PDF page counter

A page here is one page object in the PDF's page tree — exactly what a parser reports when it walks the document's own structure. That is the same number a compliant reader shows in its page control and the same number a print shop bills against, because it counts leaves in the file rather than anything about how the document might later be printed or displayed. It is not a count of sheets of paper: a duplex job puts two pages on one sheet, and an imposed booklet puts four. It is not a count of visible content either — a page object that draws nothing still counts, which is right, because blank separator pages are real pages inside the file. Each row's average kilobytes per page is derived from this same count.

A reader can disagree with this tool for three ordinary reasons. The first is page labels: a PDF can carry its own numbering scheme, so a document whose front matter runs i to iv and whose body restarts at 1 displays a final page number lower than its real page count. The second is display mode — a viewer showing two-up spreads reports spreads on screen while the file still holds individual pages. The third is repair: when a document's cross-reference table is damaged, different parsers recover different subsets of the page tree, and a reader that reconstructs more objects than this tool's parser can will report more pages. When the count matters commercially, trust the file's own structure over a viewer's status bar.

Yes — the drop zone takes multiple files, and dropping more at any point appends them to the table rather than replacing what is already there. Every file becomes a row with its page count, size in kilobytes and average kilobytes per page, while the footer keeps running totals: number of files, total pages, total size and the overall average. The × button on a row removes it and the totals recalculate immediately, which helps when you have dragged in one file too many. There is no fixed limit on how many documents you can queue, though several hundred large ones will make the tab work for it, since each is parsed in memory. This is the fast way to answer how long a discovery bundle runs, or how many pages a print run adds up to.

Average kilobytes per page is a density signal — file size divided by page count, calculated per file and again across the whole batch. It tells you at a glance which documents are heavy for their length. A text-only report sits low; a page of scanned images sits far higher, because every page carries a full-resolution bitmap. When one file in a batch shows an average several times its neighbours', that is the file worth compressing, downsampling or re-exporting from its source. It also works as a sanity check after a merge: if the average jumps, something in the input was image-heavy. The figure is deliberately coarse — it says nothing about which particular pages are large, only about the document as a whole.

Files that cannot be parsed stay in the table as a row with a red error message instead of being silently dropped. Encrypted documents are opened with the encryption flag ignored, so most password-protected PDFs still yield a correct page count; only files whose structure is genuinely unreadable fail. A failed row keeps its real file size and contributes to the total size, but contributes nothing to the page total or to the averages — counting it as zero pages would quietly understate every figure in the footer. The footer also shows, in red, how many files were unreadable next to the file count, and the copied Markdown marks each of those rows with ERROR and the parser's own message, so a report you paste elsewhere never looks complete when it is not.

Counting pages here is free, unlimited and entirely local. Each PDF is read from disk by the browser and parsed inside the tab; nothing is sent to a server, there is no account, and there is no cap on how many files or how many pages you may run through it. Copy as Markdown uses your system clipboard, so even the report never crosses the network. That matters for the batches people usually count — case bundles, tender submissions, patient records, archive scans — where uploading a folder of documents to an unknown backend just to learn how long they are would be an absurd trade. And because the tool produces no output file, there is no download that could carry a watermark or a credit line.