Background: In the UK, it is recommended by the National Institute for Health and Care Excellence (NICE) that if antipsychotics are initiated in people living with dementia, treatment should be at the lowest dose for the shortest time possible (1–3 months). In this study, we aimed to investigate how dose and duration of antipsychotic medication adhere to UK clinical guidelines and explore treatment restart details in those who stop treatment. Methods: We did a retrospective cohort study using longitudinal UK primary care data from the IQVIA Medical Research Database. We included people living with dementia aged 60–85 years who received their first antipsychotic prescription between Jan 1, 2000, and Dec 31, 2023. Individuals with any previous antipsychotic prescriptions in their records more than 1 year before a dementia diagnosis and those who had missing social deprivation information were excluded from the study. Duration of first and subsequent antipsychotic treatment episodes, medication dosage, and treatment discontinuation and reinitiation rates were investigated. Duration and discontinuation were defined by grouping consecutive prescriptions into treatment episodes using the waiting time distribution method (80% inter-arrival density, 59 days). Daily doses were derived from strength and frequency information and categorised as low or moderate or high based on established minimum effective dose equivalences. People with lived experience of dementia care contributed throughout this project, shaping the research question and advising on interpretation and dissemination strategies. Findings: In our dataset search, we identified 108 910 people with a record indicating dementia at any time. In total, 99 091 cases were excluded (ie, individuals with no antipsychotic prescription between the ages of 60 and 85 years between 2000 and 2023, a previous history of antipsychotics, missing deprivation information, or only one eligible prescription). We included 9819 people living with dementia aged 60–85 years who received their first antipsychotic prescription between 2000 and 2023 in the study. 5310 (54·1%) were female and 4509 (45·9%) were male, with a mean age of 77·1 years (SD 5·6 years), and ethnicity data were not available. The first treatment episode lasted a median of 7 months (IQR 6·6–8·7), exceeding NICE guidelines of 1–3 months and 18·1% [95% CI 17·4–18·9]) were initiated on a prescription above the minimum effective dose (ie, low dose). Of the 1781 participants who started on a moderate or high dose, 519 (29·1%) had a moderate or high dose in all quarters of the first year of treatment. 1 year after treatment initiation, 5136 (78·3%) of 6559 eligible individuals remained on medication (48·9% [95% CI 47·7–50·1] on low dose, 14·8% [13·9–15·6] on moderate or high dose of haloperidol, olanzapine, quetiapine or risperidone; and 14·6% [13·8–15·5] on other antipsychotics). Of the 5547 individuals eligible to restart treatment after initial discontinuation, 3106 (56%) restarted with a median treatment duration of 2·6 months (IQR 0·0–9·9). Interpretation: This study highlights how antipsychotic prescribing in dementia is discordant with current NICE guidelines on both duration and dose. More than half of those who discontinued their treatment then restarted treatment. These findings emphasise a persistent gap between clinical guidelines and real-world prescribing, underscoring the need for interventions that prioritise safety and person-centred dementia care. Funding: National Institute for Health and Care Research School for Primary Care Research.