Objective To explore the diagnostic value of artificial intelligence assisted score (AIAS) combined with ultrasonic shear wave elastography (SWE) in benign and malignant thyroid nodules. Methods A total of 110 patients with thyroid nodules (152 nodules) admitted to XD Group Hospital were enrolled between January 2021 and January 2023. According to random number table method, they were divided into test group (77 cases, 109 nodules) and verification group (33 cases, 43 nodules). All patients underwent pathological examination, ultrasound examination and ultrasonic SWE, and AIAS was analyzed. The general data, pathological diagnosis results, ratio of strain rate (SR) and AIAS were compared between the two groups. The models of ultrasonic SWE parameters, AIAS and combined detection were constructed, and predictive efficiency of the above models for benign and malignant thyroid nodules was evaluated by receiver operating characteristic (ROC) curves. The differences in predictive efficiency of different models were evaluated by Delong test. The fit of the combined model was detected by Hosmer-Lemeshow test. Results In the 109 nodules from test group, operation or biopsy results showed that there were 45 benign nodules (41.28%) and 64 malignant nodules (58.72%). In the 43 nodules from verification group, operation or biopsy results showed that there were 17 benign nodules (39.53%) and 26 malignant nodules (60.47%). In the two groups, SR and AIAS in patients with malignant thyroid nodules were significantly higher than those with benign nodules (P<0.05). In patients with benign or malignant thyroid nodules, there were no significant differences in SR or AIAS between test group and verification group (P>0.05). ROC curves analysis showed that AUC, sensitivity and specificity of SR combined with AIAS in test group for predicting malignant thyroid nodules in verification group were 0.917, 82.81% and 91.11%, respectively (P<0.05). In verification group, diagnostic efficiency of ultrasonic SWE parameters combined with AIAS was superior to that of single detection (P<0.05). DeLong test showed that AUC of combined model (AIAS + ultrasonic SWE parameters) for predicting malignant thyroid nodules was the greatest (P<0.05). The fit of the combined model was detected by Hosmer-Lemeshow test: Hosmer-Lemeshow χ2^{=7.692 (}P>0.05). Conclusion The combined model constructed by AIAS + ultrasonic SWE parameters has high predictive value for malignant thyroid nodules, high calibration ability and clinical application value.