OpenEvidence
by OpenEvidence
Getting-started guide
A free AI reference for verified US healthcare professionals: ask a clinical question and get an answer that cites the medical literature. Not built for patients.
Current prices: Free · See all plans
Last reviewed 2026-10-24 · for OpenEvidence
You ask a clinical question in plain language — on the web or in the app. OpenEvidence searches peer-reviewed medical literature and guidelines and writes an answer with citations you can open and check. You pick how deep it goes: a quick answer in seconds, or a longer structured report that can take five minutes or more.
OpenEvidence is an AI tool for clinicians. Its terms say it is for physicians and other healthcare professionals and is not intended for use by patients or other consumers. Its user guide describes evidence-based answers to clinical questions, grounded in peer-reviewed literature, where every answer cites its sources. Typical uses it lists include quick consults, differential diagnosis, treatment options, drug information and guideline summaries.
It has content agreements with major publishers: OpenEvidence's about page names NEJM, JAMA, Nature, NCCN and Cochrane. Its user guide describes three answer models — Osler, the fast default; Sackett, more thorough and may ask clarifying questions; and Snow, a comprehensive structured report (this replaced the older DeepConsult feature). An EvidenceGrade on each answer rates the strength of the evidence behind it.
There's more for clinicians who sign up, including Visits, which records patient encounters and drafts clinical notes, and Patient Take-Homes, plain-language handouts a clinician sends to a patient. OpenEvidence's user guide says patients who receive a take-home can read it but cannot ask their own questions. If you use Visits, its terms say you must tell others in the room you're recording where the law or your organisation requires it, and let them opt out.
The model choice matters more than it first appears. OpenEvidence's guide describes Osler as giving direct answers in seconds, Sackett as taking seconds to a minute and sometimes asking you to clarify, and Snow as producing a full structured report that takes five minutes or more. Matching the model to the question saves you waiting for depth you don't need.
It's free for clinicians — and ad-supported. OpenEvidence's advertising policy says it is committed to providing the service free to healthcare professionals, paid for by a limited number of clearly marked ads, and that advertisers cannot influence answers. Its user guide says all three answer models are available to verified clinicians with unlimited usage at no cost.
You'll be asked to prove you're a professional. Its terms say registration asks for your name, email, profession, specialty and NPI number (or another licensing number), which OpenEvidence may try to verify. Its privacy policy says the service isn't intended for anyone under 18.
Not available everywhere. OpenEvidence's terms say the service is not available in the EU, UK or EEA.
If you're not a clinician, this isn't for you. OpenEvidence says so in its terms. Patients looking for health information should talk to their own clinician; the general chat assistants on Launchpad HQ are not a substitute for medical advice either.
Do not enter patient information. Leave out names, dates of birth, record numbers, addresses, photos and any detail that could identify a patient. OpenEvidence's security page says HIPAA covered entities may choose to send protected health information under its standard Business Associate Agreement, but its terms make you solely responsible for consents and for de-identifying anything you enter. Unless your organisation has approved OpenEvidence and confirmed that agreement covers you, ask about a de-identified case instead — "a 67-year-old with…" works just as well.
OpenEvidence's terms say it is for educational and informational purposes, is not a diagnostic service, and does not replace the clinical judgment of a qualified professional. Its acceptable use policy requires a professional to review content before it's shared with a patient, and to follow your institution's AI policies.
Management of new-onset type 1 diabetes in an adult with an HbA1c of 11% and no ketoacidosis: initial insulin regimen and first-week follow-up.
Why this one: OpenEvidence's prompting guide contrasts a vague question like "management of T1DM" with one that includes the key labs. Detail gets a more useful answer — and notice there's no patient identifier in it.
Differential diagnosis for macrocytic anemia in adults, as a table ordered from most to least common, with the first-line workup for each.
Why this one: Its prompting guide recommends asking for an output format, such as a table. A clinical curbside on macrocytic anemia is one of its own examples.
Alternative treatments for acute otitis media in a child with penicillin, cephalosporin and macrolide allergies.
Why this one: This is close to an official example of stating the outcome you want. Constraints like allergies narrow the answer to what's actually usable.
Risk of Lyme disease after a tick has been attached for 5 hours.
Why this one: OpenEvidence uses this example to show EvidenceGrade. Look at the grade and open two citations — it's the habit that makes the tool safe to rely on.
Want the full picture first? See the OpenEvidence overview for pricing, the free tier and related tools.
