A PDF is not a folder of independent page images. Pages can refer to shared fonts, forms, annotations, outlines, labels and document-level objects. PDFRemove loads the container, chooses the zero-based source indices corresponding to every retained 1-based page number, copies those page objects to a blank document and saves the result. The source File object remains unchanged in browser memory until clear or reload.
Retained order is stable
The complement is built from source page 1 through source page N and filters out the selected set. Therefore page deletion never reorders retained pages. If pages 2 and 5-7 leave 1, 3, 4, 8, 9 and 10, those six become output pages 1 through 6 in exactly that order.
Output size is evidence, not a promise
Removing pages can reduce bytes, but reserialization and shared resources mean the change is not proportional to page count. The exact output byte number identifies the produced Blob for this run. It does not prove visual fidelity, compression quality, safe redaction or absence of recoverable information in every PDF feature.
The result is a new PDF
PDFRemove does not edit the source container in place. It creates a blank PDF, copies the retained page objects in their original order and serializes a separate download. Document metadata, outlines, page labels, embedded files, portfolios, scripts, optional layers, tagged structure, PDF/A or PDF/X claims, usage rights, encryption, incremental revisions and digital signatures may not carry across. Forms, links, annotations, fonts and accessibility relationships can rely on document-level resources. Keep the source and never call the output signed, certified, archival, compliant or evidentiary merely because its visible pages look correct.