Three hundred eighty-eight human immunodeficiency virus (HIV)-negative clients in Zimbabwe were retested at 3 months using 2 parallel rapid tests. One operator error (risk, 0.26%; 95% confidence interval, 0.0065%-1.4%) and no "true" seroconversions (upper 95% confidence limit, 0.96%) were detected. High-risk behavior was not significantly reduced. Policies recommending routine retesting need to be reconsidered.