AI Interview Eye Tracking Explained: What Video Interviews Actually Analyze

TL;DR: AI interview eye tracking became a flashpoint after HireVue claimed to score candidates using facial cues including gaze patterns. In March 2021, HireVue quietly discontinued facial analysis after internal research showed it contributed just 0.25% to job-performance predictions. What remains — and still matters — is basic gaze alignment to your camera. This guide explains what video interview AI actually analyzes today, which concerns are real versus overblown, and the specific camera setup adjustments that make a measurable difference.

---

The company that pioneered AI interview eye tracking killed it.

In March 2021, HireVue — the platform that had been scanning candidate faces for 58 behavioral markers including gaze patterns — announced it was discontinuing facial analysis. The reason: their own internal research showed facial expression data contributed less than 0.25% to job-performance predictions. For a feature that had generated years of candidate anxiety, that's a remarkably small number.

This context matters before you spend time trying to optimize your eye behavior for an AI that, in most cases, isn't measuring it anymore. But "most cases" isn't "all cases," and the reality of video interview AI analysis in 2026 is more nuanced than either the "AI is watching your every blink" panic or the "it's all fine" reassurance.

---

What Is AI Interview Eye Tracking?

AI interview eye tracking refers to computer vision systems embedded in video interview platforms that track where a candidate's eyes are pointed, how long they maintain camera-directed gaze, and gaze movement patterns during an answer.

The original pitch from vendors like HireVue was that gaze behavior — maintaining eye contact, not looking away when asked difficult questions, avoiding excessive blinking — correlated with traits like confidence, honesty, and engagement. These systems analyzed video footage frame-by-frame and fed gaze metrics into broader behavioral scoring models alongside speech patterns, word choice, and vocal tonality.

In theory: candidates who looked directly at the camera, maintained steady gaze, and showed minimal gaze aversion scored higher on automated screens.

In practice: the science behind this was always thin, and the consequences for candidates whose natural gaze patterns differed from the training data baseline were real.

---

The HireVue Eye Tracking Reversal

HireVue's 2021 decision to discontinue facial analysis is the most important development in this space — and the one least covered by "how to ace your HireVue interview" articles.

The timeline: EPIC (Electronic Privacy Information Center) filed an FTC complaint against HireVue in 2019, arguing that the facial analysis system lacked scientific validation and posed bias risks. HireVue disputed the characterization but commissioned an independent audit. The audit, conducted by O'Neil Risk Consulting, found the facial analysis component provided minimal predictive value. HireVue discontinued the feature in March 2021, per SHRM coverage, and Fortune reported that the platform would continue scoring candidates based on language and audio features only.

So: if you're doing a HireVue interview today, facial expression and eye tracking are not being scored. The system analyzes your words and voice. This also means the extensive advice you'll find about "maintaining 80% eye contact with the camera during HireVue interviews" is addressing a system that no longer exists in its described form.

For details on what HireVue does still monitor — including screen recording and tab switching — see our guide on what HireVue can see during your interview.

---

What AI Interview Facial Analysis Still Actually Measures

HireVue's reversal doesn't mean facial analysis has disappeared from video interview technology broadly. Several other platforms continue to use it, and their usage varies.

What some platforms still analyze:

• Emotional valence scoring — classifying facial expressions as positive, neutral, or negative using models trained on facial action coding system (FACS) data
• Engagement proxies — head movement, presence/absence of visible face in frame, gross gaze direction (looking at screen vs. looking away)
• Presentation energy — some systems combine facial expression variance with speech energy to estimate "enthusiasm" scores

What no platform reliably measures:

• Intention behind gaze patterns. Looking to the left doesn't reliably indicate deception, regardless of what pop-psychology body language guides claim.
• Cultural gaze norms. Eye contact behavior is heavily culture-dependent — what reads as confident engagement in a US interview context may read as aggressive or disrespectful in others.
• Individual variance. Autistic candidates, candidates with visual impairments, and candidates on ADHD medication all have atypical gaze patterns that bear no relationship to competence.

A 2025 peer-reviewed study published on arXiv (2505.12114) used GAN-generated counterfactual videos to demonstrate that AI interview scoring systems rated women, older candidates, and African-American candidates lower for identical performance — with facial feature analysis being a contributing factor. The bias isn't hypothetical.

AceRound AI focuses on what you actually say and how you structure answers rather than running facial analysis — the tool is designed around coaching response quality, not scoring appearance. For a guide on avoiding detection-related anxiety during AI-assisted interviews, our piece on whether AI interviews detect cheating covers the surveillance scope in detail.

---

How Video Interview Eye Contact AI Works in 2026

Setting aside the discontinued HireVue feature, there's a practical question: if eye tracking isn't being scored by AI, why do so many candidates report feeling like their eye contact "tanked" an interview?

The answer is human reviewers, not algorithms.

Most AI video interview platforms use automated screens as a first filter, then pass candidates who clear the threshold to human reviewers who watch the recorded video. Those human reviewers — hiring managers, recruiters, HR staff — absolutely do form impressions based on eye contact. Not because of algorithmic measurement, but because of ordinary human social cognition.

Video interview eye contact AI tools that exist today generally fall into two categories:

Eye-contact correction software (Nvidia RTX Eye Contact, Zoom Appearance Studio, Apple Center Stage): These use AI to synthetically redirect your gaze in real-time so that you appear to look directly at your interviewer even when you're reading notes or looking at a different screen region. They're increasingly used by candidates and by experienced interviewers alike. Whether using these tools is appropriate for job interviews is a legitimate debate — they're certainly effective.

