QExtract
Turn a PDF exam paper into a quiz, worksheet or flashcards. Drop in a PDF, DOCX or TXT. QExtract finds the questions, the answer options and the answer key, lets you correct anything before you use it, then exports a practice test, a printable worksheet, a Google Form, or an Anki deck. Your documents never leave your computer.
How it works
Four steps, none of which involve uploading anything.
- Drop in your document. PDF, DOCX, DOC or TXT — several at once if you want them merged into one bank. Scanned pages are read with on-device OCR, so a photocopied paper still works.
- QExtract reads the structure. It detects question numbers, answer options, answer keys, sections, marks and passages — including multiple choice, true/false, fill in the blank, match the following, assertion–reason, internal choice and case-study questions.
- You review and correct. Everything extracted is shown for editing first. Anything the parser was unsure about is flagged, with the exact words it could not find in your document.
- Export what you need. An interactive quiz, a printable worksheet, an answer sheet, a Google Form, an Anki or Quizlet deck, or an LMS import.
What goes in, what comes out
One paper in. As many usable formats out as you need — and no retyping in between.
Reads
Documents you already have — including question sets you exported from something else.
Writes
For the classroom, for the LMS, and for the student revising on their own.
What you can make from one paper
Practice tests, in versions
Build a new paper from your extracted questions following the original's structure — sections, marks and question counts. Up to four versions with questions and options shuffled, and each version's answer key rebuilt to match.
Printable worksheets
Classroom-ready worksheets and bubble answer sheets, in full, half or quarter length.
Flashcards and study decks
Anki and Quizlet decks, plus a built-in study mode that brings back the questions you got wrong sooner and spaces out the ones you know.
Google Forms and LMS formats
Google Forms for online assignment, plus GIFT, QTI, H5P and CSV for Moodle, Canvas, Kahoot and other learning platforms.
The hard part isn't reading a PDF
It is that a question paper's meaning lives in its layout. Numbering restarts at every section. Options are labelled a dozen different ways. Fractions are drawn as vector strokes with no characters behind them. “Attempt either 21A or 21B” is two questions, not one. And a single lost line break silently welds two questions into one. Getting that right is the entire product.
Maths that survives the trip
Fractions, radicals and formulas are rebuilt from the PDF's own geometry when the page contains no characters to read — then exported as MathML and LaTeX rather than flattened into nonsense.
Papers from 40+ boards
109 exam shapes are tested on every release: CBSE sections and internal choice, Edexcel and SQA A1./B2. numbering, bare-letter and F G H J K answer sets, “points” as well as “marks”, statement lists inside a stem, and papers numbered in ten different digit systems.
Stimulus stays with its question
Comprehension passages, case studies and diagram-labelling intros are attached to the questions that depend on them — so a question still makes sense once it is out of the paper.
Every doubt is shown, not hidden
Low-confidence extractions are surfaced for correction with the reason attached, and the corrections you make are remembered on your device.
Who it's for
Teachers and schools
You already have the papers. What you lack is a way to re-use them without retyping — as a practice test, a worksheet, a Google Form or a Moodle import.
And in a district with rules about where student material may go, “the file never leaves the device” is not a feature. It is the requirement.
Board and competitive exam prep
Built against real board papers: sections with declared marks, internal choice, assertion–reason items, case-study passages with sub-questions, and dense mathematical notation.
A 174-page question bank goes in once and becomes a bank you can re-cut into practice sets as often as you like.
Students revising
Turn last year's papers into a quiz you can actually sit, or an Anki deck. Study mode brings back what you got wrong sooner and spaces out what you already know.
It extracts and organises questions — it is not an answer engine, and it will not do your homework.
Local parsing and connected features
Document parsing runs in your browser. Google Forms and page imports use network access when you choose those features. Review the privacy policy for the details.
Documents parsed locally
Document parsing runs in your browser and can work offline. Sending selected questions to Google Forms is a separate action that you initiate.
No account, no tracking
Nothing to sign up for and no analytics on what you parse. Your review history and study progress stay on your own device and can be cleared in one click.
No AI guessing
Parsing is deterministic, so the same document always gives the same result. When QExtract cannot find something, it flags the question for review rather than inventing an answer.
Built to be accessible
Keyboard controls, a dyslexia-friendly font option, and light and dark themes support different reading needs. Automated checks are part of accessibility review; manual testing is also needed.
The permissions it asks for, and why
| Permission | What it is for |
|---|---|
| Host permissions | No persistent host permissions are declared. A user-initiated page import uses temporary active-tab access. Absence of persistent host access does not prohibit every network request. |
| identity | Only when you choose Push to Google Forms: it signs you in and creates that one form. The token is used for that operation and is not stored. |
| storage | Keeps your question banks, corrections and study progress on your own device. |
| activeTab · scripting | Reads the page you are looking at, only when you deliberately import questions from it. |
| contextMenus · sidePanel | The right-click entry, and the side panel the extension opens in. |
What it does not do
Published because a tool worth trusting is one that tells you where it stops. Each of these is measured before every release and listed in the release notes rather than rounded away.
- It will not invent an answer key. Most question papers print none, by design — the card says so instead of guessing. On the largest bank in the test corpus, 190 of 1,478 questions carry an answer.
- Scans need more review than files. A PDF with a real text layer always beats a photograph of the same page — and OCR is English-only, so a scan in another script returns unusable text.
- The interface is English only, even though the parser reads papers in many scripts.
- Some answer choices still weld onto the question — about 1.6% of multiple-choice questions on the largest corpus paper, down from 19.5% two releases ago.
- Match-the-following with coded options can split into several questions instead of one.
- Chrome only. There is no Firefox, Edge or Safari build today.
- It is not an answer engine. It extracts and organises questions; it does not solve them, and it gives no academic advice.
