An autofocus point is a small area of the frame that you tell the camera to focus on. The camera measures focus inside that area, drives the lens until that area is sharp, and ignores everything else. Choose the right point and the eye, the owl or the petal you care about is sharp; leave the choice to chance and the camera may focus on a branch in front, the background behind, or the brightest edge it can find.
The owl above is the whole problem in one frame. There are hundreds of crisp edges in it: needles, twigs, the bark of the trunk, the pale branch right in front. Each one is a valid place to focus. Only one of them is the picture. An autofocus point is how you say which.
What an autofocus point actually is
The square you see in the viewfinder or on the screen is a marker for a measuring area. What sits behind it depends on the kind of camera.
- Cameras with a separate AF module (most DSLRs) send part of the light down to a dedicated autofocus sensor under the mirror. That sensor has a fixed number of small detector strips, and each viewfinder point is one of them. The number of points is the number of real detectors, from 9 to over a hundred depending on the camera, and they cannot be anywhere else.
- Cameras with AF on the image sensor (mirrorless cameras, phones, and DSLRs in live view) build focusing pixels into the main sensor itself, spread across most of its area. Here a “point” is a selectable position in a grid. When a specification lists thousands of points, it is counting places you can move the box, not thousands of separate sensors.
Either way the point measures focus with phase detection (which tells the camera both how far out of focus the subject is and which way to move the lens) or contrast detection (which moves the lens and watches for the sharpest result), and many cameras use both. For the full mechanism, see how autofocus works. What matters on this page is the part every system shares: a point can only judge focus from detail inside its own area.
Not sure how autofocus fits with everything else your camera does? Start with what autofocus is and its main settings.
Where the points can reach: coverage
Coverage is how much of the frame has a point you can use. It decides whether you can put focus directly on a subject near the edge, or have to focus somewhere else and move the camera.
A camera with a separate module usually covers the middle half or so of the frame, wider than it is tall. A subject on a thirds line is often within reach; a subject near a corner is not. On-sensor systems usually cover most of the frame, so you can put the point on an eye at the very edge. That one difference changes technique more than any point count does, and it is a large part of why focus and recompose (below) is a habit from the DSLR era.
Why a point needs detail, and which way it runs
Every autofocus point needs contrast inside its area: an edge, a texture, a line between light and dark. Point it at a plain white wall, clear blue sky, smooth water or thick fog and there is nothing to compare, so the lens hunts back and forth and gives up. That is why autofocus usually fails for lack of contrast at the point rather than lack of light. A bright, blank sky defeats it; a dim but textured face often does not.
Line points and cross-type points
Phase detection compares two views of the same small strip of the scene and works out how far apart they are. A strip can only measure detail that changes along its own length. A point whose strip runs left to right reads vertical edges well (a tree trunk, a lamp post, the side of a face) and sees almost nothing in a horizontal line, such as a horizon, the slats of a blind or a striped shirt, because along that strip the pattern looks the same everywhere.
- Line (single-axis) points have strips in one direction only. They work on detail that crosses them and miss detail that runs along them.
- Cross-type points have strips in both directions, so they read horizontal and vertical detail. On a separate AF module, usually only some points are cross-type, and the center point is almost always one of them. Some cameras also have dual-cross center points with extra diagonal strips.
- On-sensor phase detection mostly uses pixels split left and right, which means it reads vertical edges best and is weak on pure horizontal lines. Some newer sensors split pixels both ways, and contrast detection helps fill the gap, but “every point is cross-type” is not true of most mirrorless cameras.
The fix in practice is simple. If focus hunts on a horizontal pattern, move the point to an edge that runs the other way, or tilt the camera about 45 degrees, focus, then level it again.
Sensitivity: why the center point is the strongest
On cameras with a separate AF module, how well a point works depends on the lens’s maximum aperture. Most points are designed to work with lenses of f/5.6 or brighter; the center point usually has extra, more precise strips that only switch on with a lens of f/2.8 or brighter, and it is often the only point that still works at f/8, for example with a slow telephoto and a teleconverter. That is the real reason “use the center point in low light” is good advice on those cameras. It does not apply in the same way to on-sensor systems, where the off-center positions are usually close to the center in performance.
Want to know why two strips of light can tell the lens which way to move? Phase detection autofocus explains the split images.
Point size: big enough to grab, small enough to aim
The box you see is not a pin. It covers an area, and on many cameras the area the point actually reads is a little larger than the box drawn on screen. If the point is larger than your subject, it may focus on whatever in that area has the strongest contrast. With an eye, that can be the eyelashes, the eyebrow or the edge of a pair of glasses. With a small bird in a bush, it is often the twig in front.
