OBJECTIVES Anxiety and mood disorders are highly prevalent in Primary Care but research shows that general practitioners (GPs) fail to diagnose up to half of cases. In this study we try to determine the diagnostic accuracy of GPs’ diagnoses of anxiety disorders with and without any help from diagnostic (assisted vs unassisted diagnosis). METHODS We searched for articles published from January 1980 to June 2014 in 7 databases. We included studies in English, Spanish, French, and German reporting the ability of GPs to identify any anxiety disorder (DSM III/ IV/ IV-TR diagnostic criteria) in Primary Care community samples. We excluded studies from general population and those addressing specific physical or mental disorders, along with vignette and case-series studies.    Two authors independently performed abstract and full-text reviews and data extraction. Study was assessed with the QUADAS-2. Coupled forest plots summarized estimated studies’ sensitivity and specificity and 95% confidence intervals. We fitted random-effects meta-analysis models and undertook a bivariate meta-analysis to construct a summary Receiver Operator Characteristic Curve (sROC). RESULTS From a total 17.964 detected papers, 443 were included for full text review. So far, we have analyzed 111 papers, out of which 8 studies were included with N= 3608 patients with pooled anxiety prevalence 26% (CI=25-27%). Preliminary results shows an overall ROC curve with lower GP diagnostic accuracy when performing unassisted diagnoses for a total diagnostic accuracy 80% (CI=79.4-80.1) with overall sensitivity=49% (CI=45-53), and Specificity = 92% (CI=90-94). GP’s accuracy was higher with assisted diagnoses (86.7%, CI=85%-89%) than unassisted diagnoses (45.5%; CI=43.7%-47.3%). Specificity was lower in assisted (89.15%; CI=87.9-90.4) than unassisted diagnosis (92.5%; CI=91.9-93.1). CONCLUSIONS Low diagnostic sensitivity might hinder the adequate detection and management of anxiety in primary care. Results suggest that detection might be improved by using diagnostic tools. Results for all included articles will be presented.