Mapping the landscape of AI-driven feedback in education: a scoping review

作者：A. Lipnevich, C. Taranto, Christopher DeLuca, Ephraim Nukpetsi, Nathan Rickey, Therese N. Hopfenbeck, Emma Carter, J. McGrane · 发表于：Frontiers in Education · 年份：2026 · DOI：10.3389/feduc.2026.1799346

Artificial intelligence (AI) is increasingly used in educational contexts to support learning, assessment, and instructional decision-making, particularly through automated feedback. While AI-generated feedback offers scalability, the rapid growth of this work has produced a fragmented empirical literature that warrants a scoping review. We conducted a scoping review following the PRISMA guidelines. A systematic search of ProQuest Dissertations & Theses Global and Web of Science yielded 32,580 records after deduplication, which were screened using ASReview, an AI-assisted screening tool. Following full-text review, 104 empirical studies published between 2008 and 2024 met inclusion criteria. Data were extracted on study characteristics, methodological approaches, AI tools, demographics, and key findings across K–12 and higher education settings. The reviewed literature was dominated by experimental ( n  = 49) and mixed-methods designs ( n  = 44), frequently incorporating content analysis to evaluate the nature and quality of AI-generated feedback. A substantial proportion of studies ( n  = 55) originated from Asian countries, with China ( n  = 30) as the leading contributor, followed by the US ( n  = 15). ChatGPT emerged as the most commonly examined AI tool ( n  = 41), followed by bespoke AI systems ( n  = 22). Findings revealed considerable variability in students’ and educators’ perceptions of AI-generated feedback; however, learning outcome e...