Look at the Lens When You Open and When You Land

Who Is Watching Your Eyes, and Who Isn't?

Two different readers meet a recorded interview, and they take in different things. HireVue's 2025 explainability statement says its AI relies only on what the candidate says. The same document adds that it uses no video analysis, and that it does not assess facial expressions, body language, background or tone of voice ¹. By the vendor's own account, no video analysis means eye contact is not an input on that path.

Vervoe's candidate help article says the same about its own scorer. On audio and video questions the AI reviews the transcript only ².

Two statements, both written by the company whose scorer they describe. Neither is an audit, and neither reaches a platform that has published nothing.

Vervoe's page carries the second half of the answer too. The video and audio are transcribed and handed to the employer along with the recording ². HireVue's candidate FAQ says its interviews are intended as a first step that helps recruiters make a first cut, with hiring decisions made by people after that ³. That is the design, not a promise about any one employer.

So the score can be blind to your face while the employer is not.

For a platform that publishes nothing, there is no published answer to check. Researchers who decomposed their own automated video interview models, built from mock interviews with students and crowd workers, found verbal behavior weighted highest for four of the five personality traits, with extraversion the exception ⁴. On extraversion, the share given to words swung widely between samples ⁴.

Those were academic models rather than any vendor's, and the instability is the finding.

Finding out what your own interview scores comes first, and What an AI interview scores, and what it does not covers how. It also names the notice the law asks an employer to give when an AI tool is used on a role based in Illinois or New York City.

Where Should Your Eyes Go While You Answer?

Meet the lens as you start an answer and again as you land it, and let your eyes drift where they naturally go while you think. That reads as a person talking to someone. A locked stare at a small black dot for two minutes reads as strain instead, and it takes attention you need for the answer itself.

The reason to bother at all is the person who plays the recording back.

In an experiment with 114 participants, interviewees were rated lower in a videoconference interview than face-to-face ⁵. Differences in perceived social presence, perceived eye contact and impression management contributed to that gap ⁵. Ratings made live were also higher than ratings made from a recording ⁵.

That was a 2021 lab study comparing formats, not a test of eye-contact coaching, and it says nothing about what an algorithm reads. It does say human raters noticed.

Thinking time is allowed to look like thinking time. Eyes up and away for a beat before a hard answer is what people do in a room, and nothing cited here ties a brief glance to a lower rating.

Eye-line discipline has a cost when it is overdone. A candidate who spends the whole answer managing an eye-line arrives at the end with a smooth delivery and a thin example, and the example is the part a second viewer can use. Getting the camera right once, before the call, is what keeps the question off your mind while you are talking. The eye contact then happens on its own, which is how it happens in a room.

Set it up, then forget about it.

Put the Lens Where Your Notes Are

Camera placement does the work that gaze discipline is trying to do. A laptop flat on a desk aims its lens up at you from below, so looking at the interviewer's face means looking well under the camera. No amount of practice closes that angle. Raise the machine until the lens sits near eye level, put the face you are watching just beneath it, and the two directions almost converge.

• Stack books or a box under the laptop until the lens is level with your eyes
• Drag the video window to the top of the screen, directly beneath the lens
• Keep notes to a few words in large type, in that same strip near the camera
• Skip the second monitor for this call, because a glance sideways travels a long way off-lens

The whole fix costs ten minutes and nothing else.

Once the gap is an inch, the difference is hard to see on a recording.

A second screen is the setup that makes this hard to solve, since the angle between the two screens is far wider than the gap between a lens and the window beneath it. If the call has to run that way, put the interviewer on the screen with the camera in it and leave the other one dark. The rule underneath is the same in every setup: whatever you look at most should sit under the lens.

In a recorded round there is nobody to look at, so the lens is the natural place for your eyes. The self-view is the thing most likely to pull them down, and turning it off helps where the platform allows it. Preparing for a one-way video interview in an hour covers the rest of the mechanics worth checking before you press record.

Do You Need a Gaze-Correction Tool?

Probably not, and the reason sits in who such a tool is aimed at. Software that redraws your eyes so they appear to meet the lens changes nothing about the transcript, which is what both vendors above publish that they score ¹². It is aimed at the human viewer, and a camera at eye level reaches that same viewer with no third party inside the recording.

Practice tools grade delivery, which is what a viewer notices. Big Interview's VideoAI page says the tool evaluates pace of speech, vocabulary, eye contact and body language in a recorded practice answer ⁶. That is a product page with no method or accuracy figure behind it.

Of that list, three are things HireVue states its own scorer does not use ¹.

Practicing against delivery feedback prepares you for the person more than for the transcript.

The other chair has been told something about this too. The employer-side piece on a candidate whose eyes keep moving off-camera argues that gaze does not track deception. It adds that off-camera looks can also come from low vision, autism, a second monitor or notes written before the call, so acting on them risks screening on a disability.

Neither published scorer reads your eye-line, and a person watching has the answer itself to weigh. Whether that answer had a real decision in it is the part worth preparing. Set the camera once, and the rest of your attention goes where somebody is actually listening.

See how it works >