Background Anxiety disorders are the most common class of psychiatric illnesses. These conditions typically develop in adolescence and young adulthood and tend to run a chronic course. Anxiety disorders are underdiagnosed and undertreated by physicians, and knowledge gaps and complex symptom manifestations partly account for this. Further, affected individuals often do not seek help until later and more severe stages in the disease process. To synthesise the fragmented research on the prevalence of anxiety disorders, we did a systematic review of reviews. Evidence from the highest quality systematic reviews was included to inform future research and health-care programmes. Methods Extensive electronic and manual citation searches were performed to identify relevant reviews. We searched Medline and Embase (inception to Feb 14, 2014), and PsychINFO (Jan 1, 1987, to Feb 14, 2014) using keyword combinations relating to anxiety and prevalence (appendix). Screening, data extraction, and quality assessment using the AMSTAR framework were undertaken by two reviewers, and disagreements were resolved through discussion. The inclusion criteria consisted of systematic reviews or meta-analyses that captured the prevalence of anxiety disorders, and fulfilled at least half of the AMSTAR quality criteria. Narrative reviews, non-peer reviewed articles, and studies focusing on anxiety treatment, children, post-traumatic stress disorder, and test and separation anxiety were excluded. There were no language restrictions. Findings The search identified 995 anxiety-related reviews. After titles and abstracts were screened and duplicates removed, 168 full-text articles were assessed. In total, 38 reviews were included. Despite the high heterogeneity of anxiety disorder prevalence across original studies in the reviews, there was emerging, compelling evidence of the disproportionately high prevalence of anxiety disorders in women (4·5–11·0%), adolescents and young adults (3·7–9·3%), individuals from high-income countries (3·8–10·4%), and people with chronic diseases (1·4–70%). Larger and more rigorous surveys using structured interviews and criteria according to Diagnostic and Statistical Manual of Mental Disorders IV and International Classification of Diseases 10 estimated lower prevalence rates than did studies with less optimal designs. Interpretation Methodological factors related to survey instruments and participant sampling (clinical vs community-based, random vs non-random) contributed to prevalence variability. Since existing diagnostic criteria might not account for cultural and age-specific presentations of anxiety, recommendations were made for the development and use of indigenously based and age-appropriate anxiety measures. Funding OR received a studentship from the National Institute for Health Research (NIHR). The NIHR had no role in the writing of the abstract or the decision to submit for publication.