Cover the region, then check its edges
A mask only replaces pixels inside its rectangle.
Start with a complete visual pass
Open the fictional example and cover the customer details with one rectangle, then inspect the smaller attachment separately. Real screenshots can repeat information in a title bar, sidebar, notification, thumbnail or background. The app does not detect text, faces or sensitive content for you.
Use coordinates for deliberate boundaries
The X and Y fields start at the image's top-left corner. Width and height are image pixels, independent of the display zoom. In the 1000×600 example, X=50, Y=200, width=520, height=170 covers the three sample detail lines. Inspect the edges at a useful zoom and enlarge the box if any letters remain visible.
Dragging rounds outward to whole pixel boundaries and clips to the image. Coordinate entries instead require a whole-pixel rectangle fully inside the image. Select a mask in the list to update its bounds. Arrow keys move the selected mask by one pixel while the image control is focused; Shift plus an arrow moves it by ten pixels.
Coverage is geometry, not a privacy score
The masked-pixel percentage counts the union of rectangles, so overlaps are not double-counted. It does not measure how much sensitive information has been found or whether an image is safe to share. One uncovered name can matter even when most of an image is black.