How to prepare PDF for better recognition

When PDF to Google Sheets recognizes a table, the algorithm looks for structure in the document — so the quality of the PDF directly affects the result. In this guide, we’ll cover a few simple preparation steps that can noticeably improve recognition and take just a couple of minutes.

Two types of PDFs: text-based and scanned and how to tell the difference

PDF files usually fall into two categories:

• Text-based PDFs — regular documents. You can select text with the mouse, copy rows, and search inside the document (Ctrl+F works).
• Scans or photos — PDFs made from images. Text selection doesn’t work, and pages look like pictures.

Text-based PDFs are almost always recognized perfectly. Scans require a bit more attention — but the good news is that our app works with both types of PDFs.

What to check in a text-based PDF

To get better results, take a quick look at the following:

1. Tables aren’t split across pages in the middle of rows. If they are, it’s better to process those pages separately.
2. Column headers are visible on at least one page.
3. There are no watermarks covering the text.
4. Pages with landscape orientation are best processed separately from portrait pages.

What to check in scanned or photo-based PDFs

For scanned PDFs, image quality is the most important factor.

1. Contrast: best results come when text is clearly darker than the background. Pale gray scans significantly reduce OCR accuracy.
2. Resolution: higher is better. If the text is hard to read, it’s worth rescanning or retaking the photo with better lighting.
3. Shadows and glare: avoid shadows and reflections — they reduce recognition quality. Extra shadows may even be detected as table lines.

What most often reduces recognition quality

These recommendations apply to all PDFs:

• Avoid stamps and signatures placed over tables. If possible, use a page without overlaps.
• Instead of heavily compressed PDFs, use the original file whenever you can.
• Check merged cells: data will be extracted, but the structure may be easier to fix directly in Google Sheets.
• Normalize data after extraction if different languages, currencies, or units appear in the same column.
• Avoid stamps and signatures overlapping tables (worth repeating — they’re one of the most common issues).