Quiz helper apps for students: extensions, bots and hotkeys

Every quiz helper is sold on its features, and the features are all the same: fast, accurate, works everywhere, all question types. They tell you nothing. What actually separates these products is a structural choice made before any of that — how does it get the question? — and there are only three answers.

Once you know which of the three you are looking at, you can predict most of the rest: what it works with, what it can see of your life, what happens when the site changes, and whether it survives a PDF.

Shape 1: the browser extension

It runs inside the web page. The question is already in the page as text, so the extension reads the DOM — the structure the browser built — and sends it off.

Consequences. It is fast and it needs no screenshot, because it is reading text that already exists. It can also see elements you cannot: an answer key sitting in the page’s data, a hidden field, the next question. That is its party trick and its liability.

Where it breaks. Anything that is not a web page: a desktop testing client, a PDF, a slide deck, an image-based question, a locked-down browser. And it breaks when the site changes its markup — an extension keyed to a platform’s page structure is a maintenance relationship with somebody who does not want it to work.

The cost nobody mentions. Read-the-page permission is not granted per question. An extension with access to the sites you visit has it on your email, your bank and your messages too, all the time, in whatever code ships in its next auto-update. That may be fine. It is simply a much bigger permission than the job needs.

Shape 2: the chat window

You copy the question, or take a screenshot, and put it into a chatbot in another tab or another app.

Consequences. Nothing has permission to anything; you are handing over one question at a time, deliberately. It works on any question you can select or photograph, including a PDF and a slide, and you can ask follow-ups — which none of the other two shapes can do.

Where it breaks. The round trip. Copy, switch, paste, wait, read, switch back, find your place. Call it thirty seconds, times forty questions. The attention cost is larger than the time cost: you have to reload the question into your head every time you come back to it.

Shape 3: the screen reader with a hotkey

A small desktop application sits in the system tray doing nothing. You press a shortcut, it captures the window that is currently in front of you as an image, sends that image to an AI, and shows you the answer as a notification.

Consequences. It has no idea what application you are in and does not need to. A browser quiz, a desktop exam client, a PDF, a photographed worksheet and a slide deck are all the same thing to it: a picture of a question. Nothing changes when a platform redesigns, because it was never reading their markup. And you do not move — the thing you were reading is still in front of you when the answer arrives.

Where it breaks. It can only see what is on the screen. A question in a scroll box you have not scrolled to is not in the picture. It is a desktop install, so it is Windows and macOS only — no phone, no Chromebook. And it is one shot: there is no conversation, because there is no window to type a follow-up into.

The permission question, asked properly. “It reads your screen” sounds like the largest permission of the three, and whether it is depends entirely on two details worth checking in any product of this shape. First: when does it capture — continuously, or only in the moment you press the key? Second: the active window, or the whole screen? The narrow answer to both means the only thing ever captured is the thing you were deliberately looking at when you asked.

Which questions to ask, whatever you pick

• Where does the image or text go? To a company’s servers, or straight to an AI provider under your own account? This decides who could read it, not who promises not to.
• Is anything written to disk? A screenshot saved as a file is a screenshot that outlives the question.
• Is it kept? “We do not store your questions” should be checkable against something more than a sentence on a marketing page.
• What happens with no internet? Most of these are a blank screen. A few can fall back to a model running on your own machine.
• What does it cost when you use it a lot? Per-question credits and unlimited plans behave very differently at forty questions an evening.

Where we stand, so you can check it

OneTapAnswer is the third shape, with the narrow answer to both permission questions: it captures only in the moment you press the key, and only the active window — not the screen, not the other monitor, not anything behind. The image is held in memory and sent to the AI you signed into under your own account; it does not pass through a server of ours, which is why we can say we cannot read it rather than promising not to. It answers offline through a local model if you set one up, and the free tier is ten answers a month with no card. The privacy policy lists every third party involved by name, and the docs describe the capture in more detail than most people will want.

Whether using any of this is permitted on a particular test is a matter of your institution’s rules, and worth checking in advance. This one is built for revision, practice papers and self-study.

If you would rather see the shape than read about it, the practice quiz runs in a browser with nothing installed, and the download page has the free tier.