PDF to Markdown
Turn a PDF into clean, editable Markdown — without uploading it.
A PDF does not contain a table in the way a spreadsheet does. It contains instructions for drawing text and lines at particular positions. The visual grid is real to you; to a parser, it is a collection of coordinates.
That distinction explains a common disappointment: a PDF-to-text tool returns all the right words in the wrong order. The content was extracted; the table structure was not.
What “extracting a table” actually means
There are three different jobs people describe with the same sentence:
- Get the words out. Useful for search or a quick summary, but rows and columns may collapse.
- Rebuild the structure. Produce Markdown or HTML with rows and columns.
- Get a clean dataset. Produce CSV or XLSX with one value per cell, ready for analysis.
A text-first converter can do the first well and the second partially. The third usually needs a table-aware tool or manual cleanup, especially when cells span multiple lines.
A local-first workflow
1. Check whether the PDF has a text layer
Open the PDF and try to select a value inside the table. If the selection highlights individual words, there is a text layer. If it selects the whole page as an image, the PDF is a scan; run OCR first.
2. Convert the PDF to Markdown
Use PDF to Markdown. It reads the PDF locally with PDF.js and rebuilds headings, paragraphs, and simple tables. The file is never uploaded.
Open the Markdown and inspect the table before trusting it. A pipe table should have the same number of cells in every row. If a row is short, a cell probably contained a line break or a merged column.
3. Fix the structure while it is small
Small tables are easier to repair in Markdown than in a spreadsheet. Check:
- header row matches the visible columns;
- every data row has the same number of cells;
- numbers did not lose a minus sign or decimal point;
- merged cells were not silently assigned to the wrong column;
- footnotes did not become a new row.
4. Move to CSV only when the table is clean
Markdown is good for reading and pasting into notes. CSV is better for sorting, formulas, and joining. Do not export to CSV until the rows are structurally correct; a malformed table becomes a much harder cleanup problem in a spreadsheet.
5. Verify totals independently
If the table has a total row, add the column yourself. A table can look structurally perfect and still contain a transcription error. For financial or medical data, compare the extracted values with the source PDF, not with the previous version of the spreadsheet.
Why tables break
- Merged cells. A header that spans three columns is one visual cell but three logical columns.
- Wrapped text. A long description may become two lines; a naive parser treats the second line as a new row.
- Blank cells. Whitespace is not a reliable separator.
- Multi-column pages. A two-column report can interleave unrelated rows.
- Borders drawn as images. Lines that are part of a scan are not structural metadata.
- Rotated pages. Landscape tables inside a portrait PDF often need rotation before extraction.
When to use a different approach
If you need a true dataset from dozens of similar tables, a dedicated table-extraction tool or a spreadsheet import is usually better than a general PDF converter. If you only need the argument of the document, use PDF Summarizer; it reads the extracted text and returns a summary, key points, and next steps without pretending to preserve every cell.
If the table is a scanned image, OCR can recover the words but not reliable columns. Rebuild the table manually for anything important.
A note on privacy
PDFs often contain payroll, pricing, contracts, or patient data. A browser-based converter keeps the file on your device; only the extracted text is used if you choose an AI summary. Avoid uploading the original PDF to a service that requires an account before it shows you the result.
The short version
A PDF table is a visual arrangement, not a data structure. Extract the text locally, check that every row has the same number of cells, verify the numbers against the source, and only then move the data into a spreadsheet. If the table is scanned or heavily merged, plan for manual cleanup.