How to Convert a Scanned PDF to Word

To convert a scanned PDF to Word, upload the scan to an OCR tool, let it read each page, and export a DOCX file. The export holds real, editable text in place of page pictures. Before you edit it, compare names, numbers and clause references with the scanned pages, because recognition is never perfect.

A converted file is a new document that happens to say what the scan says. It was typed by software reading a picture, and it has the faults a fast, literal typist would have. The sections below explain what that typist does, how to run the conversion, and how to get from a raw export to a file you would put your name on.

What happens when a scan becomes a Word file?

When a scan becomes a Word file, software recognizes the characters in each page picture, decides which lines form paragraphs, headings, lists and tables, and writes those into a new DOCX. Nothing is carried over from the PDF except what could be read off the image. The Word file is a reconstruction.

That has three consequences worth holding on to.

The text is a reading, and readings contain errors. A clean laser-printed page converts with very few. A third-generation photocopy converts with many. The errors are not marked in the Word file. They look exactly like correct text.

Structure is inferred from appearance. The software sees large bold type on its own line and calls it a heading. It sees aligned columns and calls them a table. Mostly it is right. When it is wrong you get a heading that should be body text, or a table that should be two paragraphs.

Anything that is not text is left behind. A DOCX export from Schema Law carries headings, paragraphs, bulleted and numbered lists, tables and bold type. It does not carry the page images, so signatures, seals, letterhead logos and handwritten initials are absent from the Word file. Fonts, type sizes, italics, underlining, indentation and line spacing are not reproduced either. The document uses Word's default styles until you apply your own.

Page breaks are kept in one specific way: each page of the scan starts a new page in the Word file. That makes it easy to put the scan and the export side by side and find your place. It also means a paragraph that ran over a page break in the original arrives in two pieces.

What survives in more detail, and why, is the subject of the post on whether OCR keeps formatting.

Converting a scanned PDF to DOCX, step by step

The conversion itself is short: prepare the file, upload it, look at the parsed pages, export Word, and save the export beside the scan. Plan on a couple of minutes of your own attention for a typical agreement, plus the time the pages take to be read.

1. Confirm the file is a scan. Try to select a word. If you can, and the pasted text is accurate, the PDF already has text and a direct save to Word from your PDF program may be all you need.
2. Check size and length. Schema Law takes PDF, PNG or JPEG files of up to 20 MB. A plan sets the most pages allowed in one document. If the file is longer, split it where one document ends and the next begins.
3. Upload the file as it arrived. Do not print and rescan it, and do not run it through another converter first. Every extra generation costs sharpness.
4. Read the parsed pages before exporting. Once the document is ready, open it and look at the first page, the signature page and one page from the middle. You are looking for gross failures: a blank page, a page in the wrong order, columns woven together.
5. Export as Word. The DOCX takes its document title from the uploaded file's name.
6. Save with a name that says what it is. For example Harlow lease 2021-03-03 (OCR conversion).docx. Keep it in the same folder as the scan so the two travel together.

If you later need the same document as a searchable PDF for the file, you can export it again in that format without uploading a second time. The post on making a scanned PDF searchable explains when you want that copy as well.

A note on multi-document PDFs. A closing binder or a production volume scanned as one file will convert as one Word document. That is rarely what you want to edit. Split it first, and convert only the documents you intend to work on.

For the Harlow lease, use the Word file as the draft you mark up. Check each covenant, any assignment language and the entire agreement clause against the scan before you change a word. Read the rent figure digit by digit, since that number is what the draft will repeat.

What should you check in the Word file before editing?

Check at least four things before you change a word: that every page is present, that the defined terms and party names are right, that every number matches the scan, and that the clause numbering is intact. Do this with the scan open next to the Word file, on the same page of each.

Work through this list in order. It front-loads the checks that catch the most damaging errors.

• Page count. The scan has 34 pages; does the Word file have 34 page starts? A missing page is easy to miss in a long agreement and fatal to rely on.
• Parties and defined terms. Search for each party name and each capitalized defined term. A name read as "Dunmore Frieght" in one place will not be found by a later search-and-replace.
• Money, dates and percentages. Read every figure against the page image, digit by digit. Look for a dropped zero, a comma read as a period, and "1" swapped with "l" or "I".
• Clause numbers and cross-references. Compare the section numbers down the left margin with the scan, then test a few cross-references ("subject to Section 9.2") to see that they point where they should.
• Negatives and small words. "Not", "no", "nor", "unless" and "except" are short, and a smudge can remove one. Read the operative sentences in full.
• Tables. Count rows and columns in each schedule against the scan.
• Ends of pages. Read the last line of each page and the first of the next. Footers, page numbers and running headers sometimes land in the middle of a sentence there.

For a long document, a full proofread may not be proportionate to the task. The post on why OCR output is full of errors describes how to proofread against the original in a structured way, and how much checking different tasks warrant.

