How Accurate Is Eye Tracking? What Determines It and How to Evaluate It

The short answer

Eye tracking can be accurate enough to control a computer, select on-screen targets and support accessibility tools. However, its accuracy is not a fixed property. It depends on the type of tracker, the quality of calibration, the lighting, the user's eyes and eyewear, and how much the head moves during use. Dedicated eye trackers that use their own infrared illumination are generally more accurate and stable than tracking that relies on an ordinary webcam. Webcam-based gaze estimation usually works best for coarse interaction, such as choosing large buttons or screen regions, rather than pinpointing small details.

No verified measurements for specific products or systems were available for this guide. For that reason, it does not quote accuracy figures. If you need numbers, check the manufacturer's published specifications and test them under your own conditions, as described below.

What "accuracy" means in eye tracking

Two terms are often confused when eye tracking is evaluated:

• Accuracy: how close the measured gaze point is to where the person is actually looking. A tracker with poor accuracy consistently lands some distance away from the true target.
• Precision: how consistent repeated measurements are while the person looks at the same point. A tracker with poor precision produces a jittery or scattered gaze point, even if it is centered correctly on average.

A system can be precise but inaccurate, which means it is steady but offset. It can also be accurate but imprecise, which means it is centered correctly but noisy. Both matter for gaze interaction. An offset makes you miss targets, while jitter makes it hard to hold a selection. Accuracy is commonly expressed in degrees of visual angle rather than pixels. As a result, the on-screen error grows with viewing distance and screen size.

Factors that affect how good eye tracking is

Factor | Why it matters

Tracker type | Dedicated infrared trackers control their own illumination and image the eye in detail. Standard webcams depend on ambient light and lower-detail eye images.

Calibration | Calibration maps eye features to screen positions for a specific person. A poor or outdated calibration causes systematic offsets.

Head movement and distance | Moving outside the range the system was calibrated for, or sitting too close or too far, can reduce accuracy.

Lighting | Strong backlight, glare, uneven light or low light can degrade the eye image, especially for webcams.

Glasses and contact lenses | Reflections from lenses and frames can interfere with detection of the pupil and corneal reflections.

Screen position | Accuracy is often weaker toward the edges and corners of the screen than in the center.

Time | Gaze estimates can drift during a session as posture changes, so recalibration may be needed.

How to evaluate eye tracking for your own use

Published specifications are usually measured under favorable conditions. Your results may differ, so a practical evaluation should reflect how you will actually use the tracker:

• Calibrate in the same seating position, lighting and eyewear you will use day to day.
• Use a validation step, if the software offers one, that shows the measured gaze against known targets after calibration.
• Check targets across the whole screen, including corners and edges, not just the center.
• Observe both the offset from each target (accuracy) and how steady the gaze point is while you fixate (precision).
• Repeat the check after some time and after shifting your posture to see whether drift occurs.
• Try the interactions you actually need, such as dwell selection on buttons of the size your software uses.

If small targets are hard to hit, the usual remedies are to recalibrate, improve and even out the lighting, reduce glare from glasses, sit within the recommended range, and enlarge interface elements or increase dwell time.

What this means for webcam audio and speech capture

Eye tracking and microphone capture are separate functions, even when one webcam provides both the camera image and the built-in microphone. Gaze accuracy depends on the image of your eyes. Speech quality depends on the microphone, its distance from your mouth, room acoustics and the audio input selected in your operating system or conferencing app. A good eye tracking result says nothing about voice quality, and the reverse is also true. If you use gaze interaction alongside voice input or video calls, evaluate each one on its own. In particular, confirm that the intended microphone is selected as the input device in each application. No evidence was available on whether running gaze tracking affects audio performance on specific devices, so that should be tested directly if it is a concern.