Fish harbour many types of nutrients that are beneficial for preventing cognitive decline. Therefore, habitual fish intake might contribute to a lower risk of incident dementia. However, few prospective cohort studies have investigated fish consumption in relation to incident dementia, and their findings have been inconsistent. To investigate the association between fish consumption and the risk of incident dementia, we collected data on the consumption of fish and other foods using an FFQ in a baseline survey of individuals aged ≥65 years living in Ohsaki City, Japan. After 5·7 years of follow-up, the incidence of dementia was 1118 (8·5 %) among 13 102 participants. We then used a multivariate-adjusted Cox model to estimate hazard ratios (HR) and 95 % CI. Compared with subjects with the lowest fish intake (Q1), the multivariate HR were 0·90 (95 % CI 0·74, 1·11) for Q2, 0·85 (95 % CI 0·73, 0·99) for Q3 and 0·84 (95 % CI 0·71, 0·997) for Q4 (P trend = 0·029). Such associations were also observed even after excluding participants who were diagnosed with dementia in the first 2 years of follow-up and those who had poorer cognitive function at baseline. In conclusion, an association was observed between higher fish consumption and a lower risk of incident dementia among healthy elderly people without disability. These findings suggest that habitual fish intake may be beneficial for the prevention of dementia.