Background The Camden and Islington Impact Personality Disorder Project provided psychologically informed consultation and training to probation staff in a Public Protection Unit and staff in two ‘approved premises’ (probation hostels) working with high‐risk prison sentenced offenders on licence in the community; some direct therapeutic work was also conducted jointly with probation officers. Most of the offenders had behaviours associated with personality disorder, although had not necessarily received a diagnosis of personality disorder. Aims To evaluate this service (the intervention) and test the hypothesis that its work would be associated with a reduction in recalls to prison. Methods Local and national probation records were used to identify all prison recalls made by 10 consistently available probation officers for 1 year before the intervention and up to 3 years after its introduction; reasons for recall and evidence of new arrests or charges were also extracted. Results There was a significant decrease in rate of recalls in the first year of the intervention, sustained over the second year and, for the smaller number of officers still available, also over the third year. Non‐compliance with supervision as a reason for recall was cut by two‐thirds; other reductions were in responses to ‘challenging’ behaviours. There was no evidence of increase in serious further offending. Conclusions Our findings provide preliminary evidence that psychologically informed practice by probation officers supervising offenders at highest risk of serious harm re‐offending can reduce the rate of recall to prison, a costly alternative to maintaining offenders in the community, without apparently reducing community safety. Further research should clarify the extent to which the intervention helped the officers merely to tolerate challenging behaviours from these offenders and the extent of real change in the behaviours. Copyright © 2015 John Wiley & Sons, Ltd.