Ocular tuberculosis has traditionally been considered uncommon or anecdotal. Imprecise and variable diagnostic criteria have contributed to the confusion surrounding this topic. The increase in extrapulmonary manifestations of tuberculosis during the AIDS era established the need for a prospective study of ocular involvement in patients with all types of tuberculosis using well-defined criteria. During a 15-month period, 300 cases had culture-proven tuberculosis at our institution. We randomly selected 100 for systematic ophthalmologic evaluation. Our criteria for ocular tuberculosis were divided as follows: certainty (isolation of Mycobacterium tuberculosis from ocular specimens), probability (patients with isolation of M. tuberculosis from extraocular samples, with ocular lesions not attributable to other causes that respond to anti-tuberculous treatment), and possibility (same as probability but follow-up impossible). Ocular tuberculosis was present in 18 patients (18%) of which 10 patients fulfilled probability and 8 patients fulfilled possibility criteria. Eleven of 18 patients had HIV infection. In 11 patients, ocular involvement was asymptomatic. Almost all patients (17/18) had choroiditis, and other ocular lesions included papillitis, retinitis, vitritis, vasculitis, dacryoadenitis, and scleritis. Multivariate analysis showed as risk factors independently predicting ocular involvement in patients with ocular tuberculosis the presence of miliary disease (odd ratio 43.92, p = 0.002), ocular symptoms (odds ratio 6.35, p = 0.0143), and decreased visual acuity (odds ratio 0.04, p = 0.012). We observed an unexpectedly high (18%) incidence of ocular involvement, frequently asymptomatic, in patients with tuberculosis. Miliary disease is a clear predisposing factor in both HIV-infected and noninfected populations. Ocular examination should be routinely considered in patients with proven or suspected tuberculosis.