Eric Topol is a cardiologist and the founder and director of the Scripps Research Translational Institute. He is the author of Deep Medicine (2019) and Super Agers (2024), and writes a long-running Substack, Ground Truths, on AI in medicine. He has argued that there is little empirical evidence that large language models improve patient health outcomes outside administrative tasks.
Background and roles
Topol is a cardiologist and founder and director of the Scripps Research Translational Institute. His books include Deep Medicine (2019) and Super Agers (2024), and he publishes the Ground Truths Substack on AI in medicine.
Positions on AI in medicine
Topol's position on whether LLMs improve patient outcomes, set out in a May 3, 2026 review, is that the empirical case is thin. He wrote that there is "very little evidence for LLMs benefiting patients or doctors for health outcomes" outside administrative work such as clinical-note drafting, prior-authorization paperwork, and insurance-coding workflows (Topol/Marcus: LLMs and Patient Outcomes — three-document cluster; Source: open.substack.com).
Gary Marcus took the same position the same day, citing a Nature Medicine editorial making the same point in stronger institutional language (Source: garymarcus.substack.com). Topol and Marcus frame the industry-communications claim that "AI will revolutionize healthcare" as outpacing the clinical-outcome evidence base.
Topol's skepticism is specific to LLMs producing patient-outcome improvements rather than categorical opposition to AI in medicine; he has long advocated for AI in pattern-recognition tasks such as imaging and ECG analysis. He acknowledges productivity gains in non-clinical administrative tasks but argues that, under current evidence, those gains do not translate into improved diagnosis, treatment, or survival. He argues that randomized trials, rather than anecdotal deployments, should set the bar for clinical claims.
Topol's review is paired methodologically with the Lancet endoscopist-deskilling study, which documents an adverse outcome of AI assistance, as counter-evidence to optimistic deployment narratives in clinical settings.
Sources
- Topol/Marcus: LLMs and Patient Outcomes — three-document cluster — three-document cluster (Marcus essay + Topol review + Nature Medicine editorial) on the May 2026 patient-outcomes evidence gap.
- Gary Marcus (Substack) — Have LLMs improved patient outcomes? (May 3, 2026): garymarcus.substack.com
- Eric Topol (Substack, Ground Truths) — The Paradox of Medical AI Implementation (May 2026): open.substack.com
- Topol's own Substack and Deep Medicine book are background.
Relationships
- supports: Topol/Marcus: LLMs and Patient Outcomes — three-document cluster (the three-document cluster he co-anchors)
- related: Gary Marcus (May 3 same-day position), Healthcare — AI Deployment, Lancet Endoscopist Deskilling Study (2025), AI Deskilling
- contradicts: generalized "AI revolutionizes healthcare" framing in America's AI Action Plan and similar discourse