Why an edit can leave a visible patch on a PDF
You replaced a line and there is now a pale rectangle sitting behind it. That patch is the old text being covered, because it could not be taken out.
Why anything has to be covered
Where the exporter can prove which instructions drew the block you edited, it removes them and there is nothing to cover. A patch means it could not prove it — why that proof is not always available.
Why the patch shows
Where the pixels immediately around the text are one colour, the cover is filled with that colour and is invisible — which is the ordinary case on white paper.
Where they are not one colour — a photograph, a gradient, a tinted table row, a watermark, a scanned background — there is no correct colour to use, and the cover is filled with white. On a dark or patterned page that is exactly as visible as it sounds. This is the case the overlay only grade is warning you about.
This is measured before you type. The pixels in a thin ring around each block are sampled, and if they vary the block is graded overlay only — a warning that an edit there will be visible rather than seamless. The same grade is used when the page could not be rasterised at all, because assuming a flat background would be assuming the good case.
What you can do
- Check the grade before typing. Blocks graded overlay only are the ones that will show.
- Keep the replacement close in length. A short correction inside a longer line needs less covering than a full rewrite.
- Where the layout allows, edit the whole line rather than part of it, so the patch follows a natural boundary.
A patch is not a way to hide anything
A cover rectangle is drawn on top. It does not remove what is underneath, and the text under it may still be present in the file. Even where the original instructions are removed, that reaches one block and nothing else. This is not redaction, and must not be used as though it were.
Related: the five reasons a PDF’s text will not edit, how PDF text editing works.