The lack of robust measures of dietary exposure hinders a quantitative understanding of causal relationships between diet and health. Non-targeted metabolite fingerprinting was used to explore the relationships between citrus exposure in free-living human subjects, estimated by a FFQ, and the chemical content of urine. Volunteers (study 1, n 12; study 2, n 11) were classified into high-, medium- and low-frequency citrus consumption groups. Overnight and spot fasting urine samples were obtained after exposure to a standardised citrus-free evening meal. The urine samples were analysed by flow injection electrospray-ionisation MS followed by supervised multivariate data classification analysis to discover discriminatory features associated with the level of citrus exposure. Good separation of high and low citrus consumption classes was achieved. Deeper exploration of high-ranked explanatory mass signals revealed several correlated signals derived from proline betaine. Targeted analysis of the relative levels of proline betaine in both fasting and overnight urine samples demonstrated good correlation with FFQ exposure data. Acute exposure of volunteers to orange juice resulted in the appearance of proline betaine and several biotransformed products in postprandial urine samples. Biomarker validation showed sensitivities of 80·8–92·2 % and specificities of 74·2–94·1 % (false discovery rate-adjusted P values < 0·05) for elevated proline betaine in participants who reported high citrus consumption. Proline betaine biotransformation products displayed weaker quantitative relationships with habitual citrus exposure. Targeted screening for the presence of biotransformation products of hesperidin and narirutin, known to be abundant in oranges, revealed that they were relatively poor indicators of citrus exposure.