Load a PDF with at least four pages, keep Extract selected pages active and enter 2-4. The preview must report three copied pages and one output. After splitting, the download is a PDF rather than ZIP. Its first output page corresponds to source page 2, followed by source pages 3 and 4. The original page 1 and every page after 4 are absent. If the source has fewer than four pages, PDFSplit rejects the plan instead of clipping it silently.
A split result is a new document
PDFSplit copies selected visible pages into newly serialized PDF files; it does not cut bytes out of the original container. Document-level metadata, outlines, page labels, embedded files, portfolios, scripts, open actions, optional layers, tagged structure, PDF/A or PDF/X claims, usage rights, encryption, incremental history and digital signatures may not carry across. Forms, links and annotations can depend on shared resources. Keep the source and never describe an output as signed, certified, archival, compliant or evidentiary merely because the pages look correct.
Inspect every downloaded output
Open the exact downloaded PDF files in the reader used by their destination. Compare the first and last page of every requested range, total page counts, visible order, rotation and crop boxes. Test hyperlinks, annotations, form controls, fonts, reading order and accessibility where they matter. A successful local operation proves only that the current browser produced files with the reported page counts and byte sizes. It cannot prove semantic completeness, destination acceptance, identity, signature validity, legal filing, regulatory compliance or archival quality.