A PDF table may depend on merged cells, grouped headings, indentation, and notes outside its borders. A plain Markdown table cannot necessarily express all of those relationships. Even when the words and numbers are recognized correctly, the representation may need editing.
Separate two questions: did conversion recover the source accurately, and can the chosen Markdown form communicate it clearly? The first is an extraction check; the second is an editing decision.
Start with the simplest faithful representation
A table with one header row and consistent columns is often a good fit for pipe-table syntax supported by many Markdown editors. Keep each value with its row label and column heading.
For example, this manually written illustration has an unambiguous structure:
| Material | Mass (g) | Condition |
| --- | ---: | --- |
| Sample A | 12.4 | Dry |
| Sample B | 15.1 | Wet |
It is not a sample conversion result. It shows what a simple representation can communicate: the label, quantity, unit, and condition are visible together.
GitHub Flavored Markdown defines a table extension, while basic Markdown implementations may differ. Preview the result in the editor or publishing system you intend to use rather than assuming every renderer behaves the same way.
Flatten merged headings without losing their meaning
Pipe tables do not provide a portable way to reproduce arbitrary merged cells. A two-level PDF header may need explicit combined labels such as “2025 — Domestic” and “2025 — Export” instead of a merged year heading above several columns.
For grouped rows, repeat or restate the group label where necessary. Do not treat a visually blank source cell as a zero or missing value without checking its meaning. A label spanning several rows may apply to every row beneath it.
If flattening makes the result unwieldy, consider several smaller tables, a structured list, or retaining the original table image with a textual explanation. Choose a form that preserves the relationships instead of imitating borders at the expense of clarity.
An HTML table is another possibility in editors that support it, but support and Markdown-inside-HTML behavior vary. Do not introduce that format without testing the actual destination.
Handle page continuations as one logical table
A PDF can repeat its header on every page. In the Markdown, distinguish those repeated labels from real data rows. Check the last row before the break and the first row after it, especially where a description wraps across pages.
Keep notes and captions with the table they qualify. A note printed after the final page may apply to values much earlier. If you split the table for readability, repeat the necessary title, labels, units, and note references rather than creating fragments that cannot stand alone.
Preserve original page locations in your review notes even if you remove page breaks from the working Markdown. They are useful when a missing line or duplicated row is discovered later.
Verify headings and values after restructuring
For each sampled number, trace the full path back to the source: table title, group, column, row, and unit. Then reverse the check from the Markdown to the PDF to catch duplicates and invented entries.
Inspect decimals, minus signs, ranges, percentages, blank cells, and note markers. Do not choose a correction solely because it makes a total look reasonable. The source may contain rounding or its own error, which should not be silently “fixed” as an OCR correction.
For recognition-specific faults, use the scanned table OCR checklist. Keep editing choices separate from any uncertain source values.
Convert, edit, and preview the final use
Upload your PDF to PDF to Markdown, choose Markdown in task settings, and download the completed result. PDF Craft recognizes tables during processing; this does not guarantee that every complex printed structure becomes a perfectly portable Markdown table.
Edit a copy and keep the original export. Preview the complete table at the width where readers will use it. Check that long labels, escaped pipe characters, and notes remain understandable.
If the table will enter research notes, preserve its source citation and original PDF alongside the research Markdown workflow. A useful table is one whose meaning survives both extraction and the format you chose to present it.