BACKGROUND: Meta-analyses support an association between anxiety in older adulthood and dementia. The aim of this study was to use routinely collected health data to test whether treatment of anxiety disorders through psychological intervention is associated with a lower incidence of dementia. METHODS: In this prospective cohort study, data from nationally provided psychological therapy services in England termed Improving Access to Psychological Therapies from 2012 to 2019 were linked to medical records, including dementia diagnoses as defined by the tenth edition of the International Classification of Diseases, up to 8 follow-up years later. Inclusion criteria were as follows: (1) patients who were aged 65 years and older; (2) patients with a probable anxiety disorder; and (3) those with no previous or current diagnosis of dementia. Cox proportional hazards models were constructed to test whether reliable improvement in anxiety following psychological intervention was associated with future dementia incidence. The primary outcome was all-cause dementia and cases were identified using ICD-10 dementia codes from Hospital Episode Statistics, Mental Health Services Dataset, and mortality data. For main analyses, hazards ratios (HRs) are presented. FINDINGS: Data from 128 077 people aged 65 years and older attending a nationally provided psychological intervention service in England were linked to medical records. 88 019 (69·0%) of 127 064 participants with available gender data were women and 39 585 (31·0%) were men. 111 225 (95·9%) of 115 989 with available ethnicity data were of White ethnicity. The mean age of the sample was 71·55 years (SD 5·69). Fully adjusted models included data from 111 958 people after 16 119 were excluded due to missing data on key variables or covariates. 4510 (4·0%) of 111 958 participants had a dementia diagnosis. The remaining 107 448 (96·0%) were censored either at date of death or when the final follow-up period available for analyses was reached. People who showed reliable improvement in anxiety had lower rates of later dementia diagnosis (3·9%) than those who did not show reliable improvement (5·1%). Reliable improvement in anxiety following psychological intervention was associated with reduced incidence of all-cause dementia (HR 0·83 [95% CI 0·78-0·88]), Alzheimer's disease (HR 0·85 [0·77-0·94]), and vascular dementia (HR 0·80 [0·71-0·90]). Effects did not differ depending on anxiety disorder diagnosis. INTERPRETATION: Results showed that reliable improvement in anxiety from psychological therapy was associated with reduced incidence of future dementia. There are multiple plausible explanations for this finding and further research is needed to distinguish between these possibilities. Missing data in the sample limit reliability of findings. FUNDING: Alzheimer's Society, Medical Research Council, Wellcome Trust, and UCLH National Institute for Health and Care Research Biomedical Research Centre.