Copy-pasting a table out of a PDF gives you soup. Real extraction reconstructs rows and columns from the page's text geometry.
Digital PDFs store every text fragment with coordinates. Kameleo clusters fragments that share a baseline into rows, finds persistent horizontal gaps to derive columns, and rebuilds the grid — all in your browser. The result exports as CSV that Excel, Numbers and Google Sheets open natively.
A scan has no text coordinates — it's a photo. Run Kameleo's OCR tool first to give the document a text layer, then extract. Accuracy depends on scan quality; 300 DPI originals work best.
Merged header cells and multi-line cells are the usual suspects for misalignment. Extract page by page rather than all-at-once when a document mixes layouts, and give exported columns a quick sanity check against the original before analysis.