What counts as an annotation in a PDF (and why some of them look like content)
Sticky notes, highlights, links and form fields all live in the same list inside a PDF. Knowing which is which explains why removing them wholesale breaks documents.
Open a PDF that has been through a review and you see two kinds of thing on the page: the document itself, and everything people added to it afterwards. To your eye they sit on the same sheet. To the file they are not remotely the same thing, and that difference explains a lot of otherwise baffling PDF behaviour.
files.co now has a tool built on that distinction: Remove annotations. Before you use it, it’s worth knowing what an annotation actually is, because the category is wider than most people expect.
A page and a list
Each page of a PDF has its content — text, lines, images — and, optionally, a list of annotations hanging off it. The list is separate. A highlight is not painted into the page; it is an entry in that list saying “draw a yellow band over these coordinates, in this colour, with this author and this timestamp”.
That is why you can move a comment around in a reader without disturbing the text underneath, why an annotation can carry an author name and a date when the page itself carries neither, and why the same file can look different in two viewers. One renders the list, another renders it only on demand, a third shows every note expanded.
The surprising members of the club
If the list only held sticky notes, life would be simple. It doesn’t. All of these are annotations, in the specification’s sense:
- Sticky notes and free-text boxes.
- Highlights, underlines, strikethroughs, squiggly underlines.
- Freehand ink, arrows, lines, rectangles, circles, polygons.
- Stamps — the “APPROVED” graphic somebody dropped on page one.
- Links. Every internal jump and every external URL.
- Form fields. Every text box, checkbox and dropdown in a fillable form.
- File attachments, embedded media, 3D objects.
The one that matters most is the form fields. In PDF terms they are annotations of the widget subtype, sitting in the same list as the yellow highlight two lines above them. The interactive form is a separate structure that points at those widgets; delete the widgets and the form is left referring to objects that no longer exist. You get a document that looks like a form and cannot be filled in. If you’ve ever seen fields go missing after some tool “cleaned” a file, that is usually what happened. We went through how fillable forms are put together in a separate piece on AcroForm and XFA.
Why the tool works in groups
Because of all that, “remove all annotations” is not an option worth offering. The tool splits the list into three groups you tick individually:
Comments and sticky notes. Anything with text somebody typed in.
Highlights and drawn marks. The visual layer of a review: colour bands, strikethroughs, ink, shapes, stamps.
Links. Unticked by default. Removing links changes nothing you can see and quietly disables the table of contents and every cross-reference in the document. Almost nobody who asks to “remove the annotations” means “and break the index”, so you have to ask for it explicitly.
Form fields are not in the list of groups at all. Neither are file attachments, embedded media or the more exotic subtypes: the tool leaves anything it doesn’t understand exactly where it found it, on the principle that removing something you didn’t ask about is worse than leaving it.
The pop-up nobody thinks about
One more piece of plumbing, and it’s the one that quietly corrupts files when it’s ignored. A sticky note is normally two objects: the marker on the page and the pop-up window holding its text. They point at each other.
Remove the note and leave the window and the file now contains a pop-up belonging to nothing, plus references aimed at a deleted object. Some viewers ignore it, some draw an empty box in the corner of the page. Removal has to walk the relationship in both directions — note takes window, orphaned window goes with its note — including the case where a generator never listed the pop-up on the page in the first place.
What removal costs
The text of a comment does not survive. In the copy you download it is gone, with no way to bring it back; your original file is untouched, because the tool builds a new document rather than editing yours. If the review history has any value to you, archive the marked-up version before you clean it.
There is also a narrow edge case worth naming. A tagged PDF carries an accessibility structure tree, and that tree can reference the annotations. Take the annotations out and a strict PDF/UA validator may flag the leftovers. Readers and screen readers cope with it; automated conformance checks may not. If your documents go through formal accessibility validation, verify the output before publishing.
Removing is not the same as flattening
Two operations get confused constantly. Removing takes the annotation out of the file — the highlight is gone, the page underneath is untouched. Flattening does the opposite: it presses the annotation into the page, so the highlight becomes part of the artwork and can’t be moved, edited or deleted any more. Cleaning up after a review calls for the first. Freezing a signed or filled document calls for the second. What flattening actually does covers that side in detail.
Both run in your browser on files.co, on your own machine, with nothing uploaded.