Your Phone Is Making a Lot of Decisions Very Fast
You're at a cafe. You raise your phone toward your friend mid-laugh, and before your thumb even finds the shutter, a little green square has already settled on her face. You didn't tap. You didn't ask. The phone just decided she was the point.
That moment isn't magic. It's three competing systems running simultaneously, each one arguing with the others about what matters in the frame.
Understanding how autofocus works doesn't just satisfy curiosity. It tells you why your camera nails the shot sometimes and completely botches it other times, and what you can actually do about it.
The Three Systems Fighting Over Your Photo
Modern smartphone cameras blend up to three autofocus methods, sometimes all at once.
Contrast Detection is the oldest and simplest. The camera samples pixels across the sensor and hunts for edges: places where brightness or color changes sharply. A face against a blurred background has strong edges at the jawline and hairline. The camera shifts the lens elements, checks whether those edges got sharper or softer, then shifts again. It's like turning an old radio dial until the station comes in clean. Reliable, but slow. On a budget phone in low light, this is exactly why you sometimes watch the image hunt back and forth before it finally commits.
Phase Detection Autofocus (PDAF) is faster and smarter. The sensor carries specialized pixel pairs, each masked so it only receives light from one side of the lens. If the subject is out of focus, those two pixels see slightly different images. The gap between them tells the processor exactly how far the lens needs to move and in which direction, in a single calculation. No hunting. Many mid-range and flagship phones now embed PDAF pixels across the entire sensor, which is why they can track a moving subject across the full frame without losing lock. Worth noting: PDAF pixels are slightly less light-sensitive than standard pixels, so manufacturers place them carefully and use interpolation to fill the gaps.
LiDAR and Time-of-Flight sensors take a different approach entirely. They fire invisible infrared pulses at the scene and measure how long the light takes to bounce back. The result is a depth map: a rough three-dimensional model of everything in front of the lens, calculated before the camera even considers contrast or phase. This is why certain phones lock focus almost instantly in near-darkness. The depth sensor doesn't care about visible light at all.
Most flagships blend all three. The depth sensor sketches a rough map, PDAF gets the lens close, contrast detection fine-tunes the final position.
The Part Where AI Picks a Winner
Knowing the distance to every object in a scene is only half the problem. The camera also has to decide which object deserves focus. That decision happens in software, and it's genuinely interesting.
Scene analysis runs on a dedicated image signal processor (ISP), a chip separate from the main CPU. It classifies regions of the image in real time: face, body, pet, text, food, sky. These classifiers are trained on enormous image datasets, which is why a phone can distinguish a human face from a mannequin face with reasonable accuracy, or recognize that a dog's eyes are the priority subject when you're shooting a pet.
Face and eye detection typically override everything else. If the ISP spots a face, it anchors focus to the eyes, specifically the near eye if the face is turned at an angle. Take a portrait of someone glancing sideways and the camera will hold sharp focus on the eye closest to the lens, even as they move.
No face in the frame? The system falls back to a priority hierarchy: large objects near the center, then objects with strong contrast edges, then the closest thing the depth sensor can find. This is why photographing a flower in front of a fence sometimes focuses on the fence. The fence has stronger edges and sits at a more predictable distance. The camera isn't wrong, exactly. It's just solving a different problem than the one you had in mind.
A Tale of Two Shots
Maya has a two-year-old flagship. She points it at a birthday kid mid-lunge toward the cake. The phone's PDAF locks on the child's face in roughly 100 milliseconds, eye-detection pins focus to the near eye, and burst mode captures twelve frames in two seconds. Eight of them are sharp.
Dan has a budget phone from the same era, contrast-detection only, no dedicated depth sensor. He shoots the same moment. The camera hunts for 400 milliseconds, locks onto the cake (high contrast, centered, predictable), and delivers a beautifully sharp photo of buttercream frosting with a blurry child smeared behind it.
Technically impressive. Not what he wanted.
The difference isn't megapixels. It's the autofocus hardware and the software stack telling the camera what a birthday party actually contains.
What People Misread About Focus Lock
The most persistent misconception: tapping the screen just tells the camera where to point. It's doing more than that. A tap sends coordinates to the ISP, which anchors the entire autofocus system to that region. Tap a face and the camera locks subject tracking to that specific face, following it across the frame. Tap the background by mistake and the camera will faithfully keep the background sharp while your subject blurs as they move.
This is why a lot of candid shots go wrong even when the photographer thinks they did everything right.
Also worth knowing: autofocus and auto-exposure are linked. When you tap to set focus, you're usually setting the exposure point too. The two can be separated on most camera apps by tapping and holding, which surfaces independent focus and exposure sliders. Most people never find this. Have you ever actually used it? If you're doing it regularly, you're already ahead of the majority of smartphone photographers.
One more thing people miss: in very low light, even a phone with LiDAR will sometimes lock on the wrong subject, because the infrared depth map gets noisier as ambient light drops. The fix is simple. Tap directly on what you want. Don't trust the green square when the room is dim.
Sharpness Is a Choice the Camera Makes for You
Every autofocus decision is an editorial judgment, and the camera is making it without asking you. The phone constantly asks: what is this scene about? It answers using hardware, trained models, and a priority system designed by engineers who never saw your specific shot.
Knowing the system won't make you a better photographer by tomorrow. But it means you understand why the camera got it wrong, which is the only way you'll know exactly where to tap to overrule it.