Objectives: Postulating that efficacy of antipsychotics for agitation and psychosis in dementia is best estimated in trials among patients with these symptoms and with symptom-specific outcomes, we investigated whether clinically broader definitions affected the pooled efficacy. Study Design and Setting: Trials were searched in multiple databases and categorized according to patient population (agitated, psychotic, and mixed) and outcome scale (agitation, psychosis, and generic). Standardized mean differences with 95% confidence intervals were calculated for conventional and atypical antipsychotics separately. Results: Thirty trials met our inclusion criteria. Conventional antipsychotics might have a small effect in agitated patients on agitation scales (−0.44, −0.88, 0.01) and in psychotic patients on psychosis scales (−0.31, −0.61, −0.02). There was no effect on generic scales. Efficacy of atypical antipsychotics was not established in agitated patients on agitation scales (−0.15, −0.43, 0.13) and in psychotic patients on psychosis scales (−0.11, −0.20, −0.03) but was small in mixed patients on agitation scales (−0.29, −0.40, −0.18). Conclusion: Pooled efficacy of antipsychotics for agitation and psychosis in dementia is biased when based on trials that included patients without these target symptoms or on results measured with generic scales. This finding is important for reviewers and guideline developers who select trials for reviews.