Before
Page 2 CONFIDENTIAL Left column text... Right column text... Total $1,240.00
Readable Markdown for documents, PDFs, spreadsheets, HTML, and knowledge-base workflows.
support@mdforall.com
Markdown converter
Convert PDF to Markdown
Markdown For All converts text-based PDFs into reviewable Markdown drafts for documentation, search, and AI knowledge-base cleanup. It can retain useful text structure when the PDF has a usable text layer, but reading order, tables, custom fonts, scanned pages, names, dates, totals, and legal or financial details require manual review. Scanned PDFs should go through OCR before Markdown cleanup.
Should you convert the PDF, OCR it, or find the original file?
Choose the source and workflow before uploading. Markdown is strongest when it represents durable reading structure rather than exact visual layout or executable application logic.
| Source situation | Recommended decision |
|---|---|
| Text-based report or manual | Convert directly, then compare reading order and section boundaries. |
| Exported from Word or Google Docs | Use the original editable file when available; it usually preserves structure better. |
| Scanned or image-only PDF | Run OCR first. Markdown cleanup cannot recover characters that were never recognized. |
| Legal, financial, medical, or customer PDF | Use an approved private workflow and verify every material detail against the source. |
What PDF conversion can preserve—and what remains uncertain
| Source feature | What may carry into Markdown | What to review |
|---|---|---|
| Selectable text | Paragraphs and some inferred headings | Unicode mapping, punctuation, non-Latin characters |
| Lists | Simple bullets and numbering | Nested indentation and visual grouping |
| Tables | Regular grids may become Markdown tables | Merged cells, form layouts, totals, column order |
| Page structure | Some section boundaries | Columns, sidebars, headers, footers, page numbers |
Prepare a PDF before uploading
These visible preparation steps are also represented in the page HowTo schema. They reduce avoidable conversion errors but do not replace source comparison.
Step 1
Test text selection
Copy a paragraph into a plain editor. Empty or scrambled text indicates OCR or font-mapping problems.
Step 2
Inspect page layout
Identify columns, sidebars, forms, repeated headers, footnotes, and visual tables.
Step 3
Choose the workflow
Convert text-based pages, OCR scans, or retrieve the original DOCX when possible.
Step 4
Define verification points
List names, dates, totals, clauses, units, and page references that must match exactly.
Evidence example: a page fragment becomes a reviewable draft
After Markdown
## Section title Left column text continues in reading order. | Item | Amount | |---|---:| | Total | $1,240.00 |
What to review
Verify that the left and right columns were not interleaved, remove page furniture, and compare the amount and currency with the original page.
When another PDF workflow is safer
- • Use OCR before conversion when the PDF is scanned or selection returns no usable text.
- • Use the original Word, PowerPoint, or spreadsheet file when it still exists.
- • Use a private or self-hosted workflow for contracts, regulated records, bank statements, health information, or customer data.
- • Keep the PDF as the visual source of truth when exact positioning, signatures, annotations, or page references matter.
Technical field notes for PDF extraction
Text layers and ToUnicode maps
A PDF viewer can paint readable glyphs even when the file lacks a dependable Unicode map. Extraction tools rely on font encodings, ToUnicode CMaps, character positioning, and drawing operators. Missing mappings can turn visible words into blank text, substitutions, or unusual code points. Copy-and-paste testing is a practical signal, but it is not a guarantee that every page uses the same font mapping.
Coordinates are not document hierarchy
PDF stores text fragments at page coordinates. It does not inherently say that one line is a heading, that two fragments belong to the same paragraph, or that a right-hand sidebar follows the left column. Heading inference therefore depends on font size, spacing, repetition, and reading-order heuristics. A visually obvious section can still be ambiguous to a parser.
Ligatures, hyphenation, and line joins
Words may contain ligatures such as “fi”, soft hyphens, or line-end hyphenation. Extraction may split one word, merge two words, or keep a discretionary hyphen that should disappear. Search for suspicious character gaps, broken URLs, split identifiers, and words interrupted at page boundaries.
PDF tables are positioned text
A ruled table may contain no table object at all. Each number and label can be an independent text fragment. Column alignment can fail when cells wrap, values are right-aligned, or headers span columns. Recalculate or compare totals from the original instead of trusting a visually plausible Markdown grid.
Downloadable PDF regression sample
Test a selectable-text PDF before trusting a larger report
The public sample is a small text-layer PDF with a title, labeled facts, and a checklist. The linked Markdown is the output observed from the current release, not an idealized rewrite. It deliberately exposes line-grouping limits so users can see why PDF output must be compared with the page.
Observed result and what it proves
The converter extracts the selectable words, but the short labels and list lines are grouped into a compact draft. This proves that a text layer is only the starting point; the reviewer must restore intended paragraph and list boundaries and verify names, dates, and page order.
Failure case: image-only or multi-column PDF
If text selection returns nothing, OCR is required before Markdown cleanup. If selecting text jumps between columns or sidebars, conversion may extract the words but not the intended reading order; use the original DOCX when available or rebuild the affected sections manually.
Public samples are low-risk regression fixtures. The observed output was generated by release 2026-07-17-adsense-r9 and still requires comparison with the source.
Questions users ask before relying on this output
Can I convert scanned PDFs?
Not directly with reliable text extraction. Run OCR first, then convert and review the recognized text.
Why does PDF reading order break?
PDFs store positioned fragments rather than a semantic reading sequence. Columns, sidebars, and footnotes can be emitted in the wrong order.
Can PDF tables become exact Markdown tables?
Only regular grids convert cleanly. Visually drawn cells, merged regions, forms, and irregular headers usually need manual rebuilding.
Why are headers and page numbers in the output?
Repeated page furniture is ordinary text in many PDFs. Remove it during cleanup after confirming it is not meaningful content.
Is the original Word file better than the PDF?
Usually yes. DOCX heading styles, lists, links, and tables contain more semantic structure than the PDF export.
Should I upload contracts or financial statements?
Use an approved private workflow. If a high-risk file is converted, verify names, dates, amounts, clauses, and signatures against the source.
Continue with the right review workflow
These links are selected for this format rather than generated as a generic related-content list.
PDF conversion limits
Use this guide for columns, scans, custom fonts, forms, tables, and page artifacts.
Use the original Word file
When DOCX is available, converting it usually produces a cleaner hierarchy than converting its PDF export.
Format-by-format limits
Review the shared preservation limits and sensitive-file guidance before relying on output.
Version and review notes
Last reviewed July 17, 2026. This page documents the public converter behavior for release 2026-07-17-adsense-r9. Changes in format support, file handling, or known limits are recorded in the public changelog.