In more detail
A visible watermark is blended into the image with partial opacity: each covered pixel becomes α × mark + (1 − α) × original, where α is the opacity. Typical choices are a logo or name in a corner, or a repeated diagonal pattern across the whole picture. Documents use the same idea, such as a pale "DRAFT" or "CONFIDENTIAL" behind the text. A semi-transparent watermark across the centre is harder to crop out than one in a corner, but it also detracts from the picture.
Why it matters
A watermark deters casual copying and keeps your name attached to an image that is shared without credit; it cannot prevent determined removal by cropping, cloning or AI inpainting. Always keep an unwatermarked original, because the mark becomes part of the pixels and cannot be lifted off the saved file. Too faint and it is easy to remove; too strong and it ruins the image.
Example
White text at 30% opacity over a mid-grey pixel of value 100 gives 0.3 × 255 + 0.7 × 100 = 146.5, so about 147; over a light pixel of 200 it gives 216.5. The mark is clearly visible on dark areas and fades on light ones, which is why many tools add a thin dark outline.
Common confusion
An invisible or digital watermark hides identifying data inside the image itself, a form of Steganography, rather than drawing a visible mark.