Background: Since people with mental illness are more likely to die from cancer, we assessed whether people with mental illness undergo less cancer screening compared with the general population. Methods: In this systematic review and meta-analysis, we searched PubMed and PsycINFO, without a language restriction, and hand-searched the reference lists of included studies and previous reviews for observational studies from database inception until May 5, 2019. We included all published studies focusing on any type of cancer screening in patients with mental illness; and studies that reported prevalence of cancer screening in patients, or comparative measures between patients and the general population. The primary outcome was odds ratio (OR) of cancer screening in people with mental illness versus the general population. The Newcastle-Ottawa Scale was used to assess study quality and I 2 to assess study heterogeneity. This study is registered with PROSPERO, CRD42018114781. Findings: 47 publications provided data from 46 samples including 4 717 839 individuals (501 559 patients with mental illness, and 4 216 280 controls), of whom 69·85% were women, for screening for breast cancer (k=35; 296 699 individuals with mental illness, 1 023 288 in the general population), cervical cancer (k=29; 295 688 with mental illness, 3 540 408 in general population), colorectal cancer (k=12; 153 283 with mental illness, 2 228 966 in general population), lung and gastric cancer (both k=1; 420 with mental illness, none in general population), ovarian cancer (k=1; 37 with mental illness, none in general population), and prostate cancer (k=6; 52 803 with mental illness, 2 038 916 in general population). Median quality of the included studies was high at 7 (IQR 6–8). Screening was significantly less frequent in people with any mental disease compared with the general population for any cancer (k=37; OR 0·76 [95% CI 0·72–0·79]; I 2 =98·53% with publication bias of Egger's p value=0·025), breast cancer (k=27; 0·65 [0·60–0·71]; I 2 =97·58% and no publication bias), cervical cancer (k=23; 0·89 [0·84–0·95]; I 2 =98·47% and no publication bias), and prostate cancer (k=4; 0·78 [0·70–0·86]; I 2 =79·68% and no publication bias), but not for colorectal cancer (k=8; 1·02 [0·90–1·15]; I 2 =97·84% and no publication bias). Interpretation: Despite the increased mortality from cancer in people with mental illness, this population receives less cancer screening compared with that of the general population. Specific approaches should be developed to assist people with mental illness to undergo appropriate cancer screening, especially women with schizophrenia.