Our Take
OpenEvidence built the largest installed base of any clinical AI tool by ignoring hospital IT entirely. That strategy worked until it didn't, and now the company is asking the very gatekeepers it bypassed to let it in.
Why it matters
Clinical AI adoption has historically required hospital procurement cycles that take 18+ months. OpenEvidence proved direct-to-clinician could work at scale. But ad-based models face margin pressure, and health systems want contractual guarantees, not voluntary adoption. This pivot signals that even winner-take-most consumer health tech eventually needs institutional rails.
Do this week
CIOs: audit your clinical AI licensing terms for conflict-of-interest clauses and ad-serving restrictions before any new vendor pitch arrives this quarter.
OpenEvidence shifts from physicians to health systems
OpenEvidence, a free clinical decision-support chatbot, has accumulated roughly 650,000 active U.S. physician users in four years (company-reported). The platform reached a $12 billion valuation by capturing physician adoption without requiring hospital procurement sign-off, a traditional barrier in health tech. At the STAT Breakthrough Summit West in San Francisco this week, co-founder and CTO Zachary Ziegler announced the company is now pursuing formal relationships with health systems themselves, moving beyond its original direct-to-clinician model.
The shift reflects competitive pressure and questions about the sustainability of OpenEvidence's ad-supported business model. Free tools relying on advertising generate thin margins and create friction with institutional buyers concerned about ad-serving in clinical workflows. Health systems want contractual control, liability indemnification, and integration with existing electronic health records (EHR) systems, none of which the current model provides.
The direct-to-clinician bypass is hitting its ceiling
OpenEvidence proved that clinical AI adoption did not require navigating hospital IT and procurement departments. Thousands of physicians adopted the tool voluntarily, creating network effects that benefited both trainees (faster literature synthesis) and the company (data and user engagement). This model worked at the 100K-user scale and still works at 650K. But it has never scaled to institutional deployment, which requires different economics.
Hospitals now control clinical AI strategy through RFI and procurement processes. They want vendor accountability, data governance commitments, and fee structures that align with revenue (subscription per seat, per transaction, or shared savings). Ad-based tools cannot offer these. OpenEvidence's willingness to pursue health systems directly signals that even dominant consumer adoption in healthcare cannot ignore institutional buyers indefinitely. The two channels (individual clinician vs. health system) require different products, sales teams, and compliance postures.
What institutional buyers should verify
If OpenEvidence or competitors approach your health system, confirm three details before engaging: first, whether ad-serving or third-party data use continues during institutional deployment; second, what liability indemnification the vendor offers for incorrect clinical recommendations; and third, whether the tool integrates with your EHR at the point of care or requires separate login and workflow overhead. Free adoption at the physician level does not guarantee usable institutional deployment. Ask to audit the health system pilot results (outcomes, adoption rates, time-to-value) from at least two comparable organizations before signing a pilot agreement.