Body Language for AI Interviews: What Actually Matters in 2026
In 2026, the leading AI does not score your facial expressions. But a human will watch your video afterwards, and they do.
This is the reframe that changes how you should prepare. Most body language advice for AI interviews is based on the assumption that the AI watches your face. HireVue removed facial expression scoring in 2021. Sapia never had a visual component. But the human reviewer who watches after the AI scores it very much does see your face, your posture, and your gestures. You have two audiences. Prepare for both differently.
Two Audiences, Two Priorities
For the AI (what it measures)
- ■ Speaking rate (120-150 words per minute target)
- ■ Clear articulation (not mumbling or trailing off)
- ■ Filled pauses minimised (too many “um” reduces fluency scores)
- ■ Structured response (STAR shape produces better speech structure scores)
That is it. Everything else in your body language is for the human reviewer.
For the human reviewer
- ■ Eye contact with the camera lens
- ■ Calm, upright posture
- ■ Natural facial expression (genuine, not fixed smile)
- ■ Hand gestures that track your points
- ■ No fidgeting or distracting movement
Eye Contact: Look at the Lens, Not Your Face
The instinct is to watch your own face in the preview window on screen. From the human reviewer’s perspective, this looks like you are looking down and to the side, not making eye contact. The camera lens is the point of eye contact on video.
The sticky-dot technique
Put a small sticky note or a piece of coloured tape immediately beside your webcam lens. Look at the dot when you are speaking. This trains your gaze to the lens rather than the preview. Most built-in webcam lenses are the tiny dark circle at the top-centre of your screen bezel.
Aim for 50-70% eye contact with the lens (similar to human interview norms). Occasional glances down at your notes or to the side while thinking look natural and are not negative. A fixed unblinking stare at the lens reads as uncomfortable. Natural variation is the goal.
Posture: Upright but Not Rigid
Sit fully back in your chair rather than perching on the edge. Shoulders back and relaxed, not hunched or pulled back tightly. 2-3 feet from the camera. A slight forward lean (10-15 degrees) on a key point conveys engagement. Returning to upright conveys composure.
Good posture habits
- Sit fully back in the chair
- Shoulders relaxed, not hunched
- Head level, not tilted
- Slight forward lean on key points
- Feet flat on the floor (reduces fidgeting)
Posture problems to avoid
- Perching on the edge of the chair (looks nervous)
- Slouching (looks disengaged)
- Head tilted significantly to one side
- Leaning too close to the camera
- Swivelling on an office chair (constant motion is distracting)
Hand Gestures: Visible and Natural
Keep your hands visible in the lower portion of the frame. Gestures that track your points (counting items, showing a before-and-after, an open-palm emphasis) look natural and add clarity. Hiding your hands under the desk reads as closed or anxious on camera.
If you naturally talk with your hands, keep doing it - suppressing natural gestures creates visible tension. If you do not naturally gesture, do not force it. Unnatural added gestures read as performed. The goal is authenticity within a composed frame.
Facial Expression: for the Human Reviewer
A genuine smile at the very start of your first answer (“Hi, thanks for the opportunity to tell you about this”) and at the very end of your last answer sets a warm, professional tone. Between: natural variation in expression is correct. A fixed smile throughout the interview reads as performative and odd to the human reviewer.
During thinking time (the countdown before you start speaking), a focused, thoughtful expression is appropriate. You are allowed to look like you are thinking. Neutral attentive face while listening to question playback is natural. Animated expression during your answer where your story warrants it (genuine enthusiasm, concern, pride in a result) is genuine and positive.
Voice Pacing and Prosody
Speaking rate target: 120-150 words per minute. This is the range that reads as confident and considered without being rushed or slow. Practise timing yourself on a 90-second answer: aim for 180-225 words.
4-7-8 breathing before recording
Diaphragmatic breathing (from the belly, not the chest) before you hit record gives you a fuller, stronger voice. The 4-7-8 technique (inhale 4, hold 7, exhale 8) is specifically effective at reducing the cortisol response that makes voices shake.
Strategic pause before speaking
After the question appears on screen, take 2-3 seconds before you begin speaking. This looks like considered thinking (which it is). It does not look like freezing. Use the pause to mentally identify which STAR story fits.
Tonal variation
Monotone hurts on both AI scoring (pacing/prosody metrics) and human review (disengagement). Vary your pace slightly on key points - slow down for emphasis, return to normal for context. This happens naturally in authentic speech and tends to flatten in over-rehearsed answers.
Common Body Language Mistakes on Camera
Looking at your own face in the preview window instead of at the lens
Leaning too close to the camera (uncomfortably close framing)
Excessive head-nodding while listening to the question playback
Reading clearly off a script (the eye movement pattern is recognisable)
Using a distracting virtual background that glitches with your hair or shoulders
Crossing your arms (reads as defensive on video)
Drinking water mid-answer (wait for a question transition break)
Turning to look at a second screen during your answer
The authenticity question
Over-rehearsed body language reads as performative to human reviewers. The goal is natural, calm, and prepared - not scripted. The AI rewards authentic language markers in your content. The human reviewer rewards the impression of a genuine person who has thought carefully about their answer. Neither audience wants a performance. Practise to build comfort, not to produce a uniform delivery.