Abstract
<jats:title>Abstract</jats:title> <jats:p>Owing to the rapid advancement of artificial intelligence (AI) and machine learning (ML) technologies, computer scoring of spoken responses (CSSR) has become increasingly used in internet‐based language tests and language learning activities. Compared to a decade ago, automated speech evaluation technologies have improved in many areas such as automated speech recognition (ASR) for spontaneous non‐native learner speech, spoken grammatical error detection, aberrant response detection, and AI confidence indication for score prediction. Generally, AI has been more accepted by both the public and language assessment professionals, as a powerful tool for improving the efficiency and cost‐effectiveness of speaking assessment. Meanwhile, CSSR is faced with new issues and challenges because of the use of innovative techniques such as deep learning models and large language models (LLMs). Whichever CSSR approach is adopted, debates will continue to be ongoing as to whether computer‐scored speaking tests underrepresent the oral proficiency construct.</jats:p>