Several studies analysed the associations between dietary carbohydrate intake, glycaemic index (GI) and glycaemic load (GL) and digestive system cancers; however, the results remain controversial. This study was to perform a meta-analysis evaluating the quantitative and dose–response associations between carbohydrate intake, GI and GL, and risk of digestive system cancers. We searched medical and biological databases up to June 2018 and identified twenty-six cohort studies and eighteen case–control studies. Meta-analytic fixed or random effects models were applied to process data. We also performed dose–response analysis, meta-regression and subgroup analyses. We found that high levels of GI were significantly associated with the risk of digestive system cancers at the highest compared with the lowest categories from cohort studies (summary relative risk (RR)=1·10, 95 % CI 1·05, 1·15). Similar effects were observed from case–control studies of the comparison between the extreme categories, but the difference did not reach statistical significance (summary OR=1·28, 95 % CI 0·97, 1·69). We also observed significant dose–response association between GI and digestive system cancers, with every 10-unit increase in GI (summary RR=1·003; 95 % CI 1·000, 1·012 for cohort studies; summary OR=1·09; 95 % CI 1·06, 1·11 for case–control studies). In addition, both cohort studies and case–control studies indicated that neither dietary carbohydrate intake nor GL bore any statistical relationship to digestive system cancers from the results of the highest compared with the lowest categories analyses and dose–response analyses. The results suggest a moderate association between high-GI diets and the risk of digestive system cancers.