Prospective evidence for conversational medical AI is hard, but non-negotiable - Nature

Trust in clinical artificial intelligence (AI) cannot be benchmarked into existence. It must be earned through rigorous prospective studies in real-world clinical settings, where the hardest lessons often concern the humans and systems around the AI, not the technology itself. This is a preview of subscription content, access via your institution Access options Access through your institution Access Nature and 54 other Nature Portfolio journals Get Nature+, our best-value online-access subscription 27,99 € / 30 days cancel any time Learn more Subscribe to this journal Receive 12 print issues and online access 251,40 € per year only 20,95 € per issue Learn more Buy this article Purchase on SpringerLink Instant access to the full article PDF. 39,95 € Prices may be subject to local taxes which are calculated during checkout Fig. 1: Building trust in conversational and reasoning clinical AI. Subjects Health care Medical research References Brodeur, P. G. et al. Science 392 , 524–527 (2026). Article CAS PubMed Google Scholar Tu, T. et al. Nature 642 , 442–450 (2025). Article CAS PubMed PubMed Central Google Scholar Liévin, V. et al. Nature https://doi.org/10.1038/s41586-026-10764-5 (2026). Article PubMed PubMed Central Google Scholar Saab, K. et al. Nat. Med. 32 , 1726–1736 (2026). Article CAS PubMed PubMed Central Google Scholar O’Sullivan, J. W. et al. Nat. Med. 32 , 616–623 (2026). Article PubMed PubMed Central Google Scholar Chen, S. F. et al. Nat. Med. 32 , 1152–1159 (2026). Article CAS PubMed PubMed Central Google Scholar Tao, X. et al. Nat. Med. 32 , 934–942 (2026). Article CAS PubMed PubMed Central Google Scholar Lizée, A. et al. Preprint at https://doi.org/10.48550/arXiv.2411.12808 (2024). Zeltzer, D. et al. Ann. Intern. Med. 178 , 498–506 (2025). Article PubMed Google Scholar Brodeur, P. et al. Preprint at https://doi.org/10.48550/arXiv.2603.08448 (2026). Vedadi, E. et al. Preprint at https://doi.org/10.48550/arXiv.2507.15743 (2025). Gallifant, J. et al. Nat. Med. 31 , 60–69 (2025). Article CAS PubMed PubMed Central Google Scholar Zulman, D. M. et al. JAMA 323 , 70–81 (2020). Article PubMed Google Scholar Shekar, S., Pataranutaporn, P., Sarabu, C., Cecchi, G. A. & Maes, P. NEJM AI 2 , (2025). Walensky, R. P. & McCann, N. C. N. Engl. J. Med. 392 , 286–295 (2025). Article PubMed Google Scholar Download references Acknowledgements The authors thank the following individuals for their support in reviewing this article or enabling the studies discussed within: M. Thompson, M. Sanchez, R. Hemenway, B. Daryani, J. M. Williams, A. Goel, J. Xu, J. Lee, X. Ji, M. Nagda, R. Ruparel, W. Vaughan, T. Tiyasirichokchai, J. Breda, D. McDuff, T. Strother, V. Liévin, S. Schmidgall, L. Yang, D. Racz, D. Stutz, D. Barrett, V. Natarajan, R. Tanno, A. Karthikesalingam, E. Vedadi, C. J. Park, R. Sico, C. Kozlowski, Y. Mansour, J. Marshall, C. Boogaard, H. Spring, V. Roy, M. Eirich, P. Brodeur, J. M. Koshy, M. L. Cohen, J. Shreibati, J. Taylor, J. Hernandez, N. Hammel, L. Vardoulakis, P. Singh, K. Goel, K. Chou, J. Manyika, D. Hassabis and M. Howell. Author information Authors and Affiliations Google Research, Mountain View, CA, USA Mike Schaekermann, Anil Palepu, Ethan Goh, Yun Liu, Sunny Virmani, Christina Chen, Dale R. Webster, Avinatan Hassidim, Yossi Matias & Po-Hsuan Cameron Chen Department of Medicine, Beth Israel Deaconess Medical Center, Boston, MA, USA Adam Rodman Harvard Medical School, Boston, MA, USA Adam Rodman Included Health, San Francisco, CA, USA Ami Parekh Division of Computational Medicine, Center for Biomedical Informatics Research, Stanford University, Stanford, CA, USA Ethan Goh Stanford Clinical Excellence Research Center, Stanford University, Stanford, CA, USA Ethan Goh Google DeepMind, Mountain View, CA, USA Shekoofeh Azizi & Joëlle Barral Authors Mike Schaekermann View author publications Search author on: PubMed Google Scholar Anil Palepu View author publications Search author on: PubMed Google Scholar Adam Rodman View author publications Search author on: PubMed Google Scholar Ami Parekh View author publications Search author on: PubMed Google Scholar Ethan Goh View author publications Search author on: PubMed Google Scholar Shekoofeh Azizi View author publications Search author on: PubMed Google Scholar Yun Liu View author publications Search author on: PubMed Google Scholar Sunny Virmani View author publications Search author on: PubMed Google Scholar Christina Chen View author publications Search author on: PubMed Google Scholar Dale R. Webster View author publications Search author on: PubMed Google Scholar Joëlle Barral View author publications Search author on: PubMed Google Scholar Avinatan Hassidim View author publications Search author on: PubMed Google Scholar Yossi Matias View author publications Search author on: PubMed Google Scholar Po-Hsuan Cameron Chen View author publications Search author on: PubMed Google Scholar Corresponding authors Correspondence to Mike Schaekermann or Po-Hsuan Cameron Chen . Ethics declarations Competing interests This work was funded by Alphabet Inc. and/or a subsidiary thereof (‘Alphabet’). Authors who are employees of Alphabet may own Alphabet equity as part of the standard compensation package. Authors who are employees of Included Health may own Included Health equity as part of the standard compensation package. Rights and permissions Reprints and permissions About this article Cite this article Schaekermann, M., Palepu, A., Rodman, A. et al. Prospective evidence for conversational medical AI is hard, but non-negotiable. Nat Med (2026). https://doi.org/10.1038/s41591-026-04639-5 Download citation Published : 14 September 2026 Version of record : 14 September 2026 DOI : https://doi.org/10.1038/s41591-026-04639-5 Share this article Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy shareable link to clipboard Provided by the Springer Nature SharedIt content-sharing initiative