Yes, in almost every one-way video interview you can have notes in front of you. The recording is your webcam frame and your microphone, and a card taped beside the lens sits outside both. What gives people away is never the paper. It is eye-line, eyelids, and the flat cadence of someone reciting a sentence they wrote last night. Fix the geometry and the notes stop existing.
What the recording actually captures
An asynchronous video interview asks your browser for two permissions, camera and microphone, and records what those two devices produce. Your desk is not in the file. Your second monitor is not in the file. Anything outside the frame is outside the evidence, which is why the real question is not whether you may use notes but whether a human reviewer can see you using them.
Two things overrule that. If the invitation or the instructions screen says closed book, believe it, because some language and knowledge tests genuinely are. And if the same invitation bundles a coding or aptitude section in a locked browser, those rules belong to that section, not to the video. Platforms differ and change their defaults, so read the consent screen you click before question one rather than trusting a forum screenshot.
The eye-line problem, in degrees
Sit at a laptop and your eyes are roughly fifty to sixty centimetres from the lens. A card whose text sits five centimetres to the side of that lens costs you about five degrees of eye rotation. Notes lying flat on the desk in front of the keyboard sit twenty five to forty degrees below it, and at that angle you stop moving your eyes and start moving your head.
Direction matters more than distance, and this is the part most advice skips. When you look down, your upper eyelid follows the iris and drops. A half closed eye reads as reading even in a small thumbnail, even to a reviewer who has never thought about why. When you look slightly left or right, the eyelid does not move at all. The iris shifts a couple of millimetres inside an eye that is still fully open, which is what your face does anyway while you are recalling something. That is the whole trick. Not hiding the glance, but making the glance look like the thing your face already does when it thinks.
Duration is the second signal. A flick under a second between sentences is invisible. Three seconds of parked gaze while your mouth is moving is unmistakable, because nobody stares at a fixed point off camera while speaking fluently. Your cards have to be glanceable, not readable, and that constrains what you write far more than where you stick it.
The card layout, drawn
Here is the physical arrangement, seen as if you were sitting in your own chair looking at the laptop. Everything in the top row lives on the screen bezel, taped flat, top edge level with the lens.
| Left of the lens | The lens | Right of the lens |
|---|---|---|
| Cards 1 and 2. Story bullets, numbered large in the top corner so you can find one without shuffling. | Webcam. The top edge of every card sits level with this, never lower. | Card 3 and the facts card. One more story, plus product names, the team name, your one number. |
| Nothing | Browser window dragged to the very top of the screen, so the self-view thumbnail sits as close under the lens as the platform allows. | Nothing |
| Nothing | The desk, out of shot. Water, and nothing else. This is the surface that costs you thirty degrees. | Nothing |
Constraints on the cards themselves. Use matte paper or a large sticky note, never a phone and never a second screen, because a lit rectangle reflects in glasses and the reflection shows up in the recording. Write in marker, capitals about fifteen millimetres tall, so it is legible at arm's length without leaning forward. Keep the outer edge within roughly six centimetres of the lens on a laptop, or ten on an external monitor at seventy centimetres. Then record two seconds of nothing and check the corners of the frame for a white edge, because a card that has crept into shot is worse than no card at all.
One more thing about the self-view. Most people are not looking at the lens, they are looking at their own face in the preview, which is why so many one-way clips have the candidate apparently addressing someone off to the side. Dragging the window to the top of the screen shrinks that error to a couple of degrees for free.
Five bullets, five words, never a sentence
A card holds five bullets. Each bullet holds about five words. Written out, a card for a conflict story looks like this.
- PAYMENTS RETRY, LAST SPRING
- MAYA WANTED THE CRON
- I BENCHMARKED BOTH
- SHOWED HER THE P99
- QUEUE SHIPPED, KEPT HER DESIGN
The version that fails looks like this, and it is what almost everybody writes first: "Last spring our payments retry job kept falling over, and my colleague Maya wanted to keep the cron approach, so I benchmarked both options and showed her the p99 latency."
Three reasons the sentence loses. It takes three or four seconds of fixed gaze to consume, which is the exact signature described above. Recitation has an audible cadence, evenly stressed and suspiciously complete, missing the small self-corrections people make when they are genuinely remembering. And when you lose your place inside a sentence you stall visibly, because the sentence is the unit and you have to find the middle of it again. Lose your place in a bullet list and you just glance at the next bullet.
