Arkangel AI, OpenEvidence, ChatGPT, and Medisearch evaluated against medical standards

Peer reviewedMayo Clinic Proceedings: Digital Health 2026 · 1 min read
Reference authenticity · Arkangel AI-Deep
100%
Also reached by ChatGPT-Deep and OpenEvidence
Overall satisfaction across agents
71.1–93%
Expert evaluations
1024
128 question/answer pairs

Peer-reviewed 2026 study compares four conversational agents across 128 Q/A pairs and 1024 expert evaluations.

Jose Zea

This peer-reviewed Mayo Clinic Proceedings: Digital Health study compared Arkangel AI, OpenEvidence, ChatGPT, and Medisearch across four clinical vignettes and four question types. The design produced 128 Q/A pairs and 1024 expert evaluations using an eight-criterion Likert scale.

Overall satisfaction ranged from 71.1% to 93% across agents (Kruskal-Wallis P<.001). Reference authenticity reached 100% satisfaction for Arkangel AI-Deep, ChatGPT-Deep, and OpenEvidence. ChatGPT modes were rated unsatisfactory in 75–97% of evaluator pairs, and median response times ranged from 18 seconds to 12.8 minutes (Wilcoxon P<.05).

Results at a glance

The headline numbers from the paper, in plain language.

MetricValueWhat it means
Reference authenticity · Arkangel AI-Deep100%Also reached by ChatGPT-Deep and OpenEvidence
Overall satisfaction across agents71.1–93%
Expert evaluations1024128 question/answer pairs

How to cite

Arkangel AI. Arkangel AI, OpenEvidence, ChatGPT, Medisearch: Are They Objectively up to Medical Standards?. Mayo Clinic Proceedings: Digital Health. 2026. https://www.mcpdigitalhealth.org/article/S2949-7612(26)00022-2/fulltext

@article{arkangel2026arkangel,
  title     = {{Arkangel AI, OpenEvidence, ChatGPT, Medisearch: Are They Objectively up to Medical Standards?}},
  author    = {{Arkangel AI}},
  journal   = {Mayo Clinic Proceedings: Digital Health},
  year      = {2026},
  url       = {https://www.mcpdigitalhealth.org/article/S2949-7612(26)00022-2/fulltext}
}

More research from Arkangel AI

All research →

For hospitals and insurers

Validate these results at your institution

We run the model on your own clinical records with your team, under your privacy and governance rules.

Start a research partnership
Read paperCite