Do not accept spelling suggestions in bulk at this stage. A spell-checker will offer to change an unfamiliar surname or a defined term into a dictionary word, and it cannot see the scan.

Fixing the usual damage in a converted document

Most converted documents need the same handful of repairs: rejoining paragraphs split at page ends, removing stray headers and footers, restoring automatic numbering, and reapplying styles. Do them in that order, since later steps are easier once the text flows as one document.

Paragraphs split across pages

Because every scanned page begins a new Word page, a sentence that crossed a page boundary is cut in two. Turn on the display of formatting marks so you can see the breaks. Decide first whether you want the original pagination at all. For a working draft you usually do not, so remove the breaks and rejoin the split paragraphs. For a document you will cite by its original page numbers, leave them in.

Running headers, footers and page numbers

A scan has no concept of a header. "Page 7 of 34" and "CONFIDENTIAL" are simply text near the edge of the picture, and they can arrive as ordinary paragraphs. Delete them from the body and, if you need them, add them back as a real Word header or footer. Footnotes behave the same way, as the section on footnotes, headers and page numbers describes.

Numbering

List numbers in a converted Word file are typed characters, not Word's automatic numbering, and a list can come out renumbered. The post on whether OCR keeps formatting explains how headings and outline numbering come through conversion. Check every list against the scan. If you plan to add or remove clauses, apply your firm's numbering style so the numbers and cross-references update on their own.

Headings and styles

Recognized headings arrive as Word heading styles, which gives you a navigation pane straight away. The levels may not match your template. Reassign them, then attach the firm template so fonts and spacing follow.

Tables

A table with ruled lines usually converts into a Word table with the first row marked as its header. A borderless schedule may arrive as paragraphs with gaps. For anything numeric, a spreadsheet is the better route, and the post on getting tables from a scanned PDF into Excel covers it.

Signature blocks and exhibits

Signature blocks come through as scattered short lines: "By:", a name, a title. Since the signature image is not in the Word file, mark the block clearly, for example "[signed in original]", so that nobody reads the converted file as a signed instrument.

A common mistake is to tidy as you read, fixing each oddity where you find it. You end up doing the same repair forty times. Scan the whole document for patterns first, then fix each pattern once with find and replace or a style change.

Editing a converted contract without losing track of the original

Treat the converted file as a baseline that must be proved against the scan once, then frozen. After that, every edit is tracked against the frozen baseline, so you can always show what the signed document said and what you changed.

A workable sequence for an amendment or a restated agreement:

1. Convert and verify. Complete the page count, name, number and clause checks with tracked changes switched off. Corrections of OCR errors are not edits to the deal and should not appear as redlines.
2. Save the verified copy as the baseline. Mark it read-only and name it so its status is obvious: (verified against executed copy 2026-10-08). Note in the file who verified it and against which scan.
3. Draft in a new copy with tracked changes on. Everything marked from here is a real change.
4. Compare when in doubt. Word can compare two versions of a document and show the differences as revision marks. Microsoft's instructions for comparing two versions of a document in Word give the path as Review, then Compare, and state that the original copy remains untouched. Comparing the baseline with your draft confirms that nothing changed outside tracking.
5. Cite the scan, not the conversion. When the amendment recites existing language, take the wording from the verified baseline and the page reference from the executed scan.

Two cautions. A redline between a converted document and the other side's native Word draft will show hundreds of differences that are only formatting or OCR residue, so verify before comparing or the real changes drown. And never send the conversion out labelled as the agreement. It is your working copy of the agreement. The executed scan remains the record.

Lease work has its own pitfalls, such as rent schedules and initialed pages, which are handled in the post on converting a commercial lease PDF to Word.

Scanned PDF to Word on a free plan

You can convert a scanned PDF to Word on Schema Law's free plan. It allows one document a day of up to three pages, and Word is among its export formats, as on every plan. That is enough for a short letter or order, or for testing a hard page before paying.

Three things shape how you use it:

• Three pages per document. A longer file has to be cut down first. Extract the pages you need with your PDF program and upload that extract.
• One document a day. Choose the test with care. Pick the pages that worry you most: the faxed one, the one with the table, the signature page with a stamp across it. If those convert acceptably, the clean pages will.
• Nothing expires. A document stays in your account until you delete it, so you can come back and export it again in another format. Delete it yourself once you have what you need.

For regular work the arithmetic changes quickly. A single 40-page agreement is beyond the free plan's per-document limit. The Solo plan covers 1,000 pages a month with documents of up to 250 pages, and the pricing page lists the rest.

Free converters elsewhere vary widely in what they limit and in what they do with the files you give them. The post on free scanned PDF to Word tools compares the kinds of limits you will meet. If the document belongs to a client, read the post on whether free online OCR is safe for legal documents before you upload it anywhere.

Not sure this is the right route for your file? The guide on how to change a scanned document into Word compares the routes by document type and by what the Word file is for.