This systematic review aimed to assess the effectiveness and safety of herbal medicines (HM) for treating dementia. Databases in English and Chinese were searched from their inceptions to February 2007. References in reviews and randomized controlled trials (RCTs) were screened by hand. Trials comparing orally administered HM with placebo, no intervention or other therapy were considered. Trials on Ginkgo biloba and its extracts were excluded to avoid duplication of existing reviews. Pairs of authors independently applied eligibility criteria, extracted data and assessed methodological quality using the Jadad Scale. Thirteen RCTs met the inclusion criteria of three or above on this scale. Six trials compared herbal medicine with placebo, one with no treatment, and the remainder with pharmaceutical intervention. Meta-analyses were performed on common cognitive performance outcome measures. All studies reported HM had significant effects in improving symptoms. In studies that employed active controls, HM was at least as effective as the pharmaceutical intervention. Meta-analyses found HM more effective than no treatment or placebo and at least equivalent to control interventions, although the overall effect was small. No severe adverse events were reported. These trials provide overall positive evidence for the effectiveness and safety of certain HMs for dementia management.