Capsule endoscopy (CE) is a promising diagnostic tool for the study of patients with obscure gastrointestinal bleeding. However, the diagnostic yield of this technique has not been adequately studied. We evaluated sensitivity and specificity of CE and the outcome after CE in patients with obscure gastrointestinal bleeding. One hundred consecutive patients (all with recent negative upper and lower endoscopy; 26 with ongoing overt bleeding [group A], 31 with previous overt bleeding [group B], and 43 with guaiac-positive stools and iron-deficiency anemia [group C]) underwent CE. The yield of positive findings on CE was 92.3% in group A, 12.9% in group B, and 44.2% in group C (P < 0.0001, A vs. B, A vs. C). Angiodysplasia (29%) and Crohn's disease (6%) were the most common diagnoses. Sensitivity, specificity, and positive and negative predictive values of CE were 88.9%, 95%, 97%, and 82.6%, respectively. CE results led to treatments resolving the bleeding in 86.9% of patients undergoing the procedure while actively bleeding. Capsule retention because of unsuspected stenosis occurred in 5 patients and required surgery, which resolved the clinical problem, in 4 patients. CE is an effective diagnostic tool for patients with obscure GI bleeding. The best candidates for the procedure are those with ongoing obscure-overt bleeding or with obscure-occult bleeding. If done early in the course of the workup, CE could shorten considerably the time to diagnosis, lead to definitive treatment in a relevant proportion of patients, and spare a number of alternative investigations with low diagnostic yield.