All in the Name of Artificial Intelligence: A Commentary on Linardon (2025)

Artificial Intelligence (AI) is being rapidly integrated into healthcare, but Linardon et al. reveal a troubling gap between what AI actually is, its capabilities, and the patients' and clinicians' perceptions of it—equating AI solely with large language models. In this commentary, we discuss concerns over AI's black-box nature, its potential to perpetuate existing biases, and the blind trust some people place in its decisions, despite evidence that quantitative models outperform large language models in clinical decision-making tasks. While AI holds promise in eating disorder care, its integration requires a nuanced understanding of its capabilities, limitations, and the critical distinction between AI for administrative automation, clinical decision-making, and direct-to-patient AI. Poorly designed AI alerts risk becoming just another ignorable nuisance, while patient-facing AI could either empower individuals or drown them in notifications and misinformation. Before we anoint AI as healthcare's savior, it requires validation for accuracy, reliability, fairness, real-world usability, and its actual measurable impact on clinicians and patients. The real challenge is not whether AI will change healthcare but ensuring it does so responsibly—by integrating it thoughtfully into workflows, such that it is supporting rather than replacing clinical judgment, and maintaining accountability when things go wrong.