PDF to MCQ: MCQ Maker and Generator That Converts a PDF to Multiple Choice Questions

Which PDFs produce weak MCQs, and what to fix before you upload

Open the PDF, press select all, copy, and paste into a plain text editor. What you see there is close to what any converter sees. If the pasted text reads as clean paragraphs, your questions will be good. If sentences break mid-clause or two unrelated lines are spliced together, that is the problem, and it happens before the AI ever reads a word.

The reason is structural. A PDF records where marks sit on a page, not what order a human should read them in. Reading order only exists in the file when the PDF is tagged, meaning it carries a structure tree describing headings, paragraphs, lists and tables. W3C covers this under WCAG technique PDF3. Most textbook scans, journal articles and exported handouts are untagged, and export tools that add tags automatically often get multi-column layouts wrong anyway.

How much source material a good MCQ set needs

One page of dense body prose supports about two to three solid questions. A ten page chapter comfortably carries twenty to thirty. That ratio is not a technical limit, it is where the material runs out: past it the generator has no fresh assertions left and starts testing trivia or rewording facts it already used, which is how near-duplicate items appear in a set.

If you need more questions than the ratio allows, add source rather than raising the count. Upload the chapter together with the lecture notes covering it, and the extra angles give the generator somewhere new to go. The reverse also holds: on a very long upload, narrowing to the pages you are actually being tested on produces sharper questions than the whole book does, because everything the model reads competes for the same twenty slots.

Reference on PDF reading order and tagging: W3C, WCAG technique PDF3, Ensuring correct tab and reading order in PDF documents.

How do you know if the generated MCQs are any good?

Cover the answer options and try to answer the question from the stem alone. If you can, the item is testing whether someone learned the material. If you cannot answer until you read the options, the stem is too vague and the item is really testing reading speed. That single check is the cover-the-options rule from the National Board of Medical Examiners item-writing guide, and it catches most of the weak questions in any generated set in about ten seconds each.

The same guide is useful for a second reason. It lists the specific technical flaws that let a student who does not know the answer pick the right option anyway. These are worth knowing because they are the flaws an automated MCQ maker is most likely to produce, and because every one of them is a fast edit rather than a rewrite. Run a generated draft against this list before you give the test to anyone.

Flaw categories and the recommended fixes follow the National Board of Medical Examiners, Item-Writing Guide, chapters 3 and 4.

A distractor does not have to be completely wrong

This is the piece most people writing their first MCQ set get backwards, and it is worth knowing before you edit a generated draft. In a one-best-answer question the task is to pick the most correct option, not the only correct one. The NBME guide is explicit about it: as long as the options sit on a single continuum, from least likely to most likely, the distractors do not have to be totally wrong. A question where three options are obviously absurd tests nothing, because a student who read the chapter once will still get it right.

So when you review the draft, resist the urge to make wrong answers more wrong. The better move is the opposite: make them more tempting. Ask what someone who half learned this material would believe, and put that in an option. A distractor earns its place when a real misconception would lead a real student to choose it.

Which questions to cut after you have used the test once

Two numbers tell you almost everything about an item once you have shared the quiz with a class or run it with a team. PDFQuiz does not calculate either one, so this is a description of what to work out for yourself from the answer sheets, not a feature list. The first is the percentage of people who got the item right, what test developers call the p-value. Items above 95 percent or below 30 percent tell you very little about the group, because nearly everyone or nearly nobody got them. They are not automatically bad questions, but a test built mostly out of them cannot separate people who learned the material from people who did not.

The second is the item-total correlation, the relationship between getting one question right and scoring well on the test overall. It runs from minus 1.0 to plus 1.0, and you want it positive: the people who did well on the whole test should be the people who got this item right. A negative correlation is the useful signal, because it usually means the answer key is wrong. The strongest students picked an option you marked incorrect, which is exactly what happens when a generated item was keyed to the wrong choice or the source passage was ambiguous. Check those items first.

If you use PDFQuiz as an MCQ creator for the same course every term, this is where a generated bank starts to beat a handwritten one. Regenerating from the same PDF gives you fresh items to swap in for the ones that did not discriminate, and you keep the questions that did.

