Liver biopsy is the gold standard method to diagnose nonalcoholic fatty liver disease (NAFLD). However, ultrasound is widely recommended as the first-line imaging test for individuals with suspected NAFLD. This study aimed to estimate the accuracy of ultrasound as a screening test for NAFLD compared to liver biopsy in a cohort of patients with class II and III obesity undergoing bariatric surgery. This retrospective study included patients undergoing Roux-en-Y gastric bypass in southern Brazil between 2010 and 2019 who were screened for NAFLD with both ultrasound and liver biopsy. All samples were collected by a core biopsy needle and were analyzed by the same pathologist. Sensitivity, specificity, and positive and negative predictive values of ultrasound were estimated. The final database included 227 patients, mostly female (84%) and white (83.6%), with a mean age of 42.5 ± 10.2 years and a mean preoperative body mass index of 49.5 ± 8.4 kg/m². A total of 153 subjects (67.4%) were diagnosed with NAFLD through liver biopsies: 41 (18%) had fatty liver and 112 (49.3%) had nonalcoholic steatohepatitis. Ultrasound sensitivity was 88.9% and specificity was 44.6%. Positive and negative predictive values were 76.8% and 66.0%, respectively. Positive likelihood ratio was 1.6 (95% CI 1.30–1.98), and negative likelihood ratio was 0.25 (95% CI 0.15–0.42). Therefore, approximately three every four subjects with an ultrasound suggesting NAFLD were true positives. Ultrasound showed a good sensitivity in detecting NAFLD in patients with class II and III obesity.