Standard hospital MRI images can now be used to estimate how fast a person is biologically ageing, according to a study published in Nature Aging. Researchers from Duke University, the University of Otago and collaborators trained a machine-learning model—dubbed the Dunedin Pace of Aging Calculated from NeuroImaging (DunedinPACNI)—on scans and long-term health data from the Dunedin birth cohort, then applied it to 50,106 MRI scans drawn from the UK Biobank, Alzheimer’s Disease Neuroimaging Initiative, BrainLat and other datasets covering adults aged 22–98. People whose single-scan DunedinPACNI score indicated faster ageing were more likely to show thinner cortex, lower grey-matter volume and accelerated hippocampal atrophy. In prospective analyses, cognitively-normal adults in the fastest-ageing decile faced a 61 % higher risk of progressing to mild cognitive impairment or dementia (hazard ratio 1.49) and at least a 41 % greater risk of death over the follow-up period than peers with average scores. Faster scores also tracked with frailty, multimorbidity and lower household income, while the metric’s test–retest reliability reached 0.94 in the Human Connectome Project. The brain-based ageing ‘clock’ performed as well as or better than existing MRI age-gap and volumetric measures and, because it relies on routinely collected scans, could help screen participants for anti-ageing trials or flag patients at elevated dementia risk. The authors caution that DunedinPACNI still requires validation in more diverse populations before being adopted in clinics.