With the widespread application of generative artificial intelligence in the field of language teaching, second language writing instruction is gradually transitioning toward digitalization and intelligentization. Traditional input enhancement, which relies on manual operations, is monotonous in form and hardly adaptable to the individual differences of learners; it can no longer meet the dynamic needs of second language acquisition in the digital age. Thus, exploring an AI-driven input enhancement model has become a key issue for optimizing L2 writing instruction. This research focuses on the impact of AI input enhancement on students’ English writing. As a digital extension of traditional input enhancement, it relies on tools like ChatGPT, Grammarly, and Bing tools, and is theoretically grounded in second language acquisition theories such as the Input Hypothesis, Noticing Hypothesis, and multimodal learning theory. Guided by these theories, it strengthens target language salience through real-time feedback and integrates multimodal resources (text, images, etc.) to construct immersive learning environments, laying a theoretical basis for subsequent research and practice.​ To validate this approach, existing studies validate it with three paradigms (theoretical speculation, empirical research, mixed methods), each with pros and cons. In practice, AI input enhancement boosts writing efficiency and student confidence while adapting to diverse teaching scenarios. However, it also brings problems such as students’ over-reliance on technology, impaired writing originality, controversies over educational equity, insufficient contextual adaptability of tool feedback, and a lack of guidance for higher-order thinking.​ To address these issues, future efforts should explore differentiated tool-use rules, optimize tool contextual adaptability, and establish an “AI tools+teacher guidance” collaborative model to balance technology empowerment and student autonomy in L2 writing teaching.