Objective Although available evidence is modest, exercise could be beneficial in reducing behavioural and psychological symptoms of dementia. We aim to evaluate the cost‐effectiveness of a dyadic exercise regimen for individuals with dementia and their main carer as therapy for behavioural and psychological symptoms of dementia. Methods Cost‐effectiveness analysis within a two‐arm, pragmatic, randomised, controlled, single‐blind, parallel‐group trial of a dyadic exercise regimen (individually tailored, for 20–30 min at least five times per week). The study randomised 131 community‐dwelling individuals with dementia and clinically significant behavioural and psychological symptoms with a carer willing and able to participate in the exercise regimen; 52 dyads provided sufficient cost data for analyses. Results Mean intervention cost was £284 per dyad. For the subsample of 52 dyads, the intervention group had significantly higher mean cost from a societal perspective (mean difference £2728.60, p = 0.05), but costs were not significantly different from a health and social care perspective. The exercise intervention was more cost‐effective than treatment as usual from both societal and health and social care perspectives for the measure of behavioural and psychological symptoms (Neuropsychiatric Inventory). It does not appear cost‐effective in terms of cost per quality‐adjusted life year gain. Conclusions The exercise intervention has the potential to be seen as cost‐effective when considering behavioural and psychological symptoms but did not appear cost‐effective when considering quality‐adjusted life year gains. Copyright © 2011 John Wiley & Sons, Ltd.