How we keep explanations careful

What it does

• Reads the numbers on your report and compares each one with the range your lab printed.
• Says in plain words which results are outside that range, and how soon to ask your doctor about them.
• Gives you questions to bring to the visit and a one-page doctor brief.

What it never does

• It does not diagnose, and never says a result is definitely fine or definitely serious.
• It never names a medicine or a dose, and never suggests starting, stopping or changing one.
• It does not look at scan images. It explains the radiologist's written words only.
• If a report or a question mentions chest pain, trouble breathing, stroke signs, fainting, heavy bleeding or thoughts of self-harm, it tells you to call your emergency number first.

How we test it

We test in three separate ways, and only the last one is about clinical safety.

• Automated checks on every change. Thousands of made-up reports in US, UK, European and Indian formats, with different units, decimal commas, tables, misread numbers and unusual layouts, plus fixed safety cases such as emergency symptoms and levels labs often treat as urgent. A change ships only if these pass.
• Our own test sets. Made-up reports with known answers that we use to find and fix reading mistakes.
• Independent clinical review: not done yet. 160 made-up cases and the rules behind them are prepared for a licensed clinician to review. Until that is finished, nothing here has been clinically validated.

Passing more automated tests does not make the explanations clinically safe. It means the reader does what we designed it to do on reports like the ones we tested. Real reports can still be misread.

Its limits

• Photos can be misread. The app shows what it read so you can check it against the original.
• Ranges differ by lab, age and sex. Your doctor knows your history; the app does not.
• The Spanish wording is a draft that is still being reviewed.
• The explanations have not yet been reviewed by a clinical advisory board. We will say so here when they have.

Your data

Reports are read on your own phone or computer. Nothing is uploaded unless you save it to an account, share a doctor link, or ask the help chat. The help chat sends your question to our AI provider, Anthropic, to write the answer. Our server does not save your question or the answer; Anthropic may keep them for a limited time under its own data terms. We never sell your data or use it for ads. Read the privacy policy or delete your data at any time.