Plainmark guide
How to preserve reading order when converting multi-column PDFs to Markdown.
Use the source PDF to verify column boundaries, full-width elements, captions, sidebars, and footnotes before exporting Markdown.
The short answer
To preserve reading order, identify full-width elements first, then read each column from top to bottom in the document’s intended direction. Verify headings, captions, sidebars, footnotes, and tables against the source before export.
Why do multi-column PDFs break reading order?
A PDF can contain positioned text fragments without recording a semantic article flow. A converter may therefore follow object order or page coordinates instead of the sequence a human reader understands.
- Left and right columns may alternate line by line.
- A caption can appear before the paragraph that introduces its figure.
- A sidebar or pull quote may interrupt the main argument.
- Footnotes may be inserted midway through the body.
- A full-width heading may be attached to only one column.
Document-conversion systems such as Docling expose layout-aware processing because reading order is a separate interpretation problem, not merely text extraction.
How should you reconstruct the reading order?
- Find full-width elements. Mark titles, abstracts, section headings, figures, and tables that span both columns.
- Mark the column boundaries. Confirm where each column begins and ends, including pages that change layout halfway down.
- Read one column at a time. For left-to-right documents, finish the left column before moving to the right unless a spanning element clearly interrupts the flow.
- Place attached content deliberately. Keep captions near their figures and tables; move sidebars after the paragraph or section they support.
- Compare before export. Read the Markdown from top to bottom while checking each transition against the original page.
Which layout patterns need the closest review?
| Layout pattern | Intended reading order | Common failure | Review action |
|---|---|---|---|
| Full-width heading above two columns | Heading, left column, right column | Heading attached to one column | Keep the heading before both columns |
| Two equal text columns | Top to bottom in the first column, then the second | Lines alternate across columns | Regroup paragraphs by column |
| Sidebar or pull quote | After the related paragraph or as a separate note | Inserted inside a sentence | Move it to a clear boundary |
| Figure spanning both columns | Near the paragraph that introduces it | Caption detached from image | Keep image and caption together |
| Footnotes | After the section or document body | Mixed into body paragraphs | Collect and label them consistently |
| Wide table | After its introduction, before interpretation | Cells merge with nearby columns | Compare every row with the source |
How do you fix interleaved Markdown?
Look for a sentence or heading that appears to jump to another topic and then returns. That usually marks a column boundary error.
Interleaved result
## Left-column heading First left-column paragraph. ## Right-column heading Second left-column paragraph.
Corrected sequence
## Left-column heading First left-column paragraph. Second left-column paragraph. ## Right-column heading First right-column paragraph.
After regrouping the content, check heading levels, lists, links, tables, and blank lines. If you are preparing the result for retrieval, follow the additional provenance and structure checks in the RAG-ready Markdown guide.
What should you check before exporting?
- Every heading precedes the section it describes.
- No paragraph alternates between two columns.
- Captions remain next to the correct figure or table.
- Sidebars and pull quotes do not split sentences.
- Footnotes appear in a predictable location.
- Tables retain the correct row and column relationships.
- The Markdown reads coherently without the visual page.
For table-specific failures, use the PDF table extraction checklist. For broader structural checks, start with the core PDF-to-Markdown workflow.
Questions about multi-column PDF conversion
Why is reading order difficult in multi-column PDFs?
A PDF may store positioned text fragments instead of a semantic article flow, so extraction order can differ from the order a reader sees on the page.
Can Markdown preserve a two-column layout?
Standard Markdown is linear. A good conversion preserves the intended sequence and hierarchy rather than reproducing the visual columns.
Does Plainmark use OCR for multi-column PDFs?
No. Plainmark currently works best with text-based PDFs that already contain selectable text.
Where should captions and sidebars go in Markdown?
Keep a caption next to the figure or table it explains. Place a sidebar after the paragraph it supports unless it is clearly a separate section.
When should I use a specialist document parser?
Use one when a document contains dense scholarly layouts, formulas, irregular columns, or scanned pages that require OCR.
Review a multi-column PDF beside its Markdown.
Convert a text-based PDF, compare the source and result, then correct the reading order before export.
Convert a PDF