Advancing brain-computer interfaces with generative AI: A review of state-of-the-art and future outlook

Brain-Computer Interface (BCI) technology is rapidly emerging as a promising tool to empower individuals with severe disabilities and enhance their independence by translating brain neural signals into actionable commands. However, its development and application face challenges such as low signal-to-noise ratios, overfitting from limited training data, and the non-stationarity of brain signals, which can compromise system stability. The integration of Generative Artificial Intelligence (Generative AI, GAI) offers potential solutions by improving signal processing, generating high-fidelity synthetic data, and developing adaptive algorithms that maintain accuracy over time. Despite these advancements, existing literature lacks systematic discussion on the comprehensive integration of GAI in BCI development. To address this gap, this study examines over 170 articles published from 2020 to 2025 that leverage GAI techniques in BCI research. The analysis highlights the latest developments in techniques such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Transformers, Diffusion Models (DMs) and their hybrid models. It systematically examines the applications of artificial intelligence across various stages of BCI development, proposes an AI-driven future application framework tailored to BCI needs, and highlights the significant potential of GAI on the field. This review provides insights and a systematic overview to guide future research and applications in this interdisciplinary domain.