Introduction Mild cognitive impairment (MCI) is regarded as a prodrome to dementia. Various cognitive tests can help with diagnosis; meta‐analysis of diagnostic accuracy studies would assist clinicians in choosing optimal tests. Methods We searched online databases for “mild cognitive impairment” and “diagnosis” or “screening” from 01/01/1999 to 01/07/2017. Articles assessing the diagnostic accuracy of a cognitive test compared with standard diagnostic criteria were extracted. Risk of bias was assessed. Bivariate random‐effects meta‐analysis was used to evaluate sensitivity and specificity. Results Eight cognitive tests (ACE‐R, CERAD, CDT‐Sunderland, IQCODE, Memory Alteration Test, MMSE, MoCA, and Qmci) were considered for meta‐analysis. ACE‐R, CERAD, MoCA, and Qmci were found to have similar diagnostic accuracy, while the MMSE had lower sensitivity. Memory Alteration Test had the highest sensitivity and equivalent specificity to the other tests. Discussion Multiple cognitive tests have comparable diagnostic accuracy. The Memory Alteration Test is short and has the highest sensitivity. New cognitive tests for MCI diagnosis should not be compared with the MMSE.