Numbering PDF Pages - And Which Ones Not To Count

Adding a page number sounds like the simplest possible PDF edit - until the document has a cover page, or a photo inserted mid-way through that isn't really "page 14" so much as an appendix that happens to sit there. Numbering every page identically usually gets this wrong. Which pages get a number, and what number they show, is really two separate decisions.

Why a cover page usually goes unnumbered

A cover or title page is identified by the absence of a number - the reader's eye treats the first visible folio as where the document actually starts. Numbering the cover "1" doesn't break anything mechanically, but it reads as a mistake to anyone familiar with the convention, the same way starting a book's page count on its title page would.

Roman numerals for front matter, Arabic for the body

Longer documents often split the difference further: a table of contents or preface gets small Roman numerals (i, ii, iii...), and numbering resets to Arabic (1, 2, 3...) at the first page of actual content. It's a signal a reader can register without reading a word - front matter looks different from the body, on sight.

Skipping a page in the middle, not just the start

The harder case is a page that isn't at the beginning at all - a full-page photo or diagram inserted into an otherwise text-numbered report. Numbering it like any other page makes the count technically correct but visually odd; leaving it out entirely means whatever numbered the rest of the document needs to treat the sequence as having a gap, not just a starting offset. That's the difference between a tool that only supports "start numbering from page N" and one that supports arbitrary page ranges - HerePDF's page numbers tool does the latter: define more than one range, and the pages outside all of them are simply skipped, with numbering continuing correctly on the other side of the gap.

What "Page X of Y" should actually count

When a document shows "Page 3 of 12," most readers assume Y is however many pages they're paging through - not the PDF's raw physical page count. If three pages are deliberately excluded from numbering, the total should reflect that: the last numbered page should read "of 9," not "of 12." Getting this wrong is a subtle enough mismatch that a reader will notice something is off without immediately being able to say what.

What we measured: where a number really lands

"Bottom centre" sounds like the same place on every page, but a PDF page can be rotated, cropped or have its coordinates shifted, and a number placed against the stored page size ends up somewhere else. We stamped every page of 387 real documents (5,387 pages from a fixed collection of 2009-era US government PDFs; of the files with 3 to 40 pages and at most 3 MB, the 400 with the lowest SHA-256 of their thread/name) with HerePDF's page-number code and with the plain pdf-lib way, which lays the number out against the unrotated page size. Then we read both outputs back and checked that each number was level, inside the page, and within 2 points of where it was asked to go.

One page in ten is not the page you think it is. 512 of the 5,387 pages (9.5%), in 45 of the 387 files, were rotated (397 pages), cropped (262), shifted (8) or had no page size written at all (3), and some were more than one. Placed at the bottom centre, the plain method put 4,890 of 5,387 numbers where they were asked to go, and only 15 of the 512 on those pages. Every miss was on one of them, and 43 files had at least one. HerePDF lays the number out in the page as you see it, and placed all 5,387 correctly, and again all of them at the top right.

Which corner is safest depends on the document. We also checked, on the 5,186 pages that had a text layer, how often a two-digit number at each position would sit on top of text already there:

The default, bottom centre, hit text on 8.6% of pages. Top centre was the most crowded position, and top right the least. The data does not say why, and your document is not this sample, so use the preview before you apply.

Half of these pages already had a number. 2,556 of the 5,186 text pages, in 242 of the 387 files, had something that looks like a printed page number in the top or bottom tenth of the page. Adding a second gives a page with two numbers, so numbering is for a document with none, or for one combined from files whose own numbers stopped meaning anything.

The number a viewer shows can be a different number. A PDF can carry page labels, so that a viewer's page box reads "iii" or "A-4" for the third page. 155 of the 387 files had labels, and in 10 they differed from the plain position (roman numerals, a prefix, or a start other than 1). Page numbers are drawn on the page; they do not change the labels, which were kept in 155 of 155, so in such a file the viewer's box and the printed number can disagree.

The limits: these are old government documents, not yours. 13 of the 400 files are not counted: 10 were encrypted, and 2 have a page whose content is damaged, so the page-number tool fails on them too. The text-overlap figures read text boxes only, so a number could still land on a picture or a drawn line, and the 201 pages with no text layer (scans) are left out of them. The file names are not published, only counts; the data and the method are there to check against.

After merging documents

Numbering is often the last step after combining several documents into one - a set of chapters, or several scanned sections - since the individual files' own page numbers (if any) stop meaning anything once they're combined. HerePDF's merge tool combines the files; numbering the result is naturally the next step.