Font Auditor

Every font you see in a PDF is one of four kinds. Understanding the differences helps you choose better at the source.

The four kinds of fonts in a PDF.

Kind	Common name	Embedded?	Verdict
Standard 14 PostScript reference	Helvetica, Times-Roman, Courier, Symbol, ZapfDingbats and bold and italic variants	No, name only	Unsafe. PDF/UA-1 fails. Compression makes it worse.
Type 1 with FontFile	Older Adobe PostScript fonts	Yes, as FontFile	Acceptable but ageing format.
TrueType subset with FontFile2	Modern Microsoft Word, InDesign, LaTeX, Pages exports	Yes, as FontFile2	Safe. The default for most modern tools.
OpenType with FontFile3	Modern Adobe fonts, web fonts, CJK fonts	Yes, as FontFile3	Safe.

The auditor flags the first kind as BROKEN. It flags the others as SAFE if they have a Unicode mapping (a /ToUnicode CMap) and a reasonable subset coverage. A Type 3 font (a procedural font built from drawing instructions) is also flagged, because Type 3 has no Unicode mapping and screen readers cannot read it.

Compression is where PDFs that look fine become unreadable. Each compressor treats fonts differently.

How common PDF compressors handle fonts.

Tool	Setting	Effect on fonts
Ghostscript	-dPDFSETTINGS=/printer or /prepress	Preserves embedded fonts. Safe for accessibility.
Ghostscript	-dPDFSETTINGS=/screen or /ebook	Strips Standard 14 embeds, may drop other subsets. Unsafe.
Adobe Acrobat Pro	Save As Other > Reduced Size PDF	Preserves embedded fonts. Safe.
Adobe Acrobat Pro	Save As Other > Optimized PDF	Safe if "Embed all fonts" stays on and "Do not embed" list is empty.
Adobe Acrobat Pro	Save As Other > Compressed PDF	Aggressive. Can strip embeds. Avoid for accessibility.
macOS Preview	File > Export > Reduce File Size	Drops embedded fonts on documents above 1 MB. Avoid.
Smallpdf, ILovePDF, online compressors	Any setting	Mostly run Ghostscript /screen or worse. None preserve PDF/UA. Avoid.
Microsoft Word	Print to PDF (any "reduce" option)	Re-prints the document. Produces a fresh untagged PDF. Catastrophic.
Google Drive	Right click > Compress PDF	Inconsistent internal pipeline. Avoid.

Choose from this short list at the source. All are widely supported, freely licensed where noted, and embed cleanly across every common authoring tool.

Fonts that survive compression.

Role	Recommended	Notes
Sans-serif body and headings	Source Sans 3, Inter, Lato, IBM Plex Sans, Noto Sans	Open Font License. Broad Unicode coverage. Strong screen rendering.
Where you wanted Helvetica	Arial (system, Windows) or TeX Gyre Heros (LaTeX)	Subset and embed cleanly. Visually almost identical.
Serif body	Source Serif 4, EB Garamond, Lora, TeX Gyre Pagella	All subset cleanly.
Where you wanted Times-Roman	Times New Roman (system) or TeX Gyre Termes	Visually identical.
Monospace (code, tables)	JetBrains Mono, Source Code Pro	Ligatures off-by-default, which screen readers prefer.
Mathematics (LaTeX only)	Latin Modern Math, STIX Two Math	Both ship MathML mappings.
Indic and other Asian scripts	Noto Sans Devanagari, Noto Sans Tamil, Noto Sans Bengali, etc	Use only the scripts your document needs.

Avoid these PostScript names by their exact spellings: Helvetica, Helvetica-Bold, Helvetica-Oblique, Times-Roman, Times-Bold, Times-Italic, Courier, Courier-Bold, Symbol, ZapfDingbats. Visually identical replacements (Arial, Times New Roman, Courier New) embed cleanly. The names matter, not the look.

