Abstract

Background: The period between 2024 and 2026 represents a transformative era for Artificial Intelligence (AI) in medicine, characterized by the shift from narrow, task-specific algorithms to Multimodal Large Language Models (M-LLMs) and Generalist Medical AI (GMAI). While technical capabilities have advanced exponentially, the clinical integration of these tools remains complex and unevenly distributed.
Objective: This literature review synthesizes the latest research (2024-2026) to evaluate the state of AI in clinical practice, focusing on diagnostic accuracy, administrative efficiency through ambient intelligence, and the ethical-legal frameworks governing autonomous systems.
Methods: A literally review of the higher-impact articles published since January 2024 was conducted. Sources included PubMed, JMIR, and medRxiv, with a focus on bibliometric analyses, clinical trials of generative AI, and regulatory policy documents.
Result: Current literature indicates that generative AI has achieved “specialist-level” performance in written clinical reasoning and diagnostic imaging across multiple specialties, including radiology and pathology. “Ambient AI scribes” have demonstrated a significant reduction in physician burnout (approximately 40%) by automating EHR documentation. However, a significant “maturity gap” persists: despite the volume of publications, fewer than 1% of models have reached the stage of randomized controlled trials. Emerging research highlights critical concerns regarding “automation bias,” algorithmic transparency (the “black box” problem), and the potential for AI to exacerbate health inequities in lowresource settings.
Discussion: The discourse has evolved from pure technical validation to implementation science. Key themes include the necessity of “Human-in-the-Loop” systems, the impact of the EU AI Act on medical software development, and the urgent need for a radical restructuring of medical education to include AI literacy.
Conclusions: AI has transitioned from an experimental adjunct to a foundational infrastructure in modern healthcare. Future success depends not on increasing model parameters, but on achieving seamless interoperability, ensuring algorithmic fairness and establishing clear liability frameworks for autonomous medical decision-making.

Keywords:Artificial intelligence; Genomics; Clinical data; Well-being; Healthcare