The benefits of combination antiretroviral therapy (ART) for HIV cognitive disorders vary substantially between individuals. This study evaluated whether cerebrospinal fluid (CSF) drug penetration and CSF virological suppression influence the extent of neuropsychological (NP) improvement during ART. Overall performance on a battery of NP tests administered at baseline and follow‐up (median 15 weeks) was computed by using the global deficit score (GDS) methods in 31 cognitively impaired, HIV‐infected individuals who began new ART regimens. Virological suppression (attaining undetectable viral load by RT‐PCR at follow‐up) was assessed separately for plasma and CSF. Subjects on regimens containing greater numbers of CSF‐penetrating drugs showed significantly greater reduction in CSF viral load. Subjects attaining CSF virological suppression demonstrated greater GDS improvement than those who did not (median GDS change, 0.62 vs 0.23; p = 0.01). A similar trend for plasma did not reach statistical significance (p = 0.053). NP improvement was greater in ART‐naive versus treatment‐experienced subjects. In a multivariate model (overall p = 0.0008), significant, independent predictors of GDS reduction were CSF HIV RNA suppression, baseline antiretroviral history, and their interaction. Including CSF‐penetrating drugs in the ART regimen and monitoring CSF viral load may be indicated for individuals with HIV‐associated cognitive impairment Ann Neurol 2004.