AI-powered clinical knowledge platform OpenEvidence has launched a new feature that assesses the quality of published evidence cited in AI answers in real-time. This allows clinicians to understand the strength of the evidence provided before relying on it to make treatment decisions.
OpenEvidence said the new feature, called EvidenceGrade, is the latest step in the company’s “broad effort to bring rigorous evidence evaluation to clinicians at the point of care.”
In March 2026, OpenEvidence announced a partnership with Cochrane, a global nonprofit organization that produces independent reviews of medical research. Based on this partnership, EvidenceGrade was built on the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework, an approach for appraising evidence and supporting medical decision-making used by Cochrane, the World Health Organization, and other clinical guideline developers.
How EvidenceGrade works
When a physician asks OpenEvidence a question and receives an answer, EvidenceGrade first determines whether the question results in a clear evidentiary claim. The system filters out searches and queries for simple definitions that focus on tasks such as summarization.
Once a question is determined to be scorable, each published article included in the answer will be evaluated for quality, certainty, and relevance and will be given a grade from A to D. Grade A means the evidence uses a rigorous systematic review, and grade D means the evidence consists of preclinical or animal data and may be unreliable.
In a blog post accompanying the announcement, OpenEvidence explained that users have long requested a systematic evaluation of the strength of the evidence uncovered in their responses. However, manually grading medical evidence is time-consuming and difficult. As a result, only a small portion of the questions that clinicians actually ask are subject to formal evidence evaluation.
EvidenceGrade is OpenEvidence’s attempt to provide transparency in evidence grading to physician users without sacrificing speed, the blog post explains.
“Not all evidence is equally robust, and when clinicians act on answers, it is essential to understand how much weight the underlying evidence holds,” Samuel Finlayson, MD, OpenEvidence’s senior vice president of medical AI, said in the announcement.
“Expert research review teams do valuable work, but they cannot cover every question that arises in daily practice. We built EvidenceGrade to extend their methodology to as many questions as possible. We see it as a complement to traditional evidence synthesis, and we are releasing it as a starting point that can be improved through collaboration with the clinical community.”
Comparing OpenEvidence to general purpose AI tools
OpenEvidence has been extremely popular since its founding in 2021, and is now used by nearly two-thirds of U.S. physicians. However, in a study published in n years,natural medicine In June 2026, we found that general-purpose AI tools such as GPT-5.2 and Claude Opus 4.6 can potentially outperform specialized LLM-based clinical AI tools such as OpenEvidence and UpToDate Expert AI on multiple medical benchmarks.
This study raised questions about the value of creating unique clinical AI tools for specific industries versus using generic tools or developing hospital-specific LLMs.
In a LinkedIn post in response to the study, OpenEvidence argued that the paper and its benchmarks were “clearly insufficient to judge clinical decision support tools or draw conclusions about professional clinical AI,” and that the authors had “obvious” conflicts of interest and were “indifferent to underlying data contamination concerns.”
a few days later natural medicine The study was published and OpenEvidence reached out to researchers at UCSF to conduct the new study. status Reported. OpenEvidence reportedly provided the data, paid survey respondents, and co-developed the data collection plan.
The preprint results compared OpenEvidence to the general-purpose tool and found that “physicians scored highest on the specialized tool on all axes,” which is exactly the opposite. natural medicine study.
OpenEvidence has not disclosed any involvement in the study, and its executives are not listed as authors. As the AI race continues, specialized platforms like OpenEvidence will face increased competition from big tech companies like OpenAI and Anthropic.
Jill Hughes has been covering health tech news since 2021. Her areas of coverage include cybersecurity, HIPAA compliance, interoperability, AI, and EHR.
