Previous systematic reviews that examined whether atrial fibrillation (AF) is associated with dementia have relied on different study designs (including retrospective ones) and did not evaluate risk using meta-analysis. We searched Medline, Embase, and PsychINFO in September 2010 for published prospective studies reporting on the association between baseline AF and incident dementia. Pooled odds ratios for AF and dementia were calculated using the random effects model, with heterogeneity assessed using I(2). We identified 15 relevant studies covering 46,637 participants, mean age 71.7 years. One study that reported no significant difference in Mini-Mental State Examination scores between patients with or without AF could not be pooled. Meta-analysis of the remaining 14 studies showed that AF was associated with a significant increase in dementia overall (odds ratio [OR] 2.0, 95% confidence interval [CI] 1.4 to 2.7, p < 0.0001), with substantial heterogeneity (I(2) = 75%). When stratified by participants, the association was significant (with little heterogeneity) in studies focusing solely on patients with stroke (7 studies, OR 2.4, 95% CI 1.7 to 3.5, p < 0.001, I(2) = 10%), and of borderline significance (with substantial heterogeneity) for studies in broader populations (7 studies, OR 1.6, 95% CI 1.0 to 2.7, p = 0.05, I(2) = 87%). For conversion of mild cognitive impairment to dementia, one study showed a significant association with AF (OR 4.6, 95% CI 1.7 to 12.5). There is consistent evidence supporting an association between AF and increased incidence of dementia in patients with stroke whereas there remains considerable uncertainty about any link in the broader population. The potential association between AF and incident dementia in mild cognitive impairment merits further investigation.