A national census of all UK specialist alcohol treatment agencies was carried out in order to enquire about client characteristics and the broad features of the treatment service received by clients on a specific day (4 December 1996). A total of 728 agencies were eligible for the census and the overall response rate was 41%. When the effects of response bias were examined, correcting for under- and over-representations of different types of agencies made very little difference to findings. A rough estimate from the census is that 10,000 individuals were seen for treatment or advice regarding an alcohol problem on census day in the UK. Treatment services are dominated by the non-statutory sector, which accounted for almost two-thirds of all clients seen on census day. Despite demographic changes in drinking patterns over the last 10-20 years, the 'typical' client attending treatment services is still a middle-aged man and the client sample as a whole showed evidence of major social disadvantage and dislocation. Compared with national figures, the sample showed an over-representation of 'Irish' clients in English agencies but there was no under-representation of other ethnic groups. The most common category of staff providing treatment was 'counsellors', and the complaint most often presented by clients concerned 'psychological well-being'. Twenty-eight per cent of clients were recorded as being in residential treatment. The most common form of treatment received was therapy or counselling on a one-to-one basis, which accounted for two-thirds of clients receiving some form of psychosocial treatment. The majority of clients in detoxification (60%) received it as in-patients. Forty-two per cent of treatment episodes were estimated to have lasted for more than 3 months and 18% had continued for over a year. The most commonly cited use of other services was for Alcoholics Anonymous (AA), although AA was mentioned in relation to only 9% of all clients. Excluding cannabis, 85% of clients were recorded as having no illicit drug use. Suggestions are made as to how the response rate and other features of the census could be improved if the exercise were to be repeated.