Why “page 1” of a PDF isn’t always the first page
The page a tool calls “page 3” is the third sheet in the file, not the one printed 3. There are three different page numbers, and they routinely disagree.
When you tell a PDF tool "page 3", it does not look at what is printed on the paper. It counts to the third page in the order the file stores them, and on a real document that is often not the page with a "3" on it. A book has a cover, a title page, a copyright notice and a preface before its printed page 1, so the sheet stamped "1" might be the fifth sheet in the file. There are, quietly, three different page numbers in play, and most confusion with page tools comes from assuming they are the same number.
The three numbers
The first is the physical page index: the position of a page in the file's stored order, counting from the start. This is the number every tool uses. Internally it counts from zero, which is why the first page is index 0 to the code, but the tools present it to you as page 1 — either way, it is position in the file, and it takes no notice of what the page says.
The second is the printed page number: ink drawn onto the page, part of the content like any other text. It follows whatever convention the document's author chose — roman numerals for a preface, arabic starting at 1 on the first chapter, nothing at all on the cover. It is not metadata and no tool "knows" it; it is just marks that happen to look like a number, exactly as readable, and as meaningless to a machine, as the word "Chapter".
The third is the PDF page label: an optional piece of metadata that tells a viewer what to show in its own page box and thumbnails. It is what lets a good PDF of a book display "iv" while you read the preface and "1" when you reach the first chapter, independent of the physical order. Not every PDF sets labels, and when they are absent the viewer just shows the physical position instead — which is why some files show "1, 2, 3" in the box and others show "i, ii, iii, 1, 2".
Why "pages 3 to 7" surprises people
Because tools count physical position, "extract pages 3 to 7" means the third through seventh sheets in the file, not the pages a reader would call 3 to 7. On a report with three pages of front matter, those are two different sets, and the tool is not wrong — it is answering the only question it can answer, since the printed numbers are just ink it cannot read. Extracting pages and splitting both work in physical positions, and the fix is always the same: scroll to the page you mean, note which sheet it actually is in the file, and use that.
What adding page numbers actually does
Stamping numbers onto a document works on the second of the three — it draws real, selectable text onto the page as content. Our add-page-numbers tool puts genuine Helvetica text on every page, which is why the numbers stay searchable and print sharply rather than being a rendered picture of numbers; the how-to covers the positions, formats and the "start at" field in full. The one thing to hold onto here is that it stamps ink, not labels: it writes visible content, and it does not rewrite the document's internal page-label metadata.
That has a consequence worth seeing coming. If a PDF already carries page labels — a viewer showing "iv" in its box — and you stamp a "4" onto that same sheet, the two now disagree: the box says "iv", the paper says "4", and both are right about different systems. It is rarely a problem in practice, because most documents that need numbering never had labels to begin with, but it explains the occasional file where the number you added and the number the viewer shows are not the same.
The tool stamps every page, cover included, because it counts physical pages and a cover is a physical page. There is no notion of "the printed 1" for it to start from — only "the first sheet". If you want the cover left unnumbered, the honest route is to split it off, number the rest, and merge it back; the how-to walks through it. This is the same physical-versus-printed gap, showing up as a workflow instead of a surprise.
The through-line is simple once you have it: a machine sees a PDF as an ordered pile of sheets, numbered by position. The numbers humans care about — printed on the page, or shown in the viewer's box — are things laid on top of that pile, by an author or by metadata, and either can be missing or can disagree with the others. When a page tool does something you did not expect, the first question to ask is which of the three numbers you were thinking of, and which one the tool was.
Almost everything counts by position, not by print
It is not only extract and split. A viewer's thumbnail strip, its "go to page" box when a file has set no labels, and the print dialog's page range all count physical positions too. Ask a printer for "pages 2 to 4" and it prints the second through fourth sheets — cover included — not the pages a reader would number 2 to 4. Once you notice that the physical index is the number the software actually uses nearly everywhere, the odd behaviours stop being odd: they are all the same fact, that a machine counts sheets while a person reads ink.
Add Page NumbersStamp page numbers anywhere on the document.- The physical page index is position in the file; it is the number every tool counts by.
- The printed number is ink on the page — unreadable to tools, and often offset by front matter.
- The PDF page label is optional metadata the viewer shows in its box; many files do not set it.
- Adding page numbers stamps real content text on every page; it does not touch page labels, so the two can disagree.
