Vera Health: the AI medical search engine transforming access to scientific evidence

As scientific literature continues to grow at an unprecedented pace, two young French entrepreneurs have launched Vera Health, an AI medical search engine designed to find the strongest scientific evidence in seconds. The application is experiencing remarkable growth, with an impressive adoption curve among healthcare professionals.

More than 5,000 scientific papers and medical guidelines are published worldwide every day. “Keeping up has become inhuman” for doctors, say Vera Health co-founders Taieb Bennani and Maxime Allouch.

Both from families of healthcare professionals, the entrepreneurs grew up with similar striking scenes: doctor fathers spending each evening absorbed in hefty, highlighted reference books to solve a difficult case or prepare for a procedure.

Having studied at the prestigious MIT, HEC Paris and Yale, they aim to offer clinicians a straightforward, accessible solution. “By far, doctors are the best data scientists: they process hundreds of data points for every patient and have only a few seconds to make a decision. We are here to give them the best information at the exact moment they need it,” Taieb Bennani explained in an interview with Presse-citron.

A specialised medical search engine

Vera Health is a highly specialised search engine. The application searches, in real time, a database containing more than 60 million scientific articles and medicine monographs - the official, comprehensive information record for a drug. “We take every piece of metadata into account to recreate something like Google’s PageRank, but genuinely for doctors,” the young entrepreneur says.

In clinical practice, the tool replaces a frequently laborious manual research process. “A doctor sees a foreign patient taking a treatment that is unfamiliar in France. Checking drug interactions in Vidal [editor’s note: France’s leading medicines information reference] can take 5 to 10 minutes. On Vera, they ask the question and receive an answer within seconds,” Maxime Allouch explains.

“We make medical guidance actionable, so doctors do not have to read 15 or 20 PDF pages and can go straight to the sentence that matters,” Taieb Bennani adds. It is a highly promising solution. Shortly after its launch in July 2024, the US-based start-up was selected for the prestigious Y Combinator accelerator.

It subsequently completed a $3 million seed funding round with Gradient, Google’s artificial intelligence investment fund.

Vera Health and the drive to eliminate hallucinations

In medicine, errors are not an option. To prevent the “hallucinations” associated with general-purpose AI, Vera Health follows a rigorous approach: “The answer already exists in scientific publications. Our models are only there to synthesise or rephrase existing information,” Taieb Bennani stresses. For every answer, the tool assigns a colour code showing its confidence level, choosing not to respond when supporting evidence is unavailable.

Another major strength of Vera Health is its ability to adapt with great precision to each clinical setting. “Every search engine needs to be different for every doctor. That applies at country level, by specialty, and even by institution,” Taieb Bennani elaborates.

“An emergency physician does not have 25 seconds. They generally have 10 to 11 seconds between typing their question and needing to absorb the answer. For an oncologist, by contrast, the priority is receiving the most comprehensive report possible, while response time is secondary,” the co-founder illustrates. This is why the system runs between 8 and 9 specialist models in parallel, classifying the request and instantly tailoring the level of detail provided.

This discipline is paying off. In the NEJM-AI and MedXpertQA benchmark tests, the application delivers results two to three times better than the leading OpenAI and Anthropic models, reaching a 97.5% pass rate in the USMLE, the US medical examination. To preserve this level of accuracy, the application is continuously reviewed by a committee of clinicians from Harvard, Yale and the Mayo Clinic.

Rapid adoption outpacing Doctolib’s trajectory

Driven almost entirely by clinical word of mouth, the application has achieved striking success. In France, the monthly volume of processed questions rose from 7,000 in January to 200,000 in June 2026. It has 31,000 registered users nationally, with the number of doctors increasing 42-fold in six months. Its adoption curve is remarkable: Vera has already reached more than 24,000 practitioners in 6 months, whereas Doctolib had around 1,500 one year after its launch.

Internationally, the start-up says it is used by more than 400,000 clinicians. This momentum has just been recognised by Headline, which included the company in its AI Europe 100 ranking. It also caused a stir on the OpenAI DevDay stage by presenting a voice agent developed during a hackathon it won in front of Sam Altman. The company has also partnered with the American College of Emergency Physicians (ACEP), directly incorporating recommendations from the leading US emergency medicine society for its 40,000 members.

Vera is not intending to stop there and has just launched its new personalised “Spaces”. This feature allows doctors and healthcare institutions to build their own Vera Health environment by adding local documents. “Our strength is that we can search 60 million scientific papers, but also 500 pages of internal PDFs from a hospital such as Foch,” Taieb Bennani explains. As a result, clinicians can cross-reference the latest global recommendations with protocols specific to their own department in real time.

Beyond a search tool, a genuine medical community is taking shape. “The idea is to use our network to go further, rather than simply answering questions in a static way,” Maxime Allouch concludes. A clear ambition to transform access to medical knowledge for the long term.