The short answer: a text-based PDF is the best default, a clean DOCX is a close second, and the real danger isn't the extension — it's how the file was made. Here's what actually happens to each format inside a parser.
How each format gets read
Text runs pinned to coordinates. Reading order is reconstructed — layout is frozen forever.
PDF — a drawn page
<document>
<body>
<p>Experience…</p>
<list>bullets…</list>
</body>
<header>phone, email</header>
<textbox>skills</textbox>
A real document tree — but headers and text boxes live outside the body, where many parsers never look.
DOCX — an XML tree
A PDF is a page-drawing format. The parser doesn't read paragraphs; it reads positioned text runs and reconstructs reading order from coordinates. That works beautifully when the PDF came from Word, Google Docs, or LaTeX — and fails when it came from a design tool that exported your text as curves or a flattened image.
A DOCX is a zip file of XML. The parser reads the document tree directly — real paragraphs, real lists. That's more structure to work with, but also more places to hide content: text boxes, headers, footers, and shapes live outside the main body tree, and many parsers never look there.
Failure modes by format
| Failure | DOCX | |
|---|---|---|
| Text exported as image (Canva-style) | Parses as empty | Rare |
| Contact info in header/footer | Usually survives | Often skipped |
| Text boxes / floating shapes | Usually survives | Frequently skipped |
| Layout reflow changing line order | No — layout is frozen | Possible across Word versions |
| Fancy template styling lost | No | Sometimes |
| Ligature/encoding garbage (fi → junk) | Possible with exotic fonts | Rare |
Two takeaways from that table. First, the PDF column's problems are all creation-time problems — export from a text-first editor and they vanish. Second, the DOCX column's problems are all invisible — the file looks fine in Word while the parser silently drops your header.
The decision rule
- 1
The portal asks for a specific format? Send exactly that.
The employer knows what their own system parses best — this overrides every other rule.
- 2
No preference stated? Send PDF from Word or Google Docs.
Never print-to-PDF from a browser, and never a design-tool export — both can destroy the text layer.
- 3
Recruiter emailing your resume around? Use DOCX.
It travels better through the editing and re-saving workflows agencies actually use.
See the full layout rules in the ATS-friendly resume format.
Test the actual file, not the theory
The free JobFitAI checker parses your real upload the way the vendor systems do and shows you the extracted result — if your phone number, dates, or skills didn't survive the trip, you'll see it before a recruiter doesn't.
Written by
JobFitAI Team
The team building JobFitAI's deterministic scoring engine — nine evidence-anchored axes, a nine-vendor ATS parse simulation, and every point backed by receipts.