Feeding patterns, encompassing feed intake, feeding frequency, duration and rate, may show rapid and extensive deviation from basal levels during welfare challenges. As feeding patterns can be continuously recorded by electronic feeding stations, they may serve as feasible, real-time indicators of welfare. We systematically reviewed the literature on the relation between feeding patterns and growing-finishing pig welfare, covering both negative states, i.e. health and behavioural issues, and positive states. Studies on health issues have focused on the impact of clinical disease and thermal stress on feed intake, while other health issues and the behaviours underlying intake have received less attention (e.g. subclinical disease & skin lesions, feeding frequency & duration). Moreover, feeding patterns are commonly reported at group level and with a daily time step, despite extensive variation in feeding behaviour between pigs and across the day, suggesting that some individual- or time-specific impacts on feeding patterns may have been overlooked. Regarding behavioural issues, tail biting alters feeding frequency during the weeks preceding an outbreak, but may not alter feeding patterns once the outbreak has commenced. Feed competition strongly alters feeding patterns of both dominant and subordinate pigs, but it is unclear which changes reflect appropriate adaptation to the social environment and which indicate severe social stress. Few studies have considered feeding patterns and positive states concomitantly. Feed intake differences have been reported between pigs housed in barren or enriched environments, but studies using animal-based indicators of positive states, such as play or tail posture, are currently absent. We conclude that feeding patterns are promising indicators of real-time pig welfare, especially if several feeding behaviours are included, with short time steps (e.g. hourly) at the level of the individual pig. We propose that negative states may induce deviations from basal patterns, while positive states may be characterised by stable patterns.