Autism spectrum disorder (ASD) is an early onset neurodevelopmental disorder. Evidence suggests that ASD patients have abnormalities in information processing. Event-related potential (ERP) technique can directly record brain neural activity in real time. P300 is a positive ERP component which can measure the neuroelectrophysiological characteristics of human beings and has the potential to discover the pathological mechanism of ASD. However, P300 studies on ASD patients are incongruent and the disparities may be caused by several factors. By searching PubMed, Embase and Cochrane Library databases, a meta-analysis of P300 component difference between ASD group and typically developed (TD) control group was conducted. Results of amplitude and latency of P3b and P3a from included studies were synthesized. Random effect model was chosen and standardized mean difference (SMD) was calculated. Subgroup analysis was used to identify the source of heterogeneity and to test the effect of different experiment factors. A total of 407 ASD patients and 457 TD controls from 32 studies were included in this analysis. Reduced amplitude of P3b was found in ASD group (SMD = −0.505, 95 % CI −0.873, −0.138) compared with TD group, but no difference of P3b latency, P3a amplitude, or P3a latency was found between groups. Subgroup analysis showed that oddball paradigm elicited attenuated P3b amplitude in Pz electrode among ASD subjects. This meta-analysis suggests ASD patients have abnormalities in P300 component, which may represent for deficits in cognition, attention orientation and working memory processing, particularly in the decision-making processing condition.