The study addresses the persistent spelling difficulties faced by EFL learners, difficulties rooted in English orthography’s weak sound–spelling correspondence and compounded by limited classroom focus on writing skills. Noting the growing use of AI large language models (LLMs) in tasks such as text generation, machine translation, and long-text summarization, the researchers set out to apply a Generative Pre-training Transformer (LLM-GPT) to writing instruction. The intervention was designed to supply automated feedback, spelling assistance, and frequent practice while also gauging learners’ attitudes toward this reinforcement approach. Sixty EFL students participated in a between-subjects experiment, with control and experimental groups receiving traditional instruction and LLM-GPT-enhanced instruction, respectively.
Results showed that the experimental-group learners who used the LLM-GPT application outperformed their control-group peers and retained word spellings more readily on the post-test. Survey responses further indicated that participants perceived the LLM-GPT tool as having a positive influence on mastering the spelling of English words.