AI-Targeted Gambling Ads: How They Work and How to Limit Your Exposure

Seven operators, eight areas of questioning, one deadline

Massachusetts gave its seven licensed mobile sports betting operators until November 3 to explain in writing how they use artificial intelligence and machine learning in marketing, promotions and customer engagement. According to reporting by outlets such as CDC Gaming and Sports Betting Dime, the request came in an October 9 letter from Massachusetts Gaming Commission Executive Director Dean Serpa, covering eight operational areas, one of which is how operators identify potentially harmful gambling behaviour.

The question underneath it is blunt: can losses, deposits or riskier betting patterns make a customer more likely to receive a promotion? That is the point where AI targeting and gambling harm stop being separate topics. This piece explains how that kind of targeting actually works, what it looks like from the receiving end, and which of the available controls genuinely reduce your exposure, since they are not equally useful.

Per the same reporting, the MGC asked operators to disclose whether they rely on AI, algorithms, automated decision-making or comparable tools, to name the specific systems in use, state their purposes, and say whether each was built in house or supplied by a third party. The commission also asked whether those technologies make decisions about individual customers. Note what that implies: the regulator is not confirming that anything improper happened. It is asking because it does not currently know, and the operators do.

How AI-personalised gambling ads actually work

Algorithmic ad targeting in gambling is ordinary performance marketing applied to an unusually sensitive dataset. An operator already holds a detailed first-party record of every customer, feeds that record into models that predict who is likely to respond to what, and uses advertising platforms to reach those people and others who resemble them. None of the individual components are exotic. The combination is what regulators are now examining.

The data that feeds the algorithm

The richest signals come from inside the betting account itself, not from the open web. Typical inputs include:

• deposit frequency, deposit size, and whether deposits cluster after losses
• stake size and how it moves over a session
• which sports, markets and bet types you place, including in-play and parlay behaviour
• session length, time of day, and how long since you last logged in
• withdrawal behaviour, including cancelled withdrawals
• how you responded to previous pushes, emails and SMS promos
• app opens and in-app screens viewed, plus web browsing captured by advertising pixels

Outside the account, the mechanics are standard adtech. A hashed email address or phone number can be uploaded to an ad platform to match you to your social or search profile. From a seed list of existing high-value customers, the platform builds lookalike audiences of strangers who behave similarly. Dynamic creative then assembles the version of the ad most likely to work on your segment, and real-time bidding decides, in the milliseconds before a page loads, how much that impression is worth. This is why an ad can follow you into an app that has nothing to do with sport: the bid is for you, not for the page.

Why personalised timing can escalate risk

Timing, not wording, is what makes personalisation different from a billboard. A model that predicts churn can fire a reactivation bonus the week your account goes quiet. A model that predicts response can surface an odds boost on the exact market you backed and lost last weekend. An in-play notification can arrive at the minute a match turns, because the system knows you bet live and knows the fixture is running.

The optimisation target is a response, usually a deposit or a bet. The people easiest to make respond are the people already most inclined to bet. So a model doing exactly what it was built to do will, without anyone instructing it to, concentrate its attention on heavy users. That is the mechanism regulators are circling, and it does not require bad intent to produce an uncomfortable result.

Why this worries regulators and problem-gambling researchers

Personalised marketing intersects with three behaviours that problem-gambling research has long treated as risk markers: chasing losses, impulsive betting, and relapse after a period of abstinence. A promotion delivered at the moment of a loss, or to a dormant account belonging to someone who stopped on purpose, lands precisely where those behaviours are most fragile.

What is not established, at least publicly, is how much measurable harm is attributable to algorithmic targeting specifically, as distinct from gambling advertising in general. That evidence gap cuts both ways, and it is worth being honest about. Operators can reasonably point out that the same behavioural data also drives harm-detection systems, the tools that flag a customer for intervention. Dual use is real. Which is why the MGC letter asks about harm identification and marketing in the same document rather than treating them as separate files.

