Human behaviours that are motivated by and indicative of personal interests can be utilised to personalise behavioural recommendations used to promote health and well-being. Behavioural and psychological studies show that (1) personal interests are demonstrated differently in individuals’ daily activities; and (2) drawbacks of self-reporting methods, such as forgetfulness and providing socially accepted answers rather than actual ones, may negatively impact the reliability and validity of the recognition process. To address these two challenges, we propose an adaptive approach that infers personal interests from continuously- and passively-sensed smartphones location data. We evaluate our approach based on two longitudinal datasets gathered by human participants going about their normal daily activities. Our results indicate that our approach successfully identifies interests consistent with those reported by participants, matching or outperforming alternative approaches. We also see high inter-personal variation, suggesting a future role for personalisation in our approach.