Facts on BUSINESS
Article series
Social Media and the Psychology of PlayPart 3 of 4Start with Part 1: Old Rules, New Channels: Decoding Persuasion
Human Touch, Digital Reach in iGaming Marketing
The same customer data can serve marketing and player protection. What separates their purposes—and their evidence?
Personalisation can make an offer more relevant, while account analysis can help identify signs of gambling-related risk. Both may draw on the same behavioural records, but they pursue different outcomes.
That creates a practical tension for iGaming businesses. A system built to encourage another visit is not a safety system merely because it uses sophisticated technology. The important questions concern its objective, the quality of its evidence and what happens when a commercial opportunity conflicts with a protection concern.
A live dealer adds human presentation to a remotely accessed game. Product design and player protection still require separate assessment.
The Shape of Data-Driven Marketing
Customer data can inform which messages a player receives, when an offer is sent and which service issues need attention. Automation is not always AI, and the use of either varies between operators. Claims about adoption should identify the companies or sample being described.
Research also needs careful interpretation. A 2025 paper, AI Personalization and Its Influence on Online Gamblers' Behavior, compared platform data from 2016 and 2021. The authors state that the data did not record AI interventions. Its observations therefore cannot establish that AI caused changes in betting.
This distinction matters commercially and editorially. Behaviour changing over time is not proof of a particular algorithm's effect. A stronger evaluation would identify the intervention, compare appropriate groups and account for other changes in products, customers and circumstances.
Data can support different objectives. The label “AI” does not tell a reader whether a system prioritises engagement, protection or both.
The Paradox: Different Objectives, Similar Inputs
A change in deposit frequency can be relevant to a marketing model and to a risk-detection model. The conflict lies in the action that follows. One might recommend another offer; the other might prompt a review or an intervention.
These decisions need governance as well as technical accuracy. Businesses should be able to explain which signals they use, who can challenge a result, and how protection decisions affect promotional contact. A persuasive sales claim about a model is not a substitute for testing its real-world consequences.
Industry perspective
“That's why neuroscience needs to sit alongside data.”
— Paula Murphy, Mindway AI, in an interview with SiGMA News published 3 August 2025. This is a supplier's perspective on risk assessment, rather than independent evidence of a product's effectiveness.
Poker dealer at the PartyPoker LIVE MILLIONS Main Event, Copacabana Palace Hotel, Rio de Janeiro, March 2019. © Alexandre Rotenberg/Alamy Live News
Human Connection: Live Dealers
Live dealer games combine a human presenter with remote play. The dealer's presence can make the experience conversational and recognisable, while software handles the surrounding account and game interface.
Streaming, chat moderation and customer support are separate technical functions; they should not all be described as AI. Nor does a human dealer, an encrypted connection or regulatory oversight by itself establish a product's level of gambling-related risk.
A useful comparison would examine particular games, play speed, spending, session patterns and the characteristics of their players. This article does not have evidence sufficient to rank live dealer games as safer or more harmful than all other online casino formats.
Social Platforms and Gambling Promotion
Livestreams and chat communities can bring gambling promotion into a wider entertainment experience. The boundary between conversation, recommendation and paid promotion may be difficult for viewers to recognise.
Platform moderation and creator relationships should therefore be evaluated separately from an operator's own safety systems. A moderated conversation does not establish that a promoted product is suitable for a viewer, or that the operator is authorised to serve them.
Social features can support conversation, but moderation alone does not establish that gambling promotion protects vulnerable audiences.
AI Ethics: From Games to Marketing
Transparency, appropriate use of personal data and the ability to question a decision matter in marketing as well as in game design. A player should be able to understand why their account is subject to a restriction without having to understand a model's code.
Murphy's comments about neuroscience highlight a broader concern: account transactions do not capture every reason why someone gambles. Stress, circumstances and motivations require careful interpretation. They should not be inferred with unwarranted certainty from a sequence of bets.
Risk Detection: What Has Been Tested?
In a study first published online in 2022, Auer and Griffiths analysed account records for 1,287 European online casino players who completed the Problem Gambling Severity Index. Their machine-learning models identified patterns associated with self-reported problems, including more frequent deposits and higher losses per gambling session.
The result supports further work on risk identification, but a screening score is not a clinical diagnosis. A model developed from one operator's respondents also needs validation before its performance is assumed elsewhere. The authors disclosed gambling-industry consultancy relationships.
Predicting a screening outcome and demonstrating that an intervention reduces harm are different research tasks. Operators assessing a supplier should ask for evidence on both, including errors, missed cases and the consequences of acting on a flag.
Ethical Imperatives
The practical issue is accountability. A business needs to know whether a model's recommendations serve its stated purpose, whether staff can review them, and whether players can understand and challenge significant decisions.
Research into reward variability, such as Clark and Zack's 2023 conceptual review, provides a reason to examine product design carefully. It does not establish a quantified effect for a particular personalised campaign or prove that every gambling operator uses neuroscience in marketing.
Takeaways
Data-driven marketing and player protection can use similar inputs while pursuing different goals. Their effectiveness should be assessed separately.
Human contact, scientific language and an AI label are not guarantees of safety. The stronger test is whether the system has been validated for its purpose and whether its interventions produce a demonstrable benefit for players.
Sources and Further Reading
- AI Personalization and Its Influence on Online Gamblers' Behavior (2025) — observational comparison; no direct AI-intervention data.
- Auer and Griffiths: using AI algorithms to predict self-reported problem gambling — account data matched to screening responses.
- Clark and Zack (2023): Engineered highs — conceptual analysis of reward variability and behavioural addiction.
- SiGMA News interview with Paula Murphy, 3 August 2025 — source of the attributed quotation.
Correction and clarification — 14 October 2026: We removed unsupported revenue and acquisition-cost figures and claims of widespread neuroscience use. We clarified the observational nature of the personalisation research, corrected a research link that led to an unrelated chemistry paper, and replaced an unverified quotation with a traceable, attributed excerpt. We also distinguished prediction, diagnosis and evidence of harm reduction.