MedUni Exam

An AI question generator that has to prove where each question came from

Upload the material you are examined on and get exam questions written from its own sentences. Under every question is the line it was written from and the page that line sits on, and a question whose quote is not on that page never reaches you.

Try it on your own file

Free to try. No card and no account needed.

Try the sample exam, ten real questions with no upload and no account.

What comes back

Exam questions in the shapes exams use: single best answer with four or five options, clinical vignettes that open with a patient, multi-select with more than one key, true or false, and short answer graded against the source. Each one arrives with its answer, an explanation written afterward in the background, and the quote and page it was built from.

How each question is checked before you see it

Verbatim quote

Every question carries the sentence it was written from, copied out of your file, with its page number.

Quote check

Before you see a question, the quoted sentence is matched against the stored text of the page it cites. If it is not there, the question is dropped.

Blind second answer

A second model from a different family answers the question from the quote alone, without the key. Disagreement drops the item.

Item writing filter

Options like all of the above, vague which of the following is true stems, and questions about "the text" instead of the subject are rejected by pattern before they reach you.

Near duplicate index

Each new item is compared against everything already generated from the document, so a long run gives you different questions instead of twelve facts restated.

Two of these five are the checks that reject on evidence, the quote check and the blind second answer. The other three are rules about what a question must carry before those checks run.

What it reads

  • PDF. A textbook chapter, a handout, typed notes. Scanned pages are read with optical character recognition first. Up to 3,000 pages a document, of which up to 600 may be scanned. How a PDF becomes a quiz.
  • PowerPoint. Each slide is read as a slide, so a question cites the slide it came from. Uploading decks.
  • Word, plain text, pasted text and web pages. These have no pages of their own, so they are cut into numbered sections and a question cites its section.
  • Any language. Questions come out in the language of the document or the one you pick. What is checked in your language, and what is not.

Measured on real files

LaneFileWorkWall clock
Free poolAn OpenStax anatomy chapter of 44 to 60 pages, text PDF20 questions generated and checkedabout 15 to 18 seconds
Reading onlyA 900 page textbook, text PDFevery page read and storedunder 40 seconds

Measured on the live site, the chapters on September 23, 2026 and the textbook on September 12, 2026, from the moment the job started to the moment the last question was accepted. Your own times move with the length of the job and how busy the free lane is that hour.

Who it is written for

Students of medicine and the health sciences who are examined on their own course material: medicine, nursing, pharmacy, dentistry, physiotherapy and biomedical science. A bought question bank is written to a national blueprint; the quiz next week is written to the slides your instructor posted. There is a page for each of ten exams, USMLE Step 1, NCLEX-RN, PLAB 1 and the rest, each with its published format, its source and a recipe for the generate form.

What it costs

The free plan is 300 verified questions a month with an account, 60 a month as a guest with no account, on a pool of open-weight models. The Exam Pass, $19.99 once, puts the priority models on your jobs for 90 days and adds the downloads; Pro is $9.99 a month. Every plan, line by line.

Questions about the generator

What does the generator need from me?

A file: a PDF, a Word file, a PowerPoint deck, a plain text file, text pasted into the box, or the address of a public web page. Then a count, a question type or a mix of types, and, if you want, a language. The rest is done for you.

How is this different from asking a chatbot?

A chatbot writes questions from memory and hands you the list. Here every question must quote one or two lines of your file with the page they sit on, the quote is looked up on that page, and a second model from a different family answers the question blind. A question that fails either check is thrown out before you see it.

Can it write questions for my exam format?

Single best answer, clinical vignettes, multi-select, true or false and short answer. The exam pages carry a recipe for each of ten exams, written for the extra instructions box, so the items come back in that exam's shape.

How many questions can I generate?

The free plan is 300 verified questions a month with an account and 60 as a guest. Paid plans move the work onto the priority models: the Exam Pass covers 90 days and Pro is billed monthly.

Is it accurate?

It is only as accurate as your file. The checks stop a question from inventing a fact that is not on the page and from carrying a key the second model cannot reach from the quote; they cannot correct a source that is itself wrong. The rejection rates page shows how much of what the models write is thrown out.

Try it on your own file

See pricing

No account needed to try it. Registering keeps your library, grades and review queue across devices.

Published , updated .