Many cameras let you choose a smaller “pinpoint” or “spot” size. It is slower and needs some contrast, but it lets you put focus between branches, through a fence or on one eye. Use the normal size for everyday subjects and drop to the small one when something nearer is stealing focus.
Which points are active: area modes in one minute
Most cameras split focusing into up to three separate settings, and it helps to keep them apart:
- Focus mode says when focus happens: once and lock (AF-S or One Shot) or continuously while you hold the button (AF-C or AI Servo).
- AF area mode says where the camera may look: one point, a point with helpers, a zone, the whole frame, or a tracking box.
- Subject detection says what to look for: faces and eyes, animals, birds, vehicles, or nothing in particular.
The area modes, from tightest to loosest:
- Single point: What the camera does: Focuses only on the point you place. When it fails: A moving subject slides off the point.
- Expanded or dynamic point: What the camera does: Uses your point, and borrows the neighbors if the subject briefly leaves it (in continuous focus). When it fails: Something nearer crosses into the neighbors.
- Zone: What the camera does: Picks from a block of points you place, usually the nearest subject in it. When it fails: The block also holds a nearer branch, fence or foreground.
- Wide or auto area: What the camera does: Picks from the whole frame: usually the nearest, largest or most contrasty thing, or a detected face. When it fails: Your subject is not the nearest or biggest thing.
- Tracking: What the camera does: Starts at your point, then follows the subject’s shape, color or detected face across the frame. When it fails: The subject is small, or its color matches the background.
Choosing among these for portraits, sport, birds or landscapes is its own subject, covered with examples on how to choose focus points. The idea underneath all of them: the fewer points the camera may choose from, the more control you have and the less it can help with movement.
Is your subject moving, so the point keeps losing it? AF tracking follows a subject across the frame.
Focus and recompose: when it costs sharpness, in numbers
Focus and recompose means putting the center point on the subject, half-pressing (or pressing the back button) to lock focus, then turning the camera to frame the shot. It is quick and it uses the most sensitive point. But it has a cost that is easy to see once you draw it.
Focus sets a flat plane of sharpness at a fixed distance in front of the lens, square to the lens. When you turn the camera, that plane turns with it. Your subject stays where it was, so it is no longer on the plane: it is now in front of it. The distance it ends up in front is the subject distance times (1 minus the cosine of the angle you turned through).
Whether that matters depends on how deep the depth of field is in front of the subject. The table compares the shift with the front depth of field for a few common setups on a full frame camera held horizontally. “Thirds line” means the subject ends up a third of the way from the center to the side edge, which is where the vertical thirds line falls; “frame edge” means right at the side.
| Lens, aperture, distance | Sharp zone in front of the subject | Subject on a thirds line | Two thirds of the way out | Subject at the frame edge |
|---|---|---|---|---|
| 24mm, f/1.4, 1 m | 66 mm | 30 mm, sharp | 106 mm, soft | 200 mm, soft |
| 35mm, f/2, 1.5 m | 100 mm | 22 mm, sharp | 81 mm, sharp | 166 mm, soft |
| 50mm, f/1.4, 1.5 m | 36 mm | 11 mm, sharp | 41 mm, soft | 89 mm, soft |
| 50mm, f/4, 1.5 m | 98 mm | 11 mm, sharp | 41 mm, sharp | 89 mm, just sharp |
| 85mm, f/1.8, 2 m | 28 mm | 5 mm, sharp | 20 mm, sharp | 43 mm, soft |
| 200mm, f/2.8, 10 m | 202 mm | 5 mm, sharp | 18 mm, sharp | 40 mm, sharp |
Read across a row and the pattern is clear. Recomposing to a thirds line is almost always safe. Recomposing to the edge with a wide or normal lens at a wide aperture, close up, is not: the angle is large, the distance is short and the sharp zone is thin. Long lenses barely turn at all to move a subject across the frame, so they get away with it. Stop down, and the sharp zone grows while the error stays exactly the same.
How the table was worked out: the angle is the arctangent of the subject’s distance from the center on the sensor (6, 12 or 18 mm on a 36 mm wide sensor) divided by the focal length; the shift is the subject distance times (1 minus the cosine of that angle); the sharp zone uses the standard depth of field formulas with a 0.03 mm circle of confusion. Viewed at 100 percent on a high resolution sensor, the zone is effectively thinner, so treat “just sharp” as soft. You can run your own lens and distance through the depth of field calculator and compare its “in front” figure with the shift.
