Key takeaways
- A scanned PDF has no text inside it, so it needs OCR; our PDF to Word tool offers Convert with OCR once standard conversion finds no text.
- The OCR path treats a line as a table row when it splits into three or more pieces separated by wide gaps, then builds real Word table cells.
- Two-column tables, empty cells, wrapped cell text and merged headers are the usual table problems, so check those areas first after converting.
- Straight 300 DPI scans with good contrast and upright pages improve results more than any fixing afterwards, and Word can turn tabbed text into tables.
To turn a scanned PDF into Word, you need OCR, because a scan is a picture of a page with no text inside it. Upload the file to PDF to Word and press Convert to Word. When the tool reports no extractable text, a Convert with OCR button appears. It reads the pages in your browser, writes the recognised English text into a .docx, and rebuilds rows it detects as a table into real Word table cells. Fonts, images and exact column widths do not survive, so plan a few minutes in Word to tidy the tables.
If you are not sure whether your PDF is a scan, our OCR guide for scanned PDFs explains the quick highlight test.
Normal conversion vs OCR
The same page runs two very different processes:
| Feature | Standard Convert to Word | Convert with OCR |
|---|---|---|
| Works on | PDFs with selectable text | Scans and photos saved as PDF |
| Where it runs | On our server, file discarded afterwards | In your browser, nothing uploaded |
| Size limit | 3 MB | No fixed cap, limited by your device |
| Tables | Arrive as plain lines of text, cells often run together | Detected rows become real Word tables |
| Language | Any text in the file | English only |
| Images | Not carried over | Not carried over |
The OCR path renders each page at about 216 DPI, turns it into clean black and white, and then reads it with the Tesseract engine. The Word file opens with the title "OCR Converted PDF Content" and the source file name and date, followed by the recognised text. Delete those first lines if you do not need them.
How the OCR spots a table
Knowing the rule helps you predict the result. After reading a page, the converter looks at each line of text. A line counts as a table row when it splits into three or more pieces separated by wide gaps (two or more spaces, or a tab). Runs of three or more underscores, often used as cell lines in forms, are treated as gaps too. Consecutive rows like that become one Word table, with as many columns as the widest row.
That rule explains the usual problems:
- Two-column tables come out as ordinary lines, because a row needs at least three pieces.
- Empty cells shift the values to their left. A blank cell leaves no text to detect, so the next value moves into its column.
- Cells with wrapped text become extra rows, since each printed line is read separately.
- Merged header cells do not merge. You get a header row with fewer pieces than the rows below.
- Narrow gaps between columns can merge two cells into one piece.
Simple grids of numbers, such as invoice line items or a timetable, usually come through best.
Prepare the scan before converting
Small fixes to the input make a bigger difference than anything you can do afterwards.
- Scan straight and at 300 DPI. Skewed pages throw rows out of line, and the OCR treats each page as one block of text.
- Turn sideways pages upright. A landscape table scanned in portrait reads badly. Rotate PDF turns every page by 90, 180 or 270 degrees, so it suits a file where all pages face the same way.
- Watch out for pale text and shaded rows. The converter turns each page into pure black and white. Light grey text can drop out, and dark shaded headers can turn solid black and hide their words. A rescan with higher contrast helps.
- Crop away clutter. Stamps, punch holes and dark scanner edges can be read as stray characters. Crop PDF trims margins from every page.
- Use a laptop for long scans. OCR runs on your device and shows progress page by page, so a phone works but takes longer.
Fixing tables in Word
Once the .docx opens, most table repairs take a few clicks.
Text that should be a table. If a two-column block came through as lines, put a tab between the cells (or replace double spaces with tabs), then select those lines and choose Insert > Table > Convert Text to Table with Tabs as the separator.
Values in the wrong column. Where an empty cell pushed values left, click the cell where the gap belongs and insert a cell, or cut and paste the values one column across.
Wrapped cells split into rows. Select the two rows' cells in that column, then use Merge Cells, or copy the second line's text up and delete the extra row.
Appearance. The table arrives plain. Use the Table Design tab to add borders and a header style, and set column widths by dragging the borders.
Accuracy. OCR can misread characters such as 0 and O, or 1 and l, inside numbers. Check totals, dates and reference numbers against the scan. For long passages of prose, paste the OCR text and a version you trust into Text Compare to spot lines that differ.
Other routes, and when to retype
- A single photographed page. Our Image to Text tool runs English OCR on one image and gives you text to copy. It does not build tables or a Word file.
- Google Docs. In Google Drive, right-click the PDF and choose Open with > Google Docs. Google's own help page says files should be 2 MB or smaller, and that tables, lists and columns are not likely to be detected, though bold, italics and line breaks often are.
- Non-English documents. Our OCR reads English only. Google Docs detects the language and supports many more.
Retyping wins more often than people expect. A table of ten rows with many blank cells may be faster to type into a fresh Word table than to repair. Handwritten tables and faint carbon copies are also better typed by hand. Use OCR for the long, clean, printed pages, and keep it out of the way for the rest.
If your PDF turns out to have real text after all, the standard path is simpler, and our guide to converting a PDF to Word without losing text covers it.
Frequently asked questions
Upload it to PDF to Word and press Convert to Word. Because a scan holds no text, the tool reports that and shows a Convert with OCR button. Press it, and each page is read in your browser and written into a .docx as editable text. It reads English only and does not carry over images.
Often, for simple grids. The OCR path turns lines with three or more pieces separated by wide gaps into real Word table cells. Two-column tables, empty cells, wrapped text and merged headers cause errors, so expect to fix some rows in Word using Convert Text to Table or by moving values between cells.
An empty cell leaves no text for the OCR to detect, so the next value in that row moves left into the empty column. Click where the gap belongs and insert a cell, or move the values one column across. Rows with every cell filled usually line up correctly.
No. The Convert with OCR option renders and reads every page inside your browser, so the file stays on your device. Only the standard Convert to Word button sends the file to our server, where it is converted and discarded immediately. For confidential scans, OCR is the more private of the two routes.
Not accurately. The OCR on PDF to Word and Image to Text recognises English only, so other scripts and accented text will come out wrong. Google Docs, opened from Google Drive with Open with > Google Docs, detects the document language and supports many languages, though it rarely keeps tables.