The fix for almost every BROKEN finding is a setting in the tool you used to make the PDF. Here is the specific path for each common tool.

Adobe InDesign

File > Export > Adobe PDF (Print or Interactive). In the export dialog, go to Advanced > Fonts and set "Subset fonts when percent of characters used is less than: 100". This forces every used font to embed as a subset. Use the dedicated PDF/UA preset shipped with InDesign 2024 and later, or the "Smallest File Size" preset modified with the above setting.

Avoid the PDF/X-1a:2001 preset for accessibility work. That preset converts text to outlines (each letter becomes a path), which means the structure tree has no text and screen readers cannot read it.

Microsoft Word

File > Save As > PDF. Click Options, tick "Document structure tags for accessibility". Under File > Options > Save, tick "Embed fonts in the file". Word's Save As PDF embeds TrueType subsets by default when these are on.

Never style any text with the literal font name "Helvetica" in Word. Word respects the name and may not embed. Pick Arial, Calibri, Aptos, or Source Sans instead. Avoid Print to PDF in any flow that needs accessibility; it produces an untagged PDF.

LibreOffice and OpenOffice

File > Export As > Export As PDF. In the General tab, tick "Universal Accessibility (PDF/UA)". LibreOffice always embeds fonts. Defaults are accessibility-safe on a clean source document.

LaTeX

Use lualatex or xelatex with \usepackage{fontspec} and a TrueType or OpenType font, for example:

\usepackage{fontspec}
\setmainfont{Source Serif 4}
\setsansfont{Source Sans 3}
\setmonofont{JetBrains Mono}

Both engines always subset and embed. Avoid pdflatex with the historical Computer Modern Type 1 fonts unless you also load cm-super or lmodern. For mathematics, add \usepackage{unicode-math} and \setmathfont{Latin Modern Math}.

Affinity Publisher

File > Export. Choose the "PDF (for export)" preset rather than any "PDF/X" preset. Under PDF Options > More, leave "Convert all text to curves" unticked (otherwise text becomes silent paths and screen readers cannot read it). Tick "Embed fonts" and "Subset all fonts".

Apple Pages

File > Export To > PDF, image quality "Best". Pages embeds TrueType fonts by default. The result is a tagged PDF on macOS 14 and later. As with Word, never style text with the literal name "Helvetica"; Pages may substitute aggressively.

Pandoc

Pandoc is a converter, not a typesetter. Pandoc plus lualatex is the safest path for accessible PDFs. Pandoc plus weasyprint also works for simpler documents. Avoid Pandoc plus wkhtmltopdf; it produces no structure tree at all.

QuarkXPress

File > Export > Layout as PDF. In the PDF Style settings, ensure "Subset fonts below 100% of glyphs" is checked, and confirm "Include Logical Structure (PDF Tags)" is on. Use the "PDF/UA-1" output preset from QuarkXPress 2023 onwards.

Google Docs

File > Download > PDF Document. Google Docs embeds fonts when you download as PDF. The result has limited PDF tagging compared with Word or InDesign. For documents that must meet PDF/UA-1, re-author in Word or InDesign rather than relying on a Google Docs export.

• Using the font picker to select "Helvetica" by name. On most platforms this is a Standard 14 reference, not an actual font file. Pick Arial, Source Sans 3, or any other TrueType font instead.
• Converting text to outlines or paths during export. The text looks identical but the underlying letters are shapes, not characters. Screen readers and search find nothing.
• Using Print to PDF on a tagged source. Print to PDF strips the structure tree. The result has no headings, no alt text, no reading order. Always Save As PDF or Export to PDF.
• Compressing a finished PDF with an online tool. Most online compressors strip embeds. Use the original tool's "save smaller" option or Adobe Acrobat's "Reduced Size PDF" instead.
• Embedding only the characters used today. If you re-edit the file later and add characters that were not in the original subset, they may render as missing glyphs. Some tools let you choose "embed full font" rather than "subset"; for short documents this is sometimes safer.