OBJECTIVE: To investigate how smoking status is recorded in UK primary care; to evaluate whether appropriate multiple imputation (MI) of smoking status yields results consistent with health surveys. SETTING: UK primary care and a population survey conducted in the community. PARTICIPANTS: We identified 354 204 patients aged 16 or over in The Health Improvement Network (THIN) primary care database registered with their general practice 2008-2009 and 15 102 individuals aged 16 or over in the Health Survey for England (HSE). OUTCOME MEASURES: Age-standardised and age-specific proportions of smokers, ex-smokers and non-smokers in THIN and the HSE before and after MI. Using information on time since quitting in the HSE, we estimated when ex-smokers are typically recorded as non-smokers in primary care records. RESULTS: In THIN, smoking status was recorded for 84% of patients within 1 year of registration. Of these, 28% were smokers (21% in the HSE). After MI of missing smoking data, the proportion of smokers was 25% (missing at random) and 20% (missing not at random). With increasing age, more were identified as ex-smokers in the HSE than THIN. It appears that those who quit before age 30 were less likely to be recorded as an ex-smoker in primary care than people who quit later. CONCLUSIONS: Smoking status was relatively well recorded in primary care. Misclassification of ex-smokers as non-smokers is likely to occur in those quitting smoking at an early age and/or a long time ago. Those with no smoking status information are more likely to be ex-smokers or non-smokers than smokers.