Objective The advent of second‐generation antipsychotics (SGAs) in the 1990s brought optimism that neuroleptic‐induced tardive dyskinesia (TD) may become relegated to history. Whether or not this is the case remains inconclusive, and this review aims to compare the risk of TD in older adults treated with first‐generation antipsychotics (FGAs) versus SGAs. Methods Relevant papers were sourced via a range of electronic databases, with a date range from 1957 to January 2015. Included studies used both a validated rating scale and research diagnostic criteria to report on the prevalence or incidence of TD in older adults exposed to antipsychotic medications. Results For FGAs, the prevalence estimate was 53% (95% confidence interval [CI] [39.0, 68.4]) for mild TD and 38% (95% CI [25.9, 50.3]) for probable TD. Incidence estimates for probable TD with FGAs were 23% (95% CI [15.3, 30.6]) at 1 year, 42% (95% CI [24.8, 58.4]) at 2 years and 57% (95% CI [45.3, 69.1]) at 3 years. For SGAs, the incidence estimates at 1 year were 7% (95% CI [4.4, 10.2]) for probable TD and 3% (95% CI [1.5, 4.2]) for persistent TD. Conclusions The risk of probable TD is more than three times lower in older adults receiving SGAs in comparison with FGAs after 1 year of treatment (23% vs 7%). The risk of persistent TD at 1 year with SGAs is particularly low. Evidence is lacking in regard to the longer‐term risk of TD with SGAs, although the rates associated with the prolonged use of FGAs are high. Caution is therefore still required, particularly with the protracted use of both FGAs and SGAs.