Abstract Background Good reading skills are important for appropriate functioning in everyday life, scholastic performance, and acquiring a higher socioeconomic status. We conducted the first systematic review and meta-analysis to quantify possible deficits in specific reading skills in people with a variety of mental illnesses, including personality disorders (PDs). Methods We performed a systematic search of multiple databases from inception until February 2020 and conducted random-effects meta-analyses. Results The search yielded 34 studies with standardized assessments of reading skills in people with one or more mental illnesses. Of these, 19 studies provided data for the meta-analysis. Most studies (k = 27; meta-analysis, k = 17) were in people with schizophrenia and revealed large deficits in phonological processing (Hedge’s g = −0.88, p < 0.00001), comprehension (Hedge’s g = −0.96, p < 0.00001) and reading rate (Hedge’s g = −1.22, p = 0.002), relative to healthy controls; the single-word reading was less affected (Hedge’s g = −0.70, p < 0.00001). A few studies in affective disorders and nonforensic PDs suggested weaker deficits (for all, Hedge’s g < −0.60). In forensic populations with PDs, there was evidence of marked phonological processing (Hedge’s g = −0.85, p < 0.0001) and comprehension deficits (Hedge’s g = −0.95, p = 0.0003). Conclusions People with schizophrenia, and possibly forensic PD populations, demonstrate a range of reading skills deficits. Future studies are needed to establish how these deficits directly compare to those seen in developmental or acquired dyslexia and to explore the potential of dyslexia interventions to improve reading skills in these populations.