Understanding PDF page numbering
A line of text, added on every page.
What "add page numbers" really does, the conventions that look professional, and the surprise around removing them later.
It's a footer overlay.
Adding page numbers means drawing text on every page at a specified position (usually bottom-centre or bottom-right). The position and font are chosen once; the text varies per page — "Page 1 of 50", "Page 2 of 50", and so on. Mechanically it's identical to a watermark; the only difference is the variable content.
Conventions that look right.
The widely-followed shape: 9-11 point serif font, bottom centre, in mid-grey (#555 or so) rather than pure black. "Page X of Y" reads better than just "X" because it tells the reader how much is left. Skip the first page if it's a cover; restart numbering after the table of contents in long documents. These are book-publishing conventions, learned over centuries, that still apply to PDFs.
Roman numerals on the front matter.
For documents long enough to have front matter (title page, dedication, table of contents) and a main body, the convention is to number front matter with lowercase Roman numerals (i, ii, iii) and start Arabic numbering (1, 2, 3) at the first chapter or first content page. The page count restarts; the page numbers convey that "we are in the introduction" vs "we are in the actual book". Useful for long reports too.
A worked example.
A 100-page document. The processor opens each page, draws "Page n of 100" at the bottom centre in 10pt grey serif. Front cover gets nothing (skip page 1). Pages 2-7 are front matter, get "ii" through "vii" in lowercase Roman. Pages 8-100 get "1" through "93" in Arabic. The variable here isn't the position or font but the text itself, computed per page.
Number scheme
skip 1, roman 2–7, arabic 8+
Different ranges, different numbering style.
page 8 of 100 → render 'Page 1' at bottom centre
= Consistent reading aid
Headers and bates numbering.
The legal industry uses Bates numbering — a globally unique identifier on every page across a deposition or discovery production, often six-digit zero-padded (000001, 000002, ...) with a case-specific prefix. Same overlay operation, different number scheme. Modern e-discovery software does this in bulk across thousands of documents; for a single document, any PDF page-number tool covers the case.
Removing them later.
A page number added by overlay is not part of the original page content stream — it's an appended drawing instruction. Reversing the operation requires removing those instructions. Some tools track the watermark / page-number additions and let you reverse them; ones that don't, leave you needing to manually edit the content stream or convert PDF → Word → PDF round-trip to recover unnumbered versions.