Two things this changes in practice. First, back button focus does not remove the error. It stops the camera refocusing when you press the shutter, which is useful, but the plane still turns with the camera. Second, the real fix is to move the point to where the subject will be in the final frame, which on-sensor systems make easy almost anywhere.
How thin is the sharp zone with your own lens? Depth of field explains what makes it deep or shallow.
How to move the point quickly
A point you cannot move fast is a point you will not move. Most cameras give you several ways:
- A small joystick or the four-way pad moves the point step by step; pressing it in usually sends it back to the center.
- On many mirrorless cameras, dragging a thumb on the rear touchscreen while your eye is at the viewfinder moves the point. Limit the active touch area to the right half of the screen so your nose does not move it.
- Tapping the screen in live view sets the point directly.
- Many cameras can remember a separate point for horizontal and vertical framing, which saves resetting it each time you turn the camera.
Fewer, larger steps can also help. If your camera lets you reduce how many positions the point cycles through, you can cross the frame in a few presses instead of thirty.
Common mistakes with autofocus points
- Leaving the camera to choose. Wide area modes favor the nearest, biggest or most contrasty thing. If that is not your subject, you get a sharp fence and a soft animal. Pick the point yourself whenever the subject is small, behind something, or not the nearest thing in frame.
- Putting the point on a blank area. Sky, a plain wall or fog give the point nothing to measure. Aim it at an edge at the same distance: a hairline, the edge of a building, a tree trunk.
- Using a point larger than the subject. A normal-size point on a distant bird often lands on the branch beside it. Switch to the smallest point size.
- Recomposing at a wide aperture, close up. As the table shows, a 50mm at f/1.4 moved from center to edge misses an eye 1.5 m away by about 9 cm. Move the point instead.
- Believing more points means better focus. How much of the frame is covered, which points are cross-type, how sensitive they are and how well the camera tracks matter more than the count. Thousands of positions on a sensor are mostly a measure of how finely you can place the box.
- Blaming the lens for focus errors before checking the point. Most soft shots come from focus landing on the wrong thing, not from a faulty lens. Check where the point was (many viewing programs and cameras can show it) before thinking about calibration.
Focus still hunting or landing in the wrong place after all this? How to fix common autofocus problems.
Try this: the 10 minute recompose test
You need a tape measure (or a ruler and a printed page of text), a table, and your brightest lens. A 35mm or 50mm prime at its widest aperture shows the effect best.
- Lay the tape measure on the table so it runs away from you. Stand so that the camera is about 1 m from the mark you will focus on, and set the widest aperture, single point, and single-shot focus (AF-S or One Shot).
- Shot A, recompose: put the center point on the 1 m mark, half-press to lock focus, then turn the camera until the mark sits near the side edge of the frame. Take the picture.
- Shot B, move the point: frame the shot as in A, then move the point to the mark at the edge and take the picture.
- Shot C, let the camera choose: switch to wide or auto area, frame the same way and take the picture.
Judge the results at 100 percent on a computer. In each frame, find the numbers on the tape that are sharpest. In B they should sit right on your mark. In A the sharpest numbers will be a few centimeters farther away than the mark (the focus plane swung back behind it); with a 50mm at f/1.4 and the mark near the edge, expect roughly 6 cm. In C, note what the camera chose. Repeat A and B at f/8 and see the difference shrink into the depth of field.
Ready to test yourself on what you just tried? Take the focus points quiz.
Autofocus points FAQ
How many autofocus points do I need?
Fewer than the specifications suggest. What matters is whether there is a usable point where your subjects usually sit, and whether those points are sensitive enough for your lenses and light. A camera with points covering most of the frame lets you skip focus and recompose, which matters more than the exact count.
Why does my camera focus on the background instead of my subject?
Usually because the point, or the zone, covers more contrast in the background than on the subject, or the camera is in a wide area mode and chose something else. Switch to a single small point on an edge of the subject, such as the line of an eye or the outline of a beak.
Should I always use the center point?
On cameras with a separate AF module, the center point is the most sensitive and the best choice in poor light or with slow lenses, but recomposing from it costs sharpness close up at wide apertures. On cameras with autofocus on the sensor, put the point where the subject is, wherever that is.
Does eye detection make choosing a point unnecessary?
Not quite. Eye detection AF does the job well when a person or animal is clearly the subject. It does nothing for a landscape, a product or a building, and in a crowd it may pick the wrong face, so you still need to know how to place a point yourself.
Why won’t my camera focus on the sky or a white wall?
The point has nothing to measure: no edges, no texture, no contrast. Put the point on something at the same distance that has an edge, such as the edge of a cloud, a distant tree line, a window frame or a light switch, or switch to manual focus.