Circulating vitamin C and carotenoids are used as biomarkers of fruit and vegetable intake in research, but their comparative validity has never been meta-analysed. PubMed, EMBASE, CENTRAL, CINAHL and Web of Science were systematically searched up to December 2013 for randomised trials of different amounts of fruit and vegetable provision on changes in blood concentrations of carotenoids or vitamin C. Reporting followed PRISMA guidelines. Evidence quality was assessed using the GRADE system. Random effects meta-analysis combined estimates and meta-regression tested for sub-group differences. In all, nineteen fruit and vegetable trials (n 1382) measured at least one biomarker, of which nine (n 667) included five common carotenoids and vitamin C. Evidence quality was low and between-trial heterogeneity (I
2) ranged from 74 % for vitamin C to 94 % for α-carotene. Groups provided with more fruit and vegetables had increased blood concentrations of vitamin C, α-carotene, β-carotene, β-cryptoxanthin and lutein but not lycopene. However, no clear dose–response effect was observed. Vitamin C showed the largest between-group difference in standardised mean change from the pre-intervention to the post-intervention period (smd 0·94; 95 % CI 0·66, 1·22), followed by lutein (smd 0·70; 95 % CI 0·37, 1·03) and α-carotene (smd 0·63; 95 % CI 0·25, 1·01), but all CI were overlapping, suggesting that none of the biomarkers responded more than the others. Therefore, until further evidence identifies a particular biomarker to be superior, group-level compliance to fruit and vegetable interventions can be indicated equally well by vitamin C or a range of carotenoids. High heterogeneity and a lack of dose–response suggest that individual-level biomarker responses to fruit and vegetables are highly variable.