Contrast-detect AF is not trying to maximize overall brightness. It tries to maximize local differences in intensity between nearby pixels or small areas.

When an image is out of focus, detail is blurred and light is spread over a larger area. That usually reduces edge contrast rather than increasing it. In other words, blur tends to smooth transitions, so fine detail and edges become less distinct. As focus improves, those local intensity differences become stronger, and the camera looks for that maximum.

This works well in most real scenes because there is almost always some detail somewhere in the AF area: edges, textures, leaves, hair, text, etc. The amount of contrast needed can be very small.

Where it struggles is when there is little or no usable contrast to measure, such as a blank wall, a cloudless sky, or other low-texture subjects. In those cases the camera may hunt or fail to lock, because there is no clear contrast peak to optimize.

So yes, defocus does spread light from neighboring regions, but that spreading generally lowers local contrast instead of creating a better focus result.