On this page
Name the page-order layer under review
The PDF reference carries its verified publication year. The preservation page uses the documentation check date, 22 October 2026, rather than an assumed publication date; both sources were checked that day. Start by naming the artifact, file version, page index convention, and the order evidence available. A zero-based object index, a displayed page number, a chapter label, and an extracted text sequence are not interchangeable. Record each layer without rewriting the source.
| Layer | Record | Do not infer |
|---|---|---|
| Physical index | Page object sequence and count | Index equals printed number |
| Page label | Displayed label and numbering style | Roman and Arabic labels are continuous |
| Visible markers | Title, footer, appendix, insert | Marker proves previous page exists |
| Logical structure | Tags, outline, bookmarks if present | Bookmark is complete reading order |
| Extraction order | Tool, version, and sequence | Text order equals visual order |
| Missing signal | Gap, duplicate, or unresolved edge | Blank page is accidental |
The Adobe, (opens in a new tab) reference is useful for separating page-tree structure from document sections. The Library of Congress, (opens in a new tab) discussion reinforces that logical structure and visible page content can be represented independently. Use both observations to keep a page-order report explicit about its evidence layer.
Reconcile labels, inserts, and blank pages
Create one row per physical page with object index, visible label, first heading or marker, orientation, blank status, and neighboring labels. Include cover sheets, foldouts, intentionally blank pages, scanned inserts, and appendices. A printed page label can restart or use letters without changing physical order. A missing number is a finding to investigate, not proof that a page was omitted.
| Finding | Evidence to capture | Disposition |
|---|---|---|
| Restarted numbering | Label style and boundary heading | Separate label sequence |
| Insert | Physical position and insert marker | Retain in page order |
| Blank page | Blankness check and label | Ask whether intentional |
| Duplicate label | Neighboring markers and index | Do not overwrite label |
| Missing label | Object index and visible context | Unresolved numbering |
| Appendix | Section marker and page range | Logical section boundary |
Do not repair a page sequence by renaming files. Keep the source index and any proposed logical order as separate columns, with the reason for each proposed movement.
Check logical structure without treating it as truth
If the PDF contains tags, bookmarks, or an outline, record their presence and the page targets they reference. Compare those targets with physical order and visible headings. A missing tag or bookmark is a metadata gap; it does not establish that the visible page is unreadable. Likewise, extracted text that interleaves columns may reflect a tool order rather than authorial order.
- Record physical page order and page count.
- Record labels and section markers without normalization.
- Record logical tags, outline targets, and bookmarks if present.
- Compare extraction order with visible headings on synthetic pages.
- Mark every disagreement as a layer-specific finding.
- Preserve the source artifact and tool version used for extraction.
Use a synthetic page sequence
Synthetic page-order fixture
object_index | label | marker | layer | finding
0 | cover | title | physical | starts sequence
1 | i | contents | printed | roman label
2 | ii | blank | printed | intentionality unknown
3 | 1 | chapter one | printed | numbering restart
4 | 2 | appendix insert | physical | insert marker
5 | 4 | chapter two | printed | label gap at 3
logical_outline: chapter one -> chapter two; appendix target unresolved
report: retain physical order; flag label and logical disagreements| Test | Expected result | Hold condition |
|---|---|---|
| Label restart | Two label sequences retained | Roman labels converted away |
| Blank page | Blank and intentionality separate | Blank dropped from count |
| Insert | Physical index retained | File renamed into sequence |
| Gap | Label gap flagged | Missing page asserted |
| Outline | Target checked separately | Outline treated as page truth |
The fixture is invented and contains no source document text. Add a rotated page, a duplicate footer, and a bookmark pointing to an appendix. The result should identify an evidence conflict without attempting OCR or semantic reconstruction.
Keep page-order evidence separate from document interpretation
Page-order review can document sequence and metadata disagreement. It cannot establish that a document is complete, authentic, readable, authoritative, or suitable for a decision. Preserve the original file, layer-specific observations, and unresolved gaps for a qualified document review.
Pair this guide with document OCR layout coordinate review for extraction geometry and the data provenance log for artifact history. After a supported non-health scope, authority to describe the collection, and a controlled document packet are established, the qualified appraisal request path may be a preparation step. It does not reorder pages, run OCR, create a quote, match a buyer, or grant a license.
Takeaways
- Keep physical index, printed label, logical structure, and extraction order separate.
- Record inserts, blanks, restarts, duplicate labels, and gaps explicitly.
- Treat proposed reordering as a derived decision with a reason.
- Use synthetic page fixtures before exposing source text.
- Do not infer completeness from one page-order layer.