How to Convert a PDF to Word and Keep the Tables, Images and Layout
Converting a PDF to Word is the single most requested thing people do with documents, and it is also the one where free tools disappoint most often. You upload a contract, wait, and get back a Word file where every line is its own paragraph, the table is a jumble, and there is a watermark across the first page.
It is worth understanding why, because once you know what is actually happening, you can tell a good conversion from a bad one and you stop expecting something no tool can deliver. You can follow along with the free PDF to Word Converter, which does the whole job in your browser without uploading anything.
A PDF does not contain paragraphs
This is the root of everything. People assume a PDF is a document with structure inside it, the way a Word file is. It is not.
A PDF is a set of drawing instructions. It says: put this letter at this position, in this font, at this size. Then the next letter. There is no paragraph, no heading, no list, no table, and often not even a reliable reading order. A table in a PDF is not a table - it is text that happens to have been placed in neat columns.
So "converting to Word" is not really a conversion. It is a reconstruction. The software has to look at thousands of positioned characters and work out, by inference, what a human sees instantly: that these letters form a word, these words form a line, these lines form a paragraph, this line is bigger so it is probably a heading, and these six lines are a table because their columns line up.
That inference is the entire product. It is why Adobe charges for this, and why free converters vary so wildly in quality.
What a good conversion keeps, and what it cannot
Being honest about this saves a lot of frustration.
Reliably recoverable: the text and its reading order, bold and italic (because the PDF records which font each character uses, and the font name usually says "Bold"), font sizes and therefore heading levels, bullet and numbered lists, text colour, centred and right-aligned lines, embedded logos and photographs, and tables - especially ones with drawn borders, because the border lines are recorded in the file and can be read directly rather than guessed at.
Recoverable with effort, imperfectly: multi-column layouts, tables with merged cells, footnotes, and text that flows around images.
Recoverable as pictures rather than as editable objects: charts, diagrams, boxes and arrows drawn as vector graphics. Word has no sensible editable equivalent that a converter could generate, so the honest answer is to capture them as images and put them back in the right place. They look right; you just cannot drag the bars of the chart about.
Recoverable, but only if you give up reflow: the exact visual layout. Precise positioning, coloured panels, custom spacing and complex page furniture can be reproduced - by generating a Word document full of absolutely positioned frames. That looks right and is miserable to edit freely, so it is the wrong answer for a report and the only sensible answer for an invoice. This is a real fork in the road, and it is worth its own section.
If a tool promises a pixel-perfect Word copy of any PDF that also edits like a normal document, it is overselling. You get to pick one.
The fork in the road: reflowing text or a faithful page
Most converters quietly make this decision for you, and it is why the same tool can look brilliant on a report and useless on an invoice.
Think about what an invoice or a receipt actually is. A navy bar across the top. A pale panel with three labels in it. A row of dotted boxes. Amounts pushed to the right-hand edge. None of that is a paragraph. None of it is really a table either, even though it looks like one. It is text dropped at coordinates on top of some coloured rectangles, and the structure your eye sees was never written down anywhere in the file.
Hand that to a converter built for prose and you get exactly what people complain about: the whole page swallowed into one enormous table, words split across cells - "PAYMENT M" in one, "ETHOD" in the next - and the coloured bars thrown away because there was nowhere to put them. The software was not broken. It was answering the wrong question.
So there are two honest answers, and a good converter offers both.
Flowing text rebuilds the document: paragraphs that join up, headings as real Heading styles, bullet and numbered lists, and tables as real Word tables. Type a sentence in the middle and everything below moves down, the way a document should behave. This is right for reports, letters, essays, manuscripts, contracts - anything whose value is the words and which you intend to edit properly.
Matching the layout does the opposite. It redraws the page: every filled rectangle becomes a rectangle, every rule becomes a rule, and every line of text is placed in its own frame at the position it actually occupied. The text stays real text - you click into it and retype a name or an amount - but it does not reflow, because each line is pinned where the PDF put it. This is right for invoices, receipts, certificates, forms, letterheads and statements, where the layout *is* the document and being able to correct one field matters far more than being able to rewrite three paragraphs.
There is a further wrinkle that decides whether the faithful mode actually looks faithful. Word will not have the fonts the PDF used, so it substitutes - and a substituted line is rarely the same width as the original. Left alone, that text overflows its frame and wraps, which stacks lines on top of each other and makes the page look worse than if nothing had been attempted. The fix is to measure each line in the font Word will really use and squeeze or stretch the glyphs very slightly until it ends up the width it had. It is a small detail that is the difference between a page that matches and a page that is a mess.
Similarly, panels in a PDF are usually a mosaic of touching rectangles rather than one shape. Drawn separately in Word, every shared edge shows up as a pale seam across an otherwise solid bar. Growing each rectangle by a quarter of a point makes them overlap instead of abut, and the seams vanish. Nobody notices when it is done; everyone notices when it is not.