Signs an offer was built around your account, not the market

A general promotion goes to everyone at once and looks like it. A targeted one tends to show at least one of these patterns:

• a bonus or free bet arrives within hours of a deposit, a losing run or a cancelled withdrawal
• an odds boost matches the team, league or bet type you back most often
• the same operator’s ads appear across unrelated apps and websites for days after you visit the site
• SMS or email nudges restart after a stretch of inactivity, often framed as a welcome-back or reactivation bonus
• push notifications land during live events you have bet on before, rather than at fixed times
• a friend using the same app sees a different offer, or no offer, on the same day

None of these prove a model singled you out. Together, across weeks, they are a reasonable signal that your account behaviour is feeding the promotional calendar.

Three layers of control, and what each actually does

The controls available to you work at different depths, and the shallow ones are the most heavily advertised. Opting out of ad personalisation on a social platform usually changes which ads you see, not how many. The operator still holds its first-party record of you and can still email, text and push. Ranking the layers honestly:

The verdict, if you only do two things: use the operator-level opt-out and set a deposit limit. Those reach the data that matters most, because the operator’s own record of your deposits and losses is the input no ad-network setting can scrub. Ad personalisation toggles are worth flipping, but treating them as protection is the mistake. They are a relevance filter on a channel that is not the main channel.

Platform-level steps you can take today

1. Open the ad or privacy settings on each major social and search account and switch off ad personalisation, interest-based ads and data sharing with advertising partners.
2. Remove gambling-related interests and advertiser lists where the platform lets you see and delete them.
3. On mobile, deny app tracking permission and reset or delete your advertising identifier, then repeat occasionally since new installs ask again.
4. Use a browser that blocks third-party cookies by default, and clear cookies for betting sites you have visited.
5. Mute or report individual gambling ads where the platform offers a “see fewer ads like this” control for that category.

Operator-level tools that reach further

1. In each betting account, withdraw consent for marketing across every channel separately: email, SMS, push and phone. Opting out of one does not opt you out of the others.
2. Unsubscribe from promotional emails and reply STOP to marketing SMS, then check the account settings anyway, since the two systems are not always linked.
3. Set deposit, loss, time or session limits before you need them. Reductions usually apply immediately; increases typically carry a cooling-off delay by design.
4. Use a cool-off or time-out if you want a short break without closing the account.
5. Use self-exclusion if you want the account shut and marketing contact stopped, and register with your jurisdiction’s central scheme rather than one operator at a time where such a register exists.
6. Turn off in-app push notifications at the operating system level as well as in the app.

Every one of these reduces exposure. None eliminates it. Ads bought on the open market by brands you have never dealt with can still reach you, and lookalike modelling means you can be targeted on the basis of other people’s behaviour rather than your own.

Where gambling ad regulation appears to be heading

The Massachusetts probe is an information request, not a finding, and its outcome is unknown. What operators disclosed about their systems, and what the commission does with those answers, will determine whether anything follows in the form of guidance, conditions or rules. Treat anyone forecasting a specific result as guessing.

The broader direction is easier to read. Regulators in several markets have been moving toward tighter scrutiny of bonus offers, direct marketing and affordability-style checks, and the technical question Massachusetts has put in writing, whether automated systems make decisions about individual customers, is the same question sitting underneath all of it. Expect more regulators to ask it. Expect the answers to be commercially sensitive, which is part of why it has taken a formal letter to extract them.

What understanding this actually buys you

Knowing how the targeting works makes a well-timed offer easier to recognise as a product of your own account history rather than luck. That recognition is one layer of protection, and a thin one on its own. The limits and opt-outs described above sit underneath it, and the house edge in every product being advertised sits underneath them, unchanged by any promotion.

If promotional messages are pulling you back to an account you meant to leave alone, that pattern matters more than any ad setting. Look up the official responsible-gambling support service for your jurisdiction, usually listed by the regulator that licenses the operators you use, and start there.

About The Author

See author's posts