Medidata vs Unlearn.AI (2026)
Medidata, a Dassault Systemes brand, sells the software many clinical trials run on, anchored by Rave data capture. Since July 2026 its AI comes packaged as Medidata Plus. That layer builds study forms and edit checks, reviews data and safety signals in Clinical Data Studio, redacts and checks trial images, and runs the Dot assistant for simulation and forecasting. Drug and device sponsors and contract research organizations buy it, about 2,300 customers in all. Unlearn sells one thing in depth: digital twins that forecast how each trial participant would have progressed on control. In a randomized trial the forecast becomes a covariate in the analysis, a method called PROCOVA that the European Medicines Agency qualified in 2022 for phase 2 and 3 primary analysis. Sponsors buy it as a partnership with Unlearn's scientists, to shrink the control arm or raise a trial's power. Unlearn now also offers twins as comparators in single arm studies. The two work at different layers of a trial. Medidata runs the study itself, from build to data review, and adds AI to each of those jobs. Unlearn changes the statistics, so the same question can be answered with fewer patients. A sponsor wanting AI across its study operations, on the platform it already uses, looks to Medidata. A sponsor whose problem is enrollment, cost or power in a specific disease looks to Unlearn.
- AI study build generates forms, edit checks and synthetic test patients inside Rave.
- Clinical Data Studio brings Medidata and outside data into one review with risk and safety signal detection, and ICON uses it as its single review platform.
- Health Record Connect reads patient record data through health information exchanges and FHIR into Rave forms, and nothing is submitted until the site saves.
- Its ISO/IEC 27001 certificate, valid to October 2028, and its ISO 27017, 27018 and 27701 documents are public on a trust page.
- Imaging redacts identifying text automatically and checks anatomy before images move on.
- Digital twins forecast each participant's outcome on control, and the PROCOVA analysis keeps a randomized trial unbiased even when a forecast is off.
- The European Medicines Agency qualified PROCOVA in 2022 for primary analysis of phase 2 and 3 trials with continuous outcomes.
- Reanalyses of completed phase 3 trials showed a 33 percent smaller control arm and more than four months of enrollment saved.
- Generators cover 26 conditions and can run inside the sponsor's own cloud.
- Unlearn's scientists draft the statistical analysis plan and deliver the analysis.
This comparison is published by AI Health Index, an independent research platform that compares healthcare AI vendors objectively. Medidata and Unlearn.AI are each graded against the same capability taxonomy, from each vendor's own public materials and the regulatory record, under the AI Health Index verification standard. No vendor pays for placement, and no vendor has reviewed this page. How this evidence is graded
Plain facts
| Fact | Medidata | Unlearn.AI |
|---|---|---|
| Primary category | Clinical Trials AI | Clinical Trials AI |
| Founded | Not recorded | 2017 |
| Headquarters | New York, New York | San Francisco, California, United States |
| Website | medidata.com | unlearn.ai |
Side by Side
Select any grade to read the note behind it.
Each vendor in one paragraph
Written to be quoted whole. Each paragraph states what the AI Health Index verified about the vendor, with the caveats attached.
Medidata, a Dassault Systemes brand, sells clinical trial software anchored by Rave data capture to drug and device sponsors and contract research organizations. Its AI layer, Medidata Plus, builds study forms and edit checks, reviews data and safety signals in Clinical Data Studio, redacts and checks trial images, and adds the Dot assistant for forecasting and simulation. Health Record Connect pulls patient record data into Rave forms for sites to confirm. Medidata publishes its security certificates openly, but its product figures come without method, its generative models are unnamed, and it publishes no prices.
Source: AI Health Index, October 2026
Unlearn sells Digital Twin Generators, disease models that forecast how each clinical trial participant would progress on control, to drug and biotech sponsors through partnerships with its scientists. Used as a covariate under PROCOVA, which the European Medicines Agency qualified in 2022 for phase 2 and 3 primary analysis, the forecasts let a randomized trial reach the same answer with fewer patients. Reanalyses of completed trials show a 33 percent smaller control arm in phase 3 and 92 fewer participants in an ALS study. The evidence is the company's own reanalyses, the newer single arm use sits outside the qualification, and no price is published.
Source: AI Health Index, October 2026
Questions buyers ask
How does Medidata compare to Unlearn?
Medidata and Unlearn apply AI to different layers of a clinical trial. Medidata sells the trial platform itself, anchored by Rave data capture, and packages AI for study build, data review, imaging and forecasting as Medidata Plus. Unlearn sells digital twins that forecast each participant's outcome on control, used in the analysis to shrink the control arm or raise power. Medidata suits a sponsor wanting AI across study operations.
Unlearn suits a sponsor trying to run a smaller or better powered trial. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified October 25, 2026. No vendor pays for placement.
How much do Medidata and Unlearn cost?
Neither Medidata nor Unlearn publishes a price. Medidata sells enterprise agreements scoped by study count and module, with its AI packaged as Medidata Plus, and what Medidata Plus adds to a platform license is not published. Unlearn sells partnerships across a single study, a program or a portfolio, and states value instead of price, such as about $2 million saved on one phase 1/2 study. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified October 25, 2026. No vendor pays for placement.
Have regulators accepted Medidata's or Unlearn's AI?
Unlearn's PROCOVA method holds a 2022 qualification opinion from the European Medicines Agency's CHMP for primary analysis of phase 2 and 3 trials with continuous outcomes. It covers the statistical method rather than a specific generator, and not the newer single arm use. Medidata makes no device claim for its AI and publishes a data integrity white paper on how its platform meets 21 CFR Part 11 and FDA, EMA, MHRA, Japanese and Chinese rules for computerized systems.
FDA lists Rave Imaging as a Class I image communications system. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified October 25, 2026. No vendor pays for placement.
What evidence supports Medidata's and Unlearn's AI?
Unlearn's evidence is reanalysis of completed trials: an 18 percent variance reduction and 92 fewer participants in a phase 3 ALS trial, 17 percent fewer participants in an AbbVie phase 2 Alzheimer's study, and a 33 percent smaller control arm in phase 3 trials from a top five sponsor. It also has a 2019 Nature paper on forecasting Alzheimer's progression.
Medidata's evidence is named customers such as ICON, Eisai and Worldwide Clinical Trials, speed figures such as data review cycles up to 80 percent shorter without a published method, and 2024 company papers on data mapping and event detection. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified October 25, 2026. No vendor pays for placement.
Where do Medidata and Unlearn run, and how do they handle security and patient data?
Medidata hosts its clinical cloud in its own Houston data center and publishes its ISO/IEC 27001:2022 certificate, valid to October 2028, with ISO 27017, 27018 and 27701 documents and a data processing exhibit on a public trust page. It says it uses no data for secondary purposes without written permission, and states no business associate position.
Unlearn works on trial datasets rather than health records, runs as a web application or inside the sponsor's own cloud, and lets sponsors build custom generators entirely inside their environment. It announced SOC 2 Type II in 2024 without publishing the report, and commits to HIPAA safeguards where it signs a business associate agreement. From the AI Health Index, drawn from each vendor's own published materials and the regulatory record, verified October 25, 2026. No vendor pays for placement.
Related comparisons
Other published head to head assessments involving these vendors or their closest peers.
Medidata is a Dassault Systemes brand based in New York, and Unlearn is based in San Francisco. Neither publishes a price. Medidata does not name the language model behind Dot or its other generative features, or publish a recall figure for automated image redaction, and its subprocessor list and SOC 2 report sit behind a customer login. Unlearn does not publish where a single arm comparator stops being appropriate, or how its generators perform across age, sex, race or disease stage.