Abstract Visual impairment has emerged as a potential modifiable risk factor for dementia. However, there is a lack of large studies with objective measures of vision and with more than 10 years of follow-up. We investigated whether visual impairment is associated with an increased risk of incident dementia in UK Biobank and European Prospective Investigation into Cancer in Norfolk (EPIC-Norfolk). In both cohorts, visual acuity was measured using a “logarithm of the minimum angle of resolution” (LogMAR) chart and categorized as no (≤0.30 LogMAR), mild (>0.3 to ≤0.50 LogMAR), and moderate to severe (>0.50 LogMAR) impairment. Dementia was ascertained through linkage to electronic medical records. After restricting to those aged ≥60 years, without prevalent dementia and with eye measures available, the analytic samples consisted of 62 206 UK Biobank and 7 337 EPIC-Norfolk participants, respectively. In UK Biobank and EPIC-Norfolk, respectively, 1 113 and 517 participants developed dementia over 11 and 15 years of follow-up. Using multivariable Cox proportional-hazards models, the hazard ratios for mild and moderate to severe visual impairment were 1.26 (95% confidence interval [CI]: 0.92–1.72) and 2.16 (95% CI: 1.37–3.40), in UK Biobank, and 1.05 (95% CI: 0.72–1.53) and 1.93 (95% CI: 1.05–3.56) in EPIC-Norfolk, compared to no visual impairment. When excluding participants censored within 5 years of follow-up or with prevalent poor or fair self-reported health, the direction of the associations remained similar for moderate impairment but was not statistically significant. Our findings suggest visual impairment might be a promising target for dementia prevention; however, the possibility of reverse causation cannot be excluded.