← All posts

The tool deleted the wrong page. Here’s why.

20 August 2026

You needed page 12 gone. You typed 12, ran the tool, and downloaded a file that is missing something else entirely — a page from the middle of the contents, or a blank you had not noticed, or the wrong half of a table.

The tool did not misfire. It deleted the twelfth page of the file, exactly as asked. The trouble is that the number you read off the corner of the paper and the number the tool counts are two different numbers, and in most real documents they are several pages apart.

There are actually three of them, and they only agree in a document nobody has touched.

Three numbers in one file

The position in the file. A PDF stores its pages as an ordered list. Page one is the first entry, page two is the second, and there is nothing else to it — no gaps, no names, no skipping. This is the number every page tool on this site uses, and every page tool anywhere else too. It is counted from 1 and it always runs 1, 2, 3 to the end.

The page label. A PDF may also carry a separate, invisible table saying what a reader should display for each sheet. It is a list of rules: "from sheet 1, lower-case roman starting at i", then "from sheet 13, arabic starting at 1". That is how a reader can show you "iv" or "A-3" or "1-1" in its page box. The label is not ink and it is not on the page. It is a lookup table in the file's catalogue, and most PDFs do not have one at all.

The ink. The number actually printed on the paper, put there by Word or InDesign or whatever produced the document, or by a stamping tool afterwards. To the PDF this is not a number. It is a few glyphs at some coordinates, no different from any other word on the page. Nothing reads it, nothing checks it, and nothing keeps it in step with anything else.

So when a tool asks for "pages to delete", it can only mean the first of the three. It is the only one that is guaranteed to exist.

Why they drift apart

The gap between the printed number and the position is almost always front matter. A report with a title page, a blank verso, a contents spread and a foreword numbered i to viii has eight sheets in the file before the sheet that prints "1". Printed page 12 is the twentieth entry in the list. Type 12 into a delete tool and you remove something in the contents.

Everything else that opens a gap works the same way:

• A cover added afterwards. One sheet, no number on it, and every printed number is now one behind its position.
• Exhibits dropped into a bundle. Each exhibit keeps whatever numbering it was born with, so the printed numbers restart, repeat, or jump. A 300-page bundle can contain four separate page 1s.
• A merge. Same problem, deliberately. Every source file brings its own printed numbers into the combined document — which is also what happens to the bookmarks.
• A scan of a stapled original. A missing sheet, a double feed, or the blank backs a duplex feeder photographs will all shift the count. That last one has its own fix, and it is worth doing before you start counting anything.

Once there is a gap, it does not have to stay the same size. A document with roman front matter has one offset for the preface and a different one for the body. Working out "the number is 8 out" and applying it everywhere is the second most common way to delete the wrong page.

Which number is your reader showing you?

This is the part that makes the mistake feel unfair, because your reader may well have been showing you the label all along.

Firefox's built-in viewer honours page labels: if the file has them, its page box says "iv" while you are on the fourth sheet. Acrobat does the same when Edit → Preferences → Page Display → Use logical page numbers is ticked, and it is helpful enough to show both — "iv (4 of 340)". Chrome's viewer ignores labels entirely and always counts sheets.

Which gives you a thirty-second test. Open the file in Chrome and in Firefox, go to the same sheet, and compare the page boxes. If they disagree, the document has page labels and the number you have been reading is not a position. If they agree, any remaining mismatch is the ink, and the ink is whatever the author felt like printing.

Two more ways to pin down a position:

• The thumbnail panel counts sheets. In every reader, the number under or beside a thumbnail is the position in the file, regardless of what the page box says.
• Bookmark every page. Run the file through Create PDF bookmarks with every page selected. It uses the document's own page labels where it finds them, so the bookmark list comes back reading "Page i, Page ii, … Page 1, Page 2" in file order. That list is the mapping between the two numberings, and you can read it straight off.

Our PDF inspector will give you the page count without uploading anything, which is the other number worth knowing before you start — but it reports page count, page sizes, version, metadata and text layer, not page labels. It will not settle this one for you.

The habit that stops it happening

Do not delete first. Look first.

1. Find the position of the page you want gone, using the thumbnail count or the Chrome/Firefox test above.
2. Extract it and open it. Put the file through Extract PDF pages with just that number — 20 — and open the one-page PDF that comes back. This takes fifteen seconds and it is the whole safety net: the original is untouched, and you are looking at exactly the page the delete tool would have removed.
3. Now delete it. With the position confirmed, Delete PDF pages takes the same syntax: 20, or 20-24 for a run, or 3,7,9 for scattered pages.
4. Read the count. The result says how many pages went and how many are left. Original page count minus pages removed should equal the number you are looking at. If it does not, your range meant something other than what you thought.

If you would rather see the pages while you decide, Organize PDF pages approaches it from the other end: you write the pages to keep, in order, so 1-19,21-33,35- says the same thing as deleting 20 and 34 with no arithmetic about what shifted. Both tools rebuild the document, so bookmarks do not survive and a form's field definitions are left behind.

What none of these tools will accept

Page labels, anywhere. No tool on this site takes iv or A-3 as a page number. The parser reads digits, dashes for runs, commas for lists, an open end like 12-, and the words odd and even. Everything else is refused. If your document is labelled, you convert to positions yourself — there is no option to switch it into label mode, on any tool here.

odd and even mean the position, not the printed number. In a book with an unnumbered cover, the printed odd pages sit on even positions all the way through. Deleting even from that file removes exactly the pages you wanted to keep.

A refusal is on your side here, incidentally. Type a page number the document does not have and the job stops with the real range in the message rather than doing something approximate. Typing a printed number that is larger than the file is the one version of this mistake that catches itself.

When a desktop tool is genuinely better

For reading the labels, and for the case where you are doing this to forty files:

• pdftk file.pdf dump_data | grep -A4 PageLabel prints the label rules — where each numbering section starts, what style it uses, what number it starts at. This is the fastest honest way to see the mapping, and it is the fact none of the forum threads about this ever quote.
• qpdf --empty --pages in.pdf 1-11,13-z -- out.pdf deletes page 12 by position, where z is the last page. Free, scriptable, and it keeps the document structure rather than rebuilding it.
• pdftk in.pdf cat 1-11 13-end output out.pdf does the same job if you already have pdftk.
• Acrobat's Page Thumbnails → Page Labels, or the free PDF-XChange Editor on Windows, are what to use if you want to set labels so the reader's page box finally agrees with the paper.

Leaving it better than you found it

If the document is going to somebody else, print the numbers on it. Add page numbers to PDF stamps a figure into each page's content — real ink, so it prints, it survives a merge, and it comes out with the text. Set the range to 9- and the start to 1 and the ninth sheet reads "1", which is the arrangement most reports want; the mechanics of that particular pair of settings are worked through here.

Two things it cannot do, and both matter on exactly this document. It offers 1, Page 1, 1 of N and Page 1 of N — no roman numerals, so front matter numbered i to viii is not something it can produce. And it stamps ink only: it does not set the PDF's page labels, so the reader's page box will keep counting 1, 2, 3 whatever the paper now says. For that half of the job you need Acrobat or PDF-XChange above.

Stamped numbers that match the positions are still the single most useful thing you can do to a shared document. The next person types 12, gets page 12, and never finds out why that was ever in doubt.