Many studies have been conducted to identify potentially useful behavioural markers of harmful behaviour using online gambling data sourced from operators. However, less is known about how such markers could be used to identify higher risk products. The study examined whether certain categories and subcategories of product are more strongly associated with behavioural markers of harm than others. Analyses were based on 6 months of data ( N = 100,000 individual gamblers) drawn from the population of UK users of an online gambling website in 2022. Measures included individual-level expenditure data across multiple forms of gambling and a series of literature-informed behavioural markers of harm including declined deposits, easing/removing responsible gambling settings, within-session repeat deposits (or ‘top-ups’), bonus-seeking behaviour and gambling at unusual hours. Negative binomial models examined how well the number of active days playing different products predicted behavioural markers of harm. All markers apart from easing/removing responsible gambling settings appeared to covary with the number of active days engaging in specific product types, most notably slots, in-playing betting and some most forms of combination bets on sports. These findings highlight the potential value of using measurable markers to differentiate the risk and potential harm associated with different online products.