Accurately identifying periodontal landmarks via acoustic imaging is increasingly important. Here, we evaluated the accuracy of cementoenamel junction (CEJ) identification using ultrasound by comparing it to clinical methods in 153 extracted human teeth. The distance between the CEJ to a reference point was measured using two clinical methods (visual examination and tactile sensation) as well as ultrasound imaging. Statistical analyses were performed across all teeth and sub-groups, including tooth types (incisors, cuspids, and molars/premolars), and two classifications: A- vs. B- (visually detectable or undetectable CEJ, respectively); and CL-S vs. CL-D (shallow or deep cervical lesions). In A- or CL-S teeth, ultrasound measurements highly agreed with clinical measurements, showing a 1.72-mm 95% CI for A- and 1.99-mm 95% CI for CL-S compared to visual examination, and a 1.77-mm 95% CI for A- and a 2.10-mm 95% CI for CL-S compared to tactile sensation, respectively. For 80% of A- and 76% of CL-S teeth, the difference between ultrasound and visual examination was within ± 20%. For 81% of A- and 80% of CL-S teeth, the difference between ultrasound and tactile sensation was within ± 20%. The variance of ultrasound versus clinical CEJ identifications showed a significant correlation (r = 0.6607) to the cervical lesion depth. The errors between ultrasound and clinical measurements show no significant bias across different tooth types.