Generative Artificial Intelligence as a Learning Feedback Tool: An Integrative Review of Evidence, Opportunities, and Challenges

Abstract

The incorporation of Generative Artificial Intelligence (GenAI) into education has progressively transformed teaching, learning, and assessment practices, particularly in activities requiring continuous interaction, product revision, and formative feedback. In this context, large language models have gained relevance because of their ability to analyze academic productions and provide personalized responses within short periods of time. This article aimed to critically analyze the available evidence on the use of Generative Artificial Intelligence as a learning feedback tool, with emphasis on its effects on academic performance, product revision, self-regulation, and feedback quality. An integrative review of recent scientific literature was conducted, complemented by systematic reviews, meta-analyses, and international reports addressing the educational use of generative AI. The synthesis shows generally favorable but heterogeneous results. Recent research indicates that GenAI can contribute to improving revision quality, linguistic accuracy, writing productivity, engagement, and selected learning outcomes. A meta-analysis of 68 studies reported a moderate overall positive effect of GenAI on learning outcomes (SMD = 0.45; 95% CI [0.43, 0.47]), accompanied by substantial heterogeneity (I² = 95%). At the same time, reviews focused specifically on large language model-based feedback indicate that AI performs particularly well in identifying errors and structural aspects, whereas teacher feedback retains advantages for argument development, contextual interpretation, prioritization of revision needs, and dialogic guidance. It is concluded that GenAI has greater educational value when used as a complementary mechanism within a hybrid feedback model, in which technology contributes immediacy, availability, and iterative interaction, while teachers retain responsibility for contextualizing, validating, and deepening the learning process.