PDF to mock test, question paper or exam: what changes

The generation step is identical for all three. What changes is the settings you apply afterwards, and it takes about a minute. The same PDF becomes a mock test by adding a timer and shuffling the order, a question paper by exporting to PDF or Word with the answer key on a separate page, or a graded exam by sharing one link and reading the scores as they come in.

One thing worth deciding up front: a mock test should be harder than the real one or it teaches false confidence. The cheapest way to do that here is not to raise the difficulty setting but to keep the timer tight and to leave in the questions candidates got wrong last time. If you want the full walkthrough for exam rehearsal specifically, the PDF to practice test page goes through it, and PDF to test covers the graded classroom version.

PDF to MCQ: frequently asked questions

How do I convert a PDF to MCQs?

Upload your PDF on this page and click generate. The AI reads the document, pulls out the key terms and facts, and writes multiple choice questions with one correct answer and three distractors, plus an answer key, in seconds. You can then edit any question and export the set to PDF or Word. There is nothing to install and no formatting work on your end.

Can AI generate MCQs from a PDF?

Yes. PDFQuiz uses AI to read the content of your PDF and generate multiple choice questions based on what the document actually says, including definitions, dates, names and concepts. It writes the stem, marks the correct answer and drafts three plausible distractors, then leaves every item open for you to edit before you export it.

How many MCQs can I generate from one PDF?

You set the number. A rough guide is three to five solid MCQs per page of substantive content, so a ten page PDF might yield thirty to fifty questions and a full textbook chapter many more. There is no page cap, so you can convert a long document at once and regenerate for a fresh batch later to grow your question bank.

Can I edit the multiple choice questions it generates?

Yes, every part of every question is editable. Rewrite the stem, replace a distractor that is too easy to rule out, change the correct answer, adjust the wording or delete an item entirely. The AI gives you a strong first draft and you keep full control over the final MCQ set before exporting.

Does the PDF to MCQ converter work with scanned PDFs?

It works best with PDFs that contain real, selectable text. A scanned PDF is a picture of a page rather than selectable text, so it has to be read visually, and the generator does that directly. You do not need an external OCR tool or to retype anything first: turn the OCR option on and upload the scan as it is. Legibility is what decides the result: a clean scan or a straight, well lit photo produces solid MCQs, while a dark or skewed one produces fewer.

Can I export the MCQs to Word, Canvas or Moodle?

PDFQuiz exports your multiple choice questions and answer key to PDF or Word, not to a Canvas or Moodle import package. To use them in your LMS, generate the set here, then paste the questions into your LMS quiz builder or attach the exported file. The hard part, writing the questions from your PDF, is already finished.

Why did my PDF produce bad multiple choice questions?

Usually the text came out of the PDF in the wrong order. A two-column academic paper extracts as interleaved lines unless the file is tagged, so sentences from the left and right columns get spliced together and the generator writes questions from text that was never a sentence. Copy the PDF into a plain text editor to see exactly what a converter sees, then export a single-column copy or upload the source Word file instead.

How many pages of PDF do you need for a good MCQ set?

About one page of dense body prose per two to three questions. A ten page chapter comfortably supports twenty to thirty items. Asking for forty questions from three pages forces the generator to test trivia and near-duplicates, because it runs out of assertions worth testing and starts rewording the same fact in a new stem.

What makes a good multiple choice question?

A good MCQ has a clear stem, one defensibly correct answer, and distractors that are plausible enough that someone who has not learned the material could believe them. It should test understanding rather than trick the reader. PDFQuiz drafts each question to that standard, and you can sharpen any distractor that reads as obviously wrong.

Which parts of a PDF make the best MCQs?

Body prose that states definitions, causes, procedures, and comparisons produces the strongest items. Tables and figure captions produce good factual questions. What produces weak questions is front matter, reference lists, and pages that are mostly headings, because there is no assertion to test. If a chapter opens with ten pages of introduction, upload only the section you actually taught.

Can you make a printed test from the generated MCQs?

Yes. Export the set to PDF and print it, keeping the answer key as its own file rather than a hidden column in the same document. Roughly ten to twelve four option items fit on a US letter page at 11 point with normal margins. The printable test maker covers the page layout, and the answer key generator handles the separate key.