Bullets work because they trigger stories you already know, rather than scripting stories you do not. If a bullet does not restart you on its own, the story is under-rehearsed and no card design fixes that. Four cards out is the ceiling, because past that the think window turns into a search. Running the company question bank generator against the specific firm and role gives you the dozen prompts most likely to appear, which is what the cards get built from.
Placement options, ranked honestly
| Setup | Eye movement | What the reviewer sees |
|---|---|---|
| Cards flanking the lens at bezel height | About five degrees, sideways, eyelid unchanged | Brief lateral flicks between sentences. Reads as thinking |
| Sticky notes on the bezel just under the lens | Eight to twelve degrees, downward | Workable, and the usual advice, but the eyelid still dips. The weaker option, not the wrong one |
| Notes flat on the desk | Twenty five to forty degrees plus a head tilt | Obvious. This is the movement everybody recognises on sight |
| Second monitor beside the laptop | Twenty to thirty degrees lateral, and sustained | Scanning eyes, plus a bright rectangle in your glasses if you wear them |
| Phone propped above the laptop lid | Small and upward, which is the safest direction | Fine until the screen dims. Disable auto lock first, and check it is not in frame |
The test recording pass
Do this once, in the chair and the lighting you will actually use, before you open the real link.
Record thirty seconds with everything final
Cards taped, window at the top of the screen, lamp where it will be, the clothes you will wear. Answer one real question at your real speaking pace. Any phone camera works, or the platform's own practice question, which is normally labelled as practice and not reviewed, though the instructions screen is worth checking.
Watch it full screen with the sound off
Muting is the entire point. Without words you stop grading your content and start seeing your eyes. Count the glances a stranger would notice. Nought to two across thirty seconds is invisible. Five or more is reading, and no rewording fixes it.
Watch the tip of your nose
The cleanest test there is. If your nose stays still, you moved your eyes. If it swings, you turned your head, which means the card is too far out and needs to come two centimetres closer to the lens.
Now listen with the screen off
Audio only. If you hear complete, evenly stressed sentences with no restarts, the card has too many words on it. Cut a bullet to three words and record again. A little untidiness in speech is what makes it sound like recall instead of recitation.
Check the frame edges and your glasses
Pause on a frame, look at all four corners for a card edge, then at your lenses for a bright rectangle. A reflected document is the one tell that survives perfect eye-line, and it is the reason paper beats a screen.
Move the cards in and repeat once
Same question, cards two centimetres nearer the lens. Compare the two clips muted, keep whichever one you would not have noticed, and leave the tape where it is.
Where OfferPilot AI is the wrong tool
Worth saying plainly, because people often land here from a different question than the one they typed. A prerecorded interview is a poor fit for a live copilot, and not for policy reasons.
- There is no interviewer to listen to. Interview Copilot works from an interviewer's voice arriving through system playback. A one-way interview gives you a text prompt, a countdown, then silence and you. Nothing is speaking.
- Capture exclusion is about screen sharing, not cameras. Our overlay is excluded from screen capture at the operating system level, so it stays out of Zoom, Teams, Meet, OBS and screenshots. A webcam pointed at your face is a completely different channel. That exclusion does nothing about it, and nothing about lockdown browsers, eye tracking, somebody watching you in the room, or a phone propped up as a second camera.
- Generated text arrives as sentences. Which is exactly the format this page has spent a thousand words telling you not to read from. A five word bullet you wrote yourself beats a fluent paragraph you have never said out loud.
- Windows only. We run on Windows 10 and 11, and plenty of people record these on a phone or a Mac.
The useful window is earlier, in the days before the link expires, turning likely questions into stories and stories into five bullet cards. That is preparation, and preparation is the only input this format rewards.
The short version
Notes are usually allowed and almost never the problem. Put them beside the lens at bezel height rather than below it, because sideways glances leave your eyelids open and downward ones do not. Five bullets, five words each, marker on matte paper, four cards maximum. Then record thirty seconds, watch it back with the sound off, and count what a stranger would notice. If the answer is nought to two, you are finished.
Related guides
- HireVue Interview Format: What to Expect, Question by Question
- Pair Programming Interview Tips: What to Expect in a Real Repo
- Prepare Behavioral Interview Answers in One Day: 4-Hour Plan
- Take-Home Follow-Up Interview Questions: The 12 They Ask
- Virtual Onsite Loop: How to Prepare on the Day, Hour by Hour