Best font for a resume: we tested 18 fonts in three PDF readers

Which font to use

Pick one plain font and use it for everything. MIT's resume checklist names Arial, Calibri and Times New Roman as easy-to-read examples. Columbia's engineering career office lists Times New Roman, Arial, Calibri, Garamond or Tahoma, and says to use one font throughout.

Every one of those read back cleanly in our test whenever the font was actually on the computer: Calibri through its twin, Garamond as EB Garamond. So did Lato, Roboto, Open Sans and IBM Plex Sans. From there, choose on looks and on how much fits on the page.

• Sans serif: Arial, Calibri, Helvetica, Verdana, Tahoma, Lato or Roboto.
• Serif: Georgia, Times New Roman, Cambria or Garamond.
• Body text at 10 to 12 pt, the range MIT and Columbia both give. Columbia puts your name at 12 to 22 pt.
• Margins of 0.5 to 1 inch, per MIT's checklist.
• Make the PDF on the computer that has your font. A Word file's words arrive intact, but if the employer's computer lacks your font, it draws them in another.

How we tested 18 fonts

We wrote one resume of 22 lines and 84 words, including 15 words with fi, fl or ff in them and 6 bullets, and set it in each font at 11 pt. Calibri and Cambria come with Microsoft Office, so we used Carlito and Caladea, free fonts that The Document Foundation lists as metrically compatible with them: same letter widths, different files. The last font, Brush Script MT, is a script font, there to see whether anything breaks.

Each version became a PDF two ways: Chrome's print to PDF, and LibreOffice Writer's export from a Word file. pdf.js, Poppler's pdftotext and MuPDF read all 36 PDFs back, 108 reads in all, and we checked every line, word, bullet and fi or fl word against what was typed. We also read each Word file's own text.

Of the 18 fonts, 17 came back exactly in all 6 of their reads. The only changed read was Inter in Chrome's PDF, read by pdf.js. Every Word file's own text came back exactly too, including the one in the missing font.

Ligatures came back as plain letters

A ligature joins letters like f and i into one shape. Ten of the fonts drew them, the same ten in both PDFs: Helvetica, Carlito, Lato, Roboto, Open Sans, IBM Plex Sans, EB Garamond, Palatino, Source Serif 4 and Brush Script MT.

In all of them, every one of the 15 words with fi, fl or ff came back as plain letters in all three readers. The ligature is not the risk. A wrong text map behind it is, and that came from a font the computer didn't have.

The one miss: Inter's punctuation

Inter adjusts its punctuation to the letters around it, a feature its site calls contextual alternates. Next to capitals and numbers it draws slightly different brackets, hyphens and dashes. In Chrome's PDF those alternate shapes were mapped to private-use codes, which stand for no character at all.

pdf.js returned those codes in place of the brackets, hyphens and en dash, so 18 of 22 lines came back exactly. pdftotext and MuPDF read the same file exactly, and LibreOffice's PDF of Inter was clean in all three.

Turning the feature off in Chrome, with the CSS font-feature-settings:'calt' 0, removed the private-use codes, and every line came back. If you build your resume as a web page in Inter, add that line, or export from Word or LibreOffice instead.

The real risk: a font the computer doesn't have

A Word file names its font. The program that opens it has to have that font, or it draws your words in something else. We asked for Garamond on a machine that did not have it, then exported to PDF. LibreOffice put Arial Bold and Linux Libertine G in its place.

The stand-in's text map was wrong. It listed letter pairs as runs like “fia” and “flo”, so pdf.js returned 14 of 22 lines exactly and MuPDF 15, with doubled letters and words split apart. pdftotext got all 22. The Word file itself still read back exactly: the damage happened in the conversion.

Harvard's career office tells students to “be sure that your formatting will translate properly if converted to a .pdf”. This is what that looks like. Our earlier dashes and ligatures test ran into the same stand-in, Linux Libertine G, when a different font was missing, and got the same doubled letters.

Size changes the page, not the text

Size didn't change what came back. Arial at 6 pt read back exactly in all three readers. Small type is a problem for the person reading, not for the software.

It did change the page. We set the same 310 words at 11 pt in every font. Palatino used 86% of the page and Source Serif 4 103%, and Open Sans and Source Serif 4 ran onto a second page. Part of that is line spacing: each font sets its own default, and we left it there.

A point size also looks different from font to font. Times New Roman's lowercase letters are 14% shorter than Arial's, so 11 pt Times New Roman looks like about 9.5 pt Arial: it reads smaller and fits more. The table gives the size in each font that matches 11 pt Arial.

What to avoid

These are the things that changed what came back, or made the page harder to read.

• A font the computer making the PDF doesn't have. It cost 8 lines in pdf.js.
• Icon fonts for your contact details. The readers returned each Font Awesome icon as a code with no meaning, such as <U+F095>, except that pdf.js turned the envelope into ✉. When the gap after an icon was styling rather than a typed space, MuPDF stuck the icon to the detail. Plain words like “Email” cost nothing.
• Script and novelty fonts. Brush Script MT read back exactly, but its lowercase letters are 29% shorter than Arial's at the same size.
• Capitals with wide letter spacing in headings. In one of our own designs, pdftotext split “SUMMARY” into “S UM M A RY”.
• Text inside an image. A picture of words has no text layer, so a text extractor gets nothing, whatever the font.

What ATSPrep's designs use

The Studio's designs use Inter, Source Serif 4, IBM Plex Sans, Lora and IBM Plex Mono. Our PDF export embeds each font with its text map, and none of its files drew ligatures or Inter's alternate punctuation. In all five designs we tested, every reader returned all 84 words, fi and fl words included.

One result is worth knowing. The default design, Clean, sets its headings in 7.6 pt capitals with extra letter spacing, and pdftotext split three of its four headings, as in “EXP E RI E N CE”. pdf.js and MuPDF read them whole, and every other design read whole in all three.

Our designs set body text at 8.9 pt to 10 pt, smaller in points than the 10 to 12 pt advice. Inter at 9.4 pt is about as tall as Arial at 9.9 pt. If your page has room, the Body size control goes up to 12 pt. Word downloads use Arial, Georgia or Calibri.

Check your own file

Most of what broke in this test shows up in a one-minute check.

• Copy everything in your PDF and paste it into a plain text editor. Look for doubled letters, missing brackets or dashes, and words split apart.
• In Adobe Acrobat Reader, File > Properties > Fonts lists the fonts the PDF embedded. If your font isn't there, the file was made with a stand-in.
• If you used icons, check that each contact detail pastes as its own word.
• Run the file through the ATS resume checker to see the text it gives back, line by line.

Common questions