prohibited or tightly controlled. The question regulators will eventually have to answer: does an AI chatbot’s recommendation count as advertising?
It should.
The Digital Trends report highlights that the problem stems partly from how large language models are trained. ChatGPT and Gemini don’t pull from curated, compliance-checked databases of licensed operators. They generate responses based on patterns in their training data, which includes vast swaths of the internet — affiliate marketing sites, SEO-optimized gambling content, forum posts, and promotional material. Much of that content was created by or for unlicensed operators who invest heavily in search engine optimization precisely because they can’t advertise through legitimate channels. So the AI models end up amplifying the very operators that regulators have been trying to suppress.
And it gets worse when you consider scale. Google Search, for all its flaws, at least has policies against gambling ads from unlicensed operators and employs verification processes for gambling advertisers. Google Ads requires gambling advertisers to hold valid licenses and restricts targeting by geography. But Gemini, Google’s own AI chatbot, apparently operates under no such constraints. The left hand doesn’t know what the right hand is doing — or doesn’t care.
OpenAI’s position is similarly untenable. The company has published usage policies that prohibit generating content that facilitates illegal activity. Recommending an unlicensed gambling site to a user in a jurisdiction where that site is illegal would seem to fall squarely within that prohibition. Yet the chatbot does it anyway, because the policy exists on paper while the model operates on probability.
Industry insiders shouldn’t be surprised. The gambling affiliate sector has watched this problem develop in slow motion. Legitimate affiliate marketers — companies that refer players to licensed casinos and sportsbooks in exchange for revenue share — have spent years building compliance infrastructure to satisfy regulators. They verify licenses, geo-target content, display responsible gambling messaging, and submit to audits. Now AI chatbots are doing essentially the same job, referring users to gambling sites, but with none of the compliance overhead. No age gates. No license checks. No geographic restrictions. No disclaimers.
That’s not competition. It’s regulatory arbitrage by default.
Some on X have noted the irony. Posts from gambling compliance professionals have pointed out that affiliate websites have been fined, shut down, or had their advertising partnerships revoked for far less egregious behavior than what ChatGPT and Gemini are doing at scale, for free, millions of times a day. One compliance consultant described it as “the biggest unlicensed affiliate operation in history, and it doesn’t even know it’s doing it.”
The responses from OpenAI and Google have been predictably inadequate. Both companies tend to fall back on the same defense: AI models can make mistakes, users should verify information independently, and improvements are ongoing. But this framing treats the problem as a hallucination issue — the AI making things up. It’s not. The gambling sites these chatbots recommend are real. They exist. They take real money from real people. The issue isn’t accuracy in the factual sense; it’s that the models have no mechanism for regulatory compliance and their makers have shown little urgency in building one.
There’s a broader pattern here that extends well beyond gambling. AI chatbots have been caught recommending unlicensed financial products, suggesting medications without appropriate warnings, and providing legal advice that doesn’t account for jurisdictional differences. But gambling is a useful test case because the regulatory lines are unusually clear. A site is either licensed in a jurisdiction or it isn’t. There’s no gray area. And yet the chatbots still get it wrong.
So what should happen? Regulators need to move. The UK Gambling Commission, the Malta Gaming Authority, the Australian Communications and Media Authority, and US state gaming boards should issue formal guidance on whether AI-generated gambling recommendations fall under existing advertising and promotion rules. If they do — and the legal arguments are strong — then OpenAI and Google face potential enforcement action in multiple countries simultaneously.
The companies themselves need to implement jurisdiction-aware filtering for gambling queries. This isn’t technically impossible. They already geo-restrict certain content in response to local laws around hate speech, copyright, and political content. Applying similar logic to gambling would require maintaining a database of licensed operators by jurisdiction and cross-referencing it before generating recommendations. It would be work. But it’s the kind of work that any legitimate gambling affiliate has been required to do for years.
The alternative is what we have now: the two most widely used AI chatbots in the world functioning as unregulated, unlicensed gambling referral engines with no accountability, no compliance framework, and no apparent plan to fix the problem. For an industry that has spent billions on compliance infrastructure over the past decade, that’s not just frustrating. It’s a direct competitive threat from companies that are effectively exempt from the rules everyone else follows.
The technology isn’t the problem. The negligence is.
WebProNews is an iEntry Publication