When the question is on your screen rather than spoken, a screenshot is faster than retyping it: press Ctrl+Shift+S and CoPilot Interview reads the problem from the screen, using GPT-6 Luna on the Pro plan. A shared coding editor, a SQL table, an architecture diagram or a photo of a whiteboard are all cases where the interviewer's voice does not contain the question, so captions alone miss it. This page covers how screenshot reading works, what to capture, how it combines with Coding mode, and how to practice with it. As with every live feature, use it only where it is allowed and follow your employer's rules.
When a screenshot beats listening
Live captions work when the interviewer says the question. In technical rounds they often do not. They paste a problem into CoderPad or a shared document, show a table and ask for a query, or draw boxes and ask what breaks at ten times the load. The spoken part is "take a look at this," which tells the assistant nothing.
A screenshot carries the whole problem at once: constraints, examples, column names, arrows. Retyping a 12-line problem under time pressure takes a minute and introduces typos; a screenshot takes one hotkey.
How screenshot reading works in CoPilot Interview
Screenshot reading is a Coding mode feature and needs a model that reads images; today that is GPT-6 Luna on Pro. The hotkeys below are the Windows versions; use Cmd instead of Ctrl on macOS.
| Hotkey | What it does |
|---|---|
| Ctrl+Shift+C | Turn Coding mode on |
| Ctrl+Shift+S | Read a problem from the screen |
| Ctrl+Shift+E | Answer the next question in words while Coding stays on |
| Ctrl+Shift+Q | Add a part the interviewer tacked on mid-answer |
| Ctrl+Shift+F | Keep the assistant window from taking focus |
Coding mode returns code with Big-O and edge cases. When the interviewer follows up with "why that data structure," Ctrl+Shift+E gives you the answer in words instead of more code. There are 11 global hotkeys in total; these five cover coding rounds.
What to capture
The model reads what is on screen when you press the hotkey, so make the whole problem visible first. Scroll so the constraints and the examples are both showing; a problem without its constraints produces a solution to a different problem. For SQL, include the column names and types, not just the question. For a diagram, include the legend and any numbers written on arrows.
Do not capture half a problem and hope. If it does not fit on one screen, reduce the browser zoom once, then capture. Our coding interview copilot page lists the editors and platforms this works with.
Whiteboard and system design diagrams
Hand-drawn diagrams are the hardest input: boxes labeled in a hurry, arrows that cross, numbers that could be 10 or 1,000. The model does well with printed diagrams and acceptably with clean handwriting; it will misread a cramped photo. Say the key facts out loud too, so the captions carry what the image may not.
Our guide to assistants for whiteboard interviews covers the in-person variant, where there is no screen to read and the rules are usually stricter.
Practice with it before any interview
- Open a published practice problem in your browser at the difficulty you expect.
- Press Ctrl+Shift+C to turn Coding mode on, then Ctrl+Shift+S to read it from the screen.
- Read the code, the Big-O and the edge cases, then press Ctrl+Shift+E and explain the approach out loud as if asked.
- Press redo, and compare the second answer with the first.
Ten of these teach you what the model gets right and where it needs your judgment. Our comparison of coding versus conceptual questions explains why the spoken follow-up matters as much as the code.
Where CoPilot Interview fits
CoPilot Interview is a desktop AI interview assistant with Coding mode and screenshot reading built in; our guide to which AI model to pick explains why GPT-6 Luna handles the image. Use it to practice on public problems first, and in a real interview only where the employer allows live help. Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.
Frequently asked questions
Can an AI interview assistant read a coding problem from a screenshot?
Yes, if the model behind it reads images. In CoPilot Interview, GPT-6 Luna on the Pro plan reads screenshots: press Ctrl+Shift+C to turn Coding mode on, then Ctrl+Shift+S to read the problem from the screen. The answer returns code with Big-O and edge cases.
What should be visible when I take the screenshot?
The entire problem: statement, constraints and examples together. For SQL, the column names and types. For diagrams, the legend and any numbers on arrows. A capture missing the constraints produces a solution to a different problem, so scroll or reduce zoom once so everything fits before pressing the hotkey.
Does screenshot reading work on the free plan?
No. Screenshot reading uses GPT-6 Luna, which is on the Pro plan at $39.99 a month. The free plan includes Mercury 2.5, 15 minutes of live help a day and AI mock interviews, which is enough to practice spoken coding questions and the explain-your-approach follow-ups.
Can it read a hand-drawn whiteboard?
Partly. Printed diagrams and clean handwriting read well; cramped or crossing arrows and ambiguous numbers do not. Say the key facts aloud so the captions carry what the image may miss. In-person whiteboard rounds usually have stricter rules, so check them before using any tool.
Is using screenshots allowed in a coding interview?
That depends entirely on the employer and the assessment platform. Some permit any reference material; many technical screens forbid a second application. Read the invite and the platform's candidate agreement, ask the recruiter if they are silent, and follow your employer's rules. Practicing on public problems is always allowed.
Try Coding mode on a practice problem
Run practice questions in the browser with no signup, then test screenshot reading in the desktop app.
Try a practice question