AI interview coaching tools: Platforms that analyze your practice interviews and give feedback including gaze direction. AceRound and similar tools in this category typically give you a prompt, record your practice response, and surface whether you looked away during key moments. This is genuinely useful feedback — not because AI is scoring your gaze during the live interview, but because gaze aversion is a real conversational signal that human reviewers notice.

---

Gaze Detection Interview: The Bias and Fairness Problem

Gaze detection in interview contexts has a documented fairness problem that goes beyond HireVue's specific implementation.

In Korean AI interview platforms like Saramin's AI screening tool, eye gaze is an explicitly scored metric — "시선 고정률" (gaze fixation rate) is measured and ranked, and HR surveys from Korean companies identify gaze aversion as the second-most-penalized interview behavior. Candidates with naturally different gaze patterns — neurodivergent candidates, candidates with anxiety disorders, candidates from cultures where sustained direct gaze is considered inappropriate — face an algorithmic penalty that has nothing to do with their ability to do the job.

Japanese career media (including Toyokeizai) warns candidates against "目線がうろついていた" (wandering eyes) making them appear "挙動不審" (suspicious) in video interviews. But Japanese candidates also report that over-rehearsed, rigid gaze — staring directly into the camera without normal conversational eye movement — reads as robotic and creates a different negative impression.

In Chinese-speaking markets, candidates on online interview prep forums describe feeling forced to "死死地盯着摄像头" (stare rigidly at the camera) while reciting memorized answers, producing the opposite problem: technically correct gaze behavior but obviously scripted delivery.

The common thread: gaze optimization advice tends to push candidates toward a narrow behavioral template that's calibrated to Western, neurotypical, high-context conversational norms — and penalizes everyone who naturally falls outside that template, regardless of their competence.

The most practical stance: optimize for natural, engaged gaze that would read well to a human reviewer, not for a specific algorithm that likely isn't measuring it anyway.

---

Interview Eye Contact Tips That Actually Work

Given everything above, here's what actually matters for eye contact in video interviews.

Camera position is more important than gaze behavior. If your laptop camera is below your eye level — which it is for most people using a laptop on a desk — you naturally look down at the screen, which reads as gaze avoidance to human reviewers. Elevate your laptop or external monitor so the camera is at eye level. This single adjustment eliminates the most common gaze problem without any behavioral change on your part.

Look at the camera lens during key moments, not your own face. Most video call interfaces show your face in a small thumbnail. New video callers habitually watch themselves instead of the camera. The camera lens is where you need to direct your gaze when delivering important points — especially the opening and closing of your answers. Looking at the interview software's face view of your interviewer is acceptable during the middle of an answer.

Don't stare. Unblinking, fixed gaze at a camera lens reads as unnatural and intense. Normal conversational gaze includes looking away briefly to "think" — this is neurologically normal and socially expected. The problem is excessive gaze aversion, not any gaze aversion at all.

Lighting matters as much as gaze. If you're backlit (window behind you), your face is shadowed and your eye area is unclear. Front-facing, diffused light (ring light, natural window light in front of you) makes your face — including your eyes — visible and well-defined. A visible, well-lit face is easier for both humans and algorithms to process positively.

Notes placement if you're using them. If you're using an AI interview copilot or reading from notes during an interview, position the text as close to your camera as possible — ideally in a narrow column at the top of your screen, directly below the camera. The gaze deviation from camera to notes is minimized, and brief glances become nearly imperceptible.

---

FAQ

How accurate is eye movement tracking in detecting candidate authenticity?

Not accurate enough to use in hiring decisions, which is why HireVue discontinued it after internal validation showed it contributed less than 0.25% to job-performance predictions. Human reviewers watching video also can't reliably distinguish "deceptive gaze" from "cultural gaze norms," "anxiety gaze," or "ADHD gaze" — all of which produce similar behavioral patterns with completely different causes.

Can eye tracking detect if a candidate is reading from a script during the interview?

Potentially yes — if you're reading from a second monitor or document placed far from your camera, gaze direction becomes visibly inconsistent with talking to an interviewer. This is detectable even without specialized eye-tracking software, by any human reviewer watching the recorded video. Solution: position your notes directly below or beside your camera, not on a separate screen.

Does eye tracking work for candidates wearing glasses or in low lighting?

Poorly. Glasses create reflections that confuse gaze-tracking algorithms, and low lighting degrades facial landmark detection accuracy significantly. This is one reason gaze-based scoring produces inconsistent results even when the underlying intent to look at the camera is identical — environmental factors create too much variance.

How do I keep eye contact while reading AI prompts during an interview?

Use a teleprompter-style window positioned directly under your camera lens. Several AI interview coaching tools and teleprompter apps support this layout. The key is keeping the reading distance from the camera as small as possible — a narrow column of text at the top of your screen, with large enough font that you can read with a brief downward glance rather than a full head tilt.

Does HireVue record your screen?

HireVue records your camera and audio. Screen recording capabilities depend on the specific employer configuration, but standard HireVue video interviews capture video and audio only — not your screen content. For a detailed breakdown of what HireVue captures in different interview modes, our guide on what HireVue can see covers this in full.

Can eye movement tracking be biased or unfair to certain candidates?

Yes — documented in peer-reviewed research. A 2025 arXiv study found AI interview scoring systems penalized women, older candidates, and African-American candidates for identical performance. Neurodivergent candidates and candidates from cultures with different eye-contact norms face similar algorithmic penalties. This is part of why HireVue's own audit led to discontinuing the feature, and why regulators in several jurisdictions have begun requiring algorithmic bias audits for AI hiring tools.

---

Author · Alex Chen. Career consultant and former tech recruiter. Spent 5 years on the hiring side before switching to help candidates instead. Writes about real interview dynamics, not textbook advice.