New AI Makes Eye Tracking Smoother and More Accurate

Eye tracking sounds cool, but there's a problem: your eyes naturally jitter and jump around, which makes an on-screen cursor shake. This is a big issue for people who rely on eye tracking to type, click, or communicate — like those with ALS or severe paralysis. A new research system called GazeFS aims to fix that by predicting where your eyes are headed, not just where they are right now.

The system works like a smart passenger who looks at the road and anticipates turns. Instead of reacting to every tiny eye twitch, GazeFS watches recent eye and head movements to guess the next target. This lets it smooth out the twitchy jumps, while still moving quickly toward the thing you actually want to look at. In tests with 30 people watching thousands of on-screen targets, the new method reduced aiming errors and made the cursor feel more stable.

The improvements sound small — roughly 0.2 to 0.4 degrees of angle — but that's significant at a normal screen distance. Think of moving your mouse through jelly: you can still get there, but frustration builds with every wobble. Removing that wobble means a person might select the right key or button the first time, without corrections. That's a big deal when every click requires time and concentration.

This is still academic research, not a product yet. But it points toward a future where eye tracking feels as natural as using your hand — for everyone, not just specialists. As VR headsets and assistive tech grow, having better, more forgiving eye tracking could make digital interactions faster and less exhausting for millions of people.