Objectives: To compare the clinical diagnosis of chronic obstructive pulmonary disease (COPD) with results of post-bronchodilator spirometry in general practice, and examine practitioner, practice and patient characteristics associated with agreement between clinical and spirometric diagnoses. Design, setting and participants: General practitioners from practices in Sydney identified eligible patients aged 40-80 years seen in the past year and prescribed respiratory medications whom they regarded as having COPD. Between November 2006 and April 2008, we collected information on the GPs and their practices, and demographic information, smoking status, comorbidities, respiratory medicines use, vaccination status, quality of life and spirometry results for participating patients. Main outcome measures: Frequency of COPD diagnosis on spirometry; odds ratios for characteristics associated with agreement between clinical and spirometric diagnoses. Results: 56 GPs from 44 practices participated in the study. Of 1144 eligible patients, 445 were recruited (mean age, 65 years; 49% male). Of these, 257 (57.8%) had post-bronchodilator spirometry consistent with COPD ± asthma, 16 (3.6%) had asthma only, 82 (18.4%) had normal spirometry, and 90 (20.2%) had other spirometric diagnoses. Having a spirometer in the practice was not predictive of agreement between clinical and spirometric diagnoses. Older patient age was significantly associated with correct diagnosis, while higher numbers of comorbidities were associated with misdiagnosis. Conclusions: A substantial proportion of patients clinically identified as having COPD in general practice do not have the condition according to spirometric criteria, with inaccurate diagnosis more common in patients with comorbidities. Policy and practice change is needed to support the use of spirometry in primary care.