Debate Over the Doctronic AI Prescription Medication Chatbot

In January 2026, Utah rolled out a new artificial intelligence (AI) prescription renewal service in partnership with Doctronic, an AI healthcare company. This partnership is one of the first of its kind, potentially breaking ground on a new era of physician–patient interaction. For now, the initial rollout focuses specifically on prescription medication renewal: the patient is asked a series of questions by the AI chatbot, which determines their eligibility for a prescription renewal. At the time of the rollout, the Doctronic chatbot had access to 192 medications, including treatments for conditions such as hypertension, depression, diabetes, and asthma. This rollout highlights a larger question in healthcare: how effective and safe are AI models at meeting the unique demands of healthcare settings?

Traditional clinical interactions usually entail a follow-up appointment with a physician to assess the efficacy of a medication and the need for a prescription refill. Doctronic bypasses this: patients interacting with the chatbot are asked a series of questions about their health and the medication they are trying to refill. The AI chatbot then performs an assessment and determines whether or not to renew the medication prescription. While there are preliminary plans to roll out a more extensive Doctronic program pending the results of the pilot study, for the time being, Doctronic does not formulate diagnoses or compose care plans.

Research on AI safety has produced mixed results, and to some extent, these disagreements reflect the highly variable judgments that are regularly required in clinical settings. For example, in-house research cited by Doctronic showed that Doctronic matched physician diagnoses 81% of the time and treatment plans 99.2% of the time (Hayat et al., 2025). In contrast, research analyzing other types of judgment that are critical in a clinical setting, such as medication counseling or identifying drug–drug interactions, shows less encouraging results. One study found that ChatGPT had a sensitivity of only 24% in detecting drug–drug interactions (Krishnan et al., 2024). Another study found that while ChatGPT scored similarly to pharmacists on medication counseling, it fared far worse in conducting prescription review (Huang et al., 2024).

With any analysis of AI efficacy, it is important to emphasize that differences across AI models and training data can result in profoundly different results. Therefore, any direct comparison between Doctronic, which is designed for specific clinical tasks, and other models like ChatGPT, which is designed for far more generalized use, needs to be considered in light of that difference. However, Doctronic is still relatively new, and independent research attempting to reproduce any capabilities it may have demonstrated during preliminary in-house testing remains extremely sparse. In fact, an analysis published in JAMA Health Forum noted that the evidence cited by Doctronic did not directly test the AI prescription-renewal chatbot system itself.

While early data may be limited, the potential utility of AI in the clinical setting should not be understated. There is expected to be a significant shortage of physicians in the coming years, which has prompted renewed interest in alternatives or supplements to existing clinical infrastructure. AI may be able to assist with routine aspects of primary care, particularly the management of stable, chronic conditions. Alternatively, AI may provide a valuable adjunct to clinical decision-making. The ability of AI models to process massive amounts of information could help identify potential drug interactions or alternative therapies that a physician may have overlooked. However, comprehensive and rigorous testing and scrutiny is necessary to ensure that no patient harm occurs—an area that the field of AI research continues to struggle with.

The rollout of an AI chatbot with decision-making power, specifically in renewing medication prescriptions, may represent a significant development in physician–patient relationships. While preliminary research is mixed, there are a myriad of potential applications for AI in clinical settings. The important question at this point may not be whether AI should have a role in the clinical setting, but rather identifying the circumstances that produce the best synergy between human and artificial intelligence while protecting patient interests and safety.

References

Aaron, D. G., & Robertson, C. (2026). The first AI drug prescriber. JAMA, 335(18), 1567–1568.

Huang, X., Estau, D., Liu, X., Yu, Y., Qin, J., & Li, Z. (2024). Evaluating the performance of ChatGPT in clinical pharmacy: A comparative study of ChatGPT and clinical pharmacists. British Journal of Clinical Pharmacology, 90(1), 232–238.

Jaiswal, N., Gorijavolu, R., Hernandez-Boussard, T., & Celi, L. A. (2026). When an algorithm renews a prescription. JAMA Internal Medicine. Advance online publication.

Krishnan, R. P. R., Hung, E. H., Ashford, M., Edillo, C. E., Gardner, C., Hatrick, H. B., Kim, B., Lai, A. W. Y., Li, X., Zhao, Y. X., & Raubenheimer, J. E. (2024). Evaluating the capability of ChatGPT in predicting drug-drug interactions: Real-world evidence using hospitalized patient data. British Journal of Clinical Pharmacology, 90(12), 3361–3366.