Questions
Everything people actually ask, answered without marketing.
Privacy and your documents
Are my documents uploaded anywhere?
Documents are parsed locally in your browser. If you choose Push to Google Forms, selected questions are sent to your Google account. Importing a web page can fetch its images. See the privacy policy for these connected features.
Does QExtract work offline?
Yes, completely. Reading, parsing, OCR, quiz building and export all run on your device. The only features that touch the network are ones you start yourself: pushing a quiz to Google Forms, and importing questions from a web page you have open.
What data does it store, and how do I delete it?
Your question banks, corrections and study progress are kept in your browser's local storage on your own device. There is no server, no account and no profile. Clearing the extension's data — or removing the extension — deletes all of it.
Why does it ask for permissions?
identity supports Google Forms sign-in, storage keeps local data, activeTab and scripting support a page import you initiate, and contextMenus and sidePanel provide the interface. QExtract declares no persistent host permissions. This limits ongoing website access; it does not mean every network request is impossible.
Is QExtract free? Why does it ask for a license key?
QExtract is free. It runs for 14 days without a key; after that it asks for a license key, which is free on our license page. Paste it once and QExtract keeps working. The key is checked on your device; nothing is sent to us to check it.
Can I use it on a phone or tablet, or without Chrome?
Yes: QExtract as a web app runs in any modern browser on phones, tablets and computers, and you can install it from the browser's menu ("Add to Home screen" or "Install app"). After the first visit it works fully offline, and your documents stay on your device. Features that need the Chrome extension — Google Forms sign-in and importing a page open in another tab — are not in the web app.
Can I use it for unreleased exams, or under a school data policy?
Local document parsing can help keep material on your device. Review the privacy policy, exported files and connected features against your institution's rules before using confidential material. Permissions alone do not establish compliance with a school data policy.
Files and formats
Which file formats can it read?
PDF, DOCX, DOC and TXT, plus Markdown, HTML, GIFT and QTI files. Scanned PDFs are handled with on-device OCR, and you can import questions from a web page you have open. Several files can be uploaded at once and merged into a single bank.
What can I export?
An interactive quiz, a printable worksheet, a bubble answer sheet, a Google Form, Anki and Quizlet decks, Kahoot CSV, GIFT, QTI, H5P, JSON, and a portable .qex question bank you can reopen later.
Can it read a scanned or photocopied paper?
Yes, through on-device OCR — and OCR output needs more reviewing than a clean typed file. A PDF with a real text layer always extracts better than a scan of the same page.
OCR is English-only by design. A scan in another script will return unusable text; a PDF in another script that has a real text layer is a different matter — see below.
Does it work on papers that are not in English?
For PDFs with a real text layer, yes, for a wide range. Papers numbered in ten different digit systems including Bengali, Devanagari and Arabic-Indic; papers that say Aufgabe, Zadanie, Oppgave or Exercice instead of “Question”; and non-Latin question words in Russian, Hindi, Bengali and Korean all extract.
Two honest caveats: the interface itself is English only, and OCR of a scan is English only.
How many pages or questions can it handle at once?
The largest paper in the accuracy corpus is a 174-page question bank that extracts 1,478 questions in a single pass. There is no page limit — very large files simply take longer, and all the work happens on your own machine.
Accuracy and limits
How accurate is the extraction?
Accuracy is measured against a corpus of real exam papers before every release, and the result is published rather than claimed: across the gated papers, 99.6% of extracted content is traceable back to text that is actually in the source document.
The caveat that belongs with that number: it measures faithfulness to the file, not the quality of the file. Clean typed PDFs work best; scans and badly encoded PDFs need more review.
Does it use AI?
No. QExtract uses deterministic parsing rather than a language model, so the same document always produces the same result and nothing is sent to an AI service. Where it cannot find something, it flags the question for review rather than inventing an answer.
What if my paper has no answer key?
Most question papers print none, and QExtract will not make one up. Those questions still work as a quiz, a worksheet, a printable paper and a bank — the card simply says no answer was found in the source, instead of showing a guess.
What does it get wrong?
Scans and badly encoded PDFs need the most review. About 1.6% of multiple-choice questions on the largest corpus paper still carry an answer choice welded onto the stem. Match-the-following tables with coded options can split into several questions.
Every known gap is listed in the release notes for the release it is still open in.
How do I report a bad extraction?
There is a Report a problem option that builds a diagnostic file describing the structure of what went wrong. It shows you the entire file first, masks identifying content by default, and nothing is sent automatically — you save it, and email it to shubix.connect@gmail.com yourself if you decide to.
Using it
Can it make different versions of the same paper?
Yes. It can assemble a new practice paper from your extracted questions following the original's structure of sections and marks, and produce up to four versions with questions and answer options shuffled, with each version's answer key rebuilt to match — so a printed Version A and Version B never share an answer order.
Is it useful for students, or only teachers?
Both. Teachers use paper assembly, worksheets and the LMS exports; students use the quiz player, the Anki and Quizlet decks, and study mode, which brings back questions you got wrong sooner and spaces out the ones you know.
It is not a homework-answering tool, and it gives no personalised academic advice.
Is it accessible?
Accessibility features include keyboard controls, a dyslexia-friendly font option, and light and dark themes. Automated checks help find issues, but do not establish complete WCAG conformance or replace testing with assistive technology.
Cost and availability
Is QExtract free?
Yes — free on the Chrome Web Store, with no paywall, no subscription, no trial and no account to create.
Is there a Firefox, Edge or Safari version?
Not currently. QExtract is Chrome-only today — which includes Chrome on Windows, macOS, Linux and ChromeOS.
Get it
Free on the Chrome Web Store. Nothing to configure, nothing to sign up for.