Urban public squares are key settings for social interaction, leisure and pedestrian movement, yet most behavioural studies rely on manual observation or surveys of limited spatial and temporal scope, and computer-vision studies generally treat square users as a homogeneous group. This study examines three tourist squares of contrasting morphology: Times Square in New York (amorphous), Senado Square in Macau (nuclear) and Piazza Santa Maria in Venice (closed). Public webcam footage was combined with open spatial data; pedestrian trajectories were extracted using YOLOv8 and BoT-SORT with an age-classification module; isovist area (IA) was calculated to quantify spatial openness; and generalised linear models were used to test how IA, points of interest (POI) and facility distance relate to the trajectories of children, adults and older adults. The results show that amorphous squares recorded the highest IA values, followed by nuclear and closed squares. Pedestrian volume rose from morning to evening in all three cases, but its spatial distribution differed by typology. Children’s trajectories were most strongly associated with proximity to facilities, adults moved in a linear, destination-oriented manner, and older adults favoured open sightlines and proximity to POI. The framework offers a transferable method for evidencebased, age-friendly square design.