Why do my fonts change when I convert Word to PDF?

You spend an afternoon on a document. You convert it to PDF. The headings are a different shape, a paragraph that ended neatly at the bottom of page 2 now spills three lines onto page 3, and a table that fitted has burst its column.

Nothing is corrupted. What happened has a specific name (font substitution) and once you understand it, it is entirely avoidable.

What a font actually is here

A typeface is a file of instructions for drawing letters. Your document does not contain that file; it contains the font's name, plus your text. Rendering the document means looking up the name on the machine doing the rendering and drawing with whatever it finds.

If the font is not there, something has to be drawn anyway. So the renderer picks the closest available substitute and carries on.

Why substitution moves everything

Every letter in a typeface has a width. Those widths are what determine where a line breaks.

Swap Calibri for Arial and the letters look broadly similar, but Arial is slightly wider. Over a paragraph, the accumulated difference is enough to push one word onto the next line. That extra line pushes the following paragraph down. Repeat over ten pages and the last heading has moved to a different page entirely.

This is why the symptom is almost never "the text looks slightly different" and almost always "the whole layout moved".

Why PDF was invented to fix exactly this

A PDF can embed its fonts. The actual letter-drawing instructions are stored inside the file. A viewer on the other side of the world with none of your fonts installed still draws every character exactly as you did, because the instructions travelled with the document.

That is the property that makes a PDF worth producing in the first place. It is also why the substitution happens at conversion time and never afterwards: once a font is embedded correctly, the layout is frozen for good.

So the whole question is whether the converter had access to your font at the moment it built the PDF.

Making sure it does

Use fonts that travel. For anything going to other people, prefer widely available typefaces. Arial, Times New Roman, Georgia, Verdana, Courier New and Calibri are present on virtually every machine and in every conversion environment.

Check before you convert. In Word, look at the font dropdown for each heading and body style. A font showing in the list but not rendering is a warning sign.

Avoid downloaded display fonts for one-off headings. They are the single most common cause of this problem, and the visual gain is rarely worth the layout risk.

Convert once, then check. Open the PDF and compare page counts with the original. A different page count is the fastest possible detection of a substitution, and takes two seconds.

Doing the conversion

Word to PDF converts the document with its fonts embedded. If a typeface is one the conversion environment does not have, the substitution happens there rather than on the reader's machine, which is better, because at least you can see the result and correct it, instead of every recipient seeing something different from each other.

After converting, verify three things: the page count matches, headings sit where you expect, and tables have not burst.

When it is already too late

If you have a PDF and the layout is wrong, you have two routes.

Fix it in the PDF. OCR PDF edits the text in place, which is right for a stray line or an overflowing cell. You are correcting the symptom, but on a finished document that is usually all you want.

Go back to the source. Fix the font in the original document and convert again. Slower, correct.

What does not work is converting the PDF to Word, fixing it, and converting back. Each conversion is a re-interpretation, and you will have introduced a second round of substitution on top of the first.

Two related surprises

Bold and italic can substitute separately. A family missing only its bold weight gets a synthesised bold (the regular weight artificially thickened) which looks subtly wrong and is measurably wider.

Non-Latin scripts are the strictest case. A font covering Latin characters may have nothing at all for Bengali, Arabic or Chinese. Instead of a substitution you get empty boxes, because there is no approximate answer to draw. If your document mixes scripts, check every one of them in the PDF, not just the first paragraph.

How to check what your PDF actually embedded

You do not have to guess. Every PDF lists its fonts, and the list tells you exactly what happened.

In Adobe Reader: File, then Properties, then the Fonts tab. Each entry shows the font name and, in brackets, whether it is embedded and what type it is.

What to look for:

• "Embedded" or "Embedded Subset" next to every font. This is what you want. A subset means only the characters you used were included, which is normal and keeps the file small.
• A font with no embedding note is being fetched from the reader's machine. That file will look different on a machine that lacks it.
• A font you do not recognise: something like ArialMT where you used Calibri. Is the substitution, named.

Two minutes with that tab settles every argument about whether a document will look right on somebody else's computer.

The subset trap

Subsetting stores only the characters actually used. That is almost always the right trade, until somebody edits the PDF.

If your document embedded a subset containing only the letters in "Invoice 2024" and you then type "Statement" into it, the letters S, t, a, e and m may not be in the file at all. Depending on the editor, you get a substitution for those characters alone, which looks like one word in a slightly different font.

This is why editing a PDF sometimes produces a line where a few letters look subtly wrong and everything else is fine. OCR PDF embeds what it needs when it writes text, which avoids it, but it is worth knowing the cause, because in other tools it shows up as an unexplainable rendering glitch.