If you are unsure which mode you want, the short version is: could you describe your document as a form? Then match the layout. Otherwise, flowing text.
Why "no upload" matters more for PDFs
Most online converters send your file to a server. For a recipe that is fine. For the documents people actually convert, it is worth a moment's thought, because the list is remarkably consistent: employment contracts, invoices, bank statements, tenancy agreements, medical letters, CVs, school reports.
Uploading means a copy of that document exists on a machine you do not control, for a period you have to take on trust. It also explains the business model of free converters: server processing costs money per file, which is why free tiers are capped, watermarked, or ask for your email address.
A browser can do this work itself now. The file is read from your disk, the text is extracted and rebuilt, and the .docx is assembled in memory and saved straight back to your downloads folder. Nothing is transmitted, there is no queue, no account and no watermark - and it keeps working if your connection drops.
Check the structure before you convert
Here is the practical tip that improves results more than anything else: look at what the converter found before you let it convert.
A good tool will tell you how many headings, paragraphs, lists and tables it detected and show you an outline. Thirty seconds spent on that tells you immediately whether it understood your document. If a twenty-page report shows two headings, heading detection has failed. If a document full of tables shows none, table detection has failed. Either way you can change a setting rather than open the result, sigh, and start again.
The settings worth knowing:
Joining lines into paragraphs should be on for ordinary prose and off for anything where the line breaks are deliberate - poetry, addresses, code listings, song lyrics. This single setting is behind most "why is every line a separate paragraph" complaints, and most "why did my address run together" ones.
Heading detection works on relative size. It goes wrong on documents that use bold rather than size for headings, or where body text sizes vary a lot.
Table detection is the one to turn off when a table comes out mangled - loose text in paragraphs is often quicker to fix by hand than a badly guessed table. Though if the thing that came out mangled is a whole invoice rather than one table, the mode is what needs changing, not this setting.
Scanned PDFs need OCR, and that is a different job
If your PDF came from a scanner, a photocopier or a phone camera, it contains no text at all. It contains a picture of text. There is nothing to extract, and a converter that does not notice will hand you a blank document.
There is an easy way to tell before you start: open the PDF and try to select a sentence with your mouse. If you can highlight the words, there is real text. If you can only draw a box over the page, it is a scan.
Reading a scan needs OCR - optical character recognition. Instead of reading characters out of the file, it looks at the shapes in the image and works out which letters they are. Twenty years ago this was expensive specialist software. Today the engine behind most of it, Tesseract, is open source and small enough to run inside a web page, which means a scanned PDF can now be converted without uploading it either.
A few things are worth knowing about what OCR can and cannot do.
Typed text is where it shines. On a clear scan of an ordinary printed document, accuracy is close to perfect. Even a badly degraded scan - skewed on the glass, slightly out of focus, washed out and heavily compressed - usually comes through intact, because the letter shapes are still unambiguous.
Handwriting is not realistically readable. General-purpose OCR is trained on printed type. Handwritten notes, signatures and filled-in forms will produce nonsense, and no free tool will do better.
Pick the right language first. The recognition model differs per language. Running a French document through the English model produces text, but a lot of it will be wrong, and the mistakes are subtle.
Read the result. OCR errors are rarely obvious gibberish. They are plausible words - "rnodern" for "modern", "1" for "l", "0" for "O" - which is exactly the kind of mistake that slips past a quick skim. Always proofread a converted scan before sending it anywhere that matters.
There is also an awkward middle case worth knowing. Some PDFs have a thin layer of poor text sitting over a scanned image, usually from cheap scanning software. Those are the worst of both worlds: a converter sees text and uses it, and the result is worse than if it had run OCR. If a conversion comes out garbled rather than empty, forcing OCR on every page usually fixes it.
After the conversion
Two habits make the result much easier to work with.
Turn on the navigation pane in Word - View, then Navigation Pane - and check the headings. If the conversion detected them properly you get a clickable outline of the whole document, which is the fastest way to confirm the structure came through and to move around a long file.
Then use Word's own styles rather than re-formatting by hand. Because the headings came through as real Heading 1, 2 and 3 styles rather than just bigger bold text, changing how they look is one click each, and the table of contents feature will work.
Try it
The PDF to Word Converter handles both kinds of PDF and both kinds of document. It rebuilds paragraphs, headings, bold and italic, lists, real tables, pictures, colours and alignment from PDFs that contain real text; it redraws forms, invoices and receipts so the Word file matches the page while the text stays editable; and it reads scanned pages with OCR in twelve languages - automatically, only on the pages that need it. It works out which mode your document wants and tells you which it chose, shows you what it found before you commit, lets you convert only the pages you need, and never uploads your file. If you only need the words rather than an editable document, the PDF Text Extractor is quicker, and the rest of the other PDF tools handle merging, splitting and the other everyday jobs.
Ready to turn a PDF into an editable document?