The purpose of the study was to evaluate the value of high-frequency ultrasound (HFUS) for differentiating invasive basal cell carcinomas (BCCs) from non-invasive BCCs. We established a prediction model based on ultrasound features and validated it further. One hundred patients in the pilot cohort and another 43 in the validation cohort were evaluated. All patients underwent HFUS examinations by the same radiologist, and then were divided on the basis of pathology into invasive and non-invasive types. With respect to growth pattern, 60.5% of invasive BCCs had an irregular pattern, whereas 89.5% of non-invasive BCCs had a nodular or crawling pattern (p < 0.001). As for the layers involved, the more invasive BCCs broke through the dermis compared with non-invasive BCCs (23.3% vs. 1.8%) (p < 0.001). With respect to intralesional hyperechoic spot distribution, invasive and non-invasive BCCs tended to be clustered and absent/scattered-like, respectively (55.8% vs. 91.2%) (p < 0.001). On the basis of the aforementioned features, a prediction model was established with accuracies of 84.0% and 76.7%, respectively, in the pilot and validation cohorts. HFUS holds promise for the differentiation of the invasiveness of BCCs and is helpful in its clinical management.