AI in iGaming: What's Actually Deployed, and What's Still Just Rules

Almost every operator says "we use AI." A survey of 151 industry specialists (by The Playa and NEXT) paints a far more honest picture: adoption is wildly uneven, most of what gets called AI is still ordinary rules, and the single biggest lesson isn't about technology at all. Let's go through the six use-cases and see where AI has actually taken root, where it makes money, and where it's still just a marketing label.

Where AI Actually Took Root: Support

The one area you can call mature is support. Here 65% of operators and platforms use AI/ML, 29% have fully deployed it, and 32% even name support the single most effective AI direction.

The specifics: 71% of those working with AI in support use chatbots, 55% use operator hints. Ticket prioritization and sentiment analysis show up more rarely, because the cost of a mistake there is higher and the compliance risks real.

The takeaway is simple: AI entered first where a mistake is cheapest. That's a logical start, but it's not an "AI-run company" yet.

The Industry's Real Lesson: a Prediction Nobody Acts On Is Worthless

The most important finding of the whole study is about predictive analytics, and it's not about models, it's about team behavior.

Predictive analytics is used on an ongoing basis by just 16% of operators, with another 24% testing. Among those who already use it, 47% work with LTV, 35% with churn, 24% with early VIP detection.

Now the lesson itself: two-thirds of operators with churn models receive signals that a player is about to leave, and do nothing about them. The prediction exists, the action doesn't, and so neither does the result.

Where the prediction is tied to follow-up communication or an offer, monthly retention rose by 2-4 percentage points, and in some cases by more than 7. The sample is small, so treat that as a signal of possible effect rather than a market benchmark, but the direction is unambiguous: the value isn't in the model, it's in what you do with its prediction.

Personalization: Most "AI" Is Still Rules

This is where there's the most marketing and the least real AI. Take bonuses. Full AI/ML bonus recommendations have been deployed by only 11% of companies, with another 25% testing. Among those using AI for bonuses, 80% personalize at the segment level, and only 7% have reached an individual bonus flow for a specific player.

Same story in segmentation: 38% use or test AI, 14% have fully deployed it. But even among those who call their segmentation AI/ML, only about a quarter let models find behavioral groups on their own. The rest still split players by preset parameters: deposit sum, activity, recency of last session.

In other words, a large share of what the industry calls AI personalization is ordinary rules under a new name. That isn't bad in itself, but it's important not to confuse the label with a real predictive model.

Where the Money Already Shows: Lobby Personalization

The most rewarding scenario by returns is lobby personalization, where AI reorders games and selections based on player behavior. 17% have fully deployed it, another 25% are testing or planning.

A telling detail: among operators with AI, 47% refresh the lobby daily or more often, versus 27% among companies without AI. Automation simply lets you do it more often than by hand.

Early results look strong: among those measuring the effect, GGR grows 3-9% and retention 2-6 percentage points. The clearest results are on active players, since there's more behavioral data on them. Again with the small-sample caveat, but it's the same logic as Netflix recommendations: putting the right content in front of the right person makes money.

Acquisition: It Works, but Almost Nobody Measures It

At the acquisition stage, AI helps predict LTV, optimize ads, and assess traffic. 40% use or test it, 11% have fully deployed it. Among users, 70% predict LTV, 65% optimize ads, 34% analyze traffic quality and fraud.

And here's the second systemic problem: 72% of those analyzing traffic don't measure the budget they save. The tool exists, but the team can't prove its value in numbers. It rhymes with the main lesson: without measurement and action, even working AI stays a cost rather than an investment.

How to Tell Real AI From a Label

Four quick checks before you believe an "AI-powered" claim, your own or a partner's:

1. Does a prediction trigger an action, or just sit in a dashboard? A churn score nobody acts on is theater.
2. Do the models find groups themselves, or do humans preset the parameters? The latter is rules, not ML.
3. Is the effect measured in money (retention, GGR, saved budget), or taken on faith?
4. Is personalization per-player, or just per-segment relabeled as "AI"?

If the answers are "action, self-found groups, measured in money, per-player," it's real. If not, it's a label, and that's fine to admit as long as you don't budget for it as if it were more.

Summary

• AI adoption in iGaming is uneven: only support is mature (65% use it, 29% fully).
• The main lesson isn't about tech: two-thirds of teams with churn models don't act on the signals, and where they do, retention rises 2-7 points.
• Most "AI personalization" is still rules: only 7% have an individual bonus flow.
• Real money already shows in lobby personalization (GGR +3-9%), yet 72% at acquisition don't even measure the effect.

AI doesn't make an operator smart automatically. The winner is whoever turns a prediction into an action and measures the result in money, not whoever says the word "AI" the loudest.