Trusted tools for responsible clinical AI
In 2026, Google published its latest evaluation of the Articulate Medical Intelligence Explorer (AMIE), 59 examining whether AI could safely support longitudinal disease management across multiple patient consultations. Unlike earlier studies focused on diagnosis, this research evaluated AI's ability to reason about investigations, treatment selection, medication prescribing, disease progression and follow-up care over time. 60 To develop and evaluate the study, Google selected just two core clinical knowledge resources for evidence based clinical practice guidance: BMJ Best Practice and National Institute for Health and Care Excellence (NICE) guidance. Together, they informed the development of 100 multivisit clinical scenarios used to assess AI performance against evidence based standards of care. 61 Explaining the decision to include evidence from BMJ Best Practice in the AMIE evaluation, the Google team said they selected "authoritative clinical knowledge sources that are widely accepted and used in clinical practice", citing the tool's "international coverage and strong evidence appraisal" as key factors in its inclusion. Google also noted that the evidence based content provided in "both human-readable PDF format and machine-parsable structured formats" was helpful for evaluating how AMIE and clinicians used the content. 62
At Google, we believe that working closely with partners that curate trusted clinical content is paramount in our journey of studying AI for clinical tasks in a responsible manner. The international coverage and strong evidence
appraisal of guideline content were additional reasons for us to partner with BMJ Best Practice.” Mike Schaekermann, Staff research scientist and research lead, Google, Toronto, Canada
23
Powered by FlippingBook