The “Self-medication hypothesis” that has been developed to explain the effect of nicotine in improving aspects of cognitive impairment in schizophrenia remains controversial. This systematic review and meta-analysis compared cognitive functions between smoking and non-smoking schizophrenia patients. The PubMed, PsycINFO, EMBASE, Web of Science, and Cochrane Library databases were systematically and independently searched. Basic demographic and clinical characteristics, smoking history and cognitive performance were recorded. Seven of the 11 studies included in the study, had meta-analyzable data. Compared to non-smoking schizophrenia patients, their smoking counterparts showed significant deficits on the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS)-immediate memory (n = 739), the RBANS-total score (n = 739) and the Continuous Performance Test-Identical Pairs (n = 157). Two of the 4 studies without meta-analysable data did not report significant group difference in performance on the Wechsler Digit Span Task and the Beck Cognitive Insight Scale, while the other 2 studies found that non-smokers outperformed than smokers in problem solving and visual learning. In conclusion, this systematic review and meta-analysis found that smoking schizophrenia patients had worse performance in certain cognitive tasks than non-smoking patients, casting doubts on the validity of the “self-medication hypothesis” that needs to be further examined.