Objective This study aims to evaluate the impact of suboptimal treatment, defined in terms of lower population coverage (percentage of total patient population receiving optimal treatment) and delay to treatment on the cost‐effectiveness of pharmacological therapies approved for the treatment of different severities of Alzheimer's disease (AD) in the UK. Methods A 5‐year Markov model was used to simulate transition to full‐time care, as delay and coverage were varied for AD patients with mild‐to‐moderate and moderate‐to‐severe dementia. The time‐varying predictive equations, resource use, utilities, treatment effects and mortality were derived using published sources. Results For the cohort with moderate‐to‐severe dementia, cost‐effectiveness was optimised when delay was minimised and coverage maximised. For mild‐to‐moderate dementia, results were similar but varied widely depending on the inputted cost of acetylcholinesterase inhibitors. Conclusions The average cost‐effectiveness of pharmacological treatments for AD is sensitive to delays to treatment and population coverage. The results of this study can inform future healthcare policy in order to maximise cost‐effectiveness of pharmacological therapies for AD. Copyright © 2017 John Wiley & Sons, Ltd.