PDF vs DOCX Resume: ATS Parsing in 2026

Most job seekers treat the file-format question as a coin flip: PDF looks sharper, Word feels safer. In 2026 that guesswork is expensive. U.S. employers had 7.1 million job openings and made 5.2 million hires in August 2026, according to the Bureau of Labor Statistics JOLTS release — a market where every posting collects a crowd and the software reading your file decides whether a human ever opens it.

The good news: the format itself is rarely why a resume disappears. What kills applications is how the file was built. A clean, text-based PDF parses as reliably as Word. A design-tool export saved as a PDF is a photograph of a resume, and applicant tracking systems read photographs badly.

This guide separates the two questions — extension and structure — using what the ATS vendors actually publish, then gives you a 90-second test you can run before you upload.

The Short Answer

Default to a text-based, single-column PDF. It preserves your layout for the human reviewer who eventually opens it, and modern parsers extract it cleanly.
Send DOCX when the portal asks for it, or when you are applying through an older Oracle Taleo deployment or a government system that explicitly prefers Word.
Never send an image. A scan, a phone photo, or a PNG-style export of a resume has no text layer, and no amount of keyword optimization survives that.
If the posting says "PDF only," send PDF. If it says "PDF or Word," send PDF. Never send a format the form does not list just because a blog post told you to.

That is the whole decision. Everything below explains why, and how to verify you got it right.

What the ATS Vendors Actually Document

Vendor documentation is the closest thing to ground truth, because it describes what the software will and will not ingest.

Greenhouse's supported-formats article lists candidate uploads as .doc, .docx, .pdf, .rtf and .txt, up to 100 MB. So the extension is not the filter — all five are accepted by design.

What Greenhouse does publish is a separate troubleshooting page on unsuccessful resume parses, and its list of causes is almost entirely structural:

Text with spaces between letters, which looks cohesive to the eye but reads as broken words
Resumes that include graphics, photos, or word art
Resumes uploaded as an image rather than as a document
Complex resumes with tables, headers and footers
Name and contact information placed in a header, footer, or text box
Columned layouts
Resumes without clear sections, or with a different format in every section
Abbreviated job titles such as "Sr. Account Exec" instead of "Senior Account Executive"

Workday is even more explicit. Its developer documentation for the Resume REST API states that the endpoint accepts DOCX and PDF, and that the PDF "must contain text that's not image-based." Workday's Upload My Experience feature accepts .pdf, .doc, .docx, .htm and .txt, and likewise does not support image-based files.

Read those three pages together and the rule writes itself: PDF is fine. A picture of a PDF is not. The file extension is a container; the parser cares whether real characters live inside it.

Why the Image PDF Is the Real Killer

A PDF can hold text two ways: as encoded characters, or as pixels arranged to look like letters. Both look identical on screen. Only one is readable by software.

Three very common workflows produce the pixel version:

Design tools. Canva, Figma, Illustrator and similar tools often export layouts where text is flattened, subset in unusual fonts, or baked into vector shapes. The file opens fine and prints fine. Extract it and you get fragments.
Scans and photos. Printing a resume and re-scanning it, or photographing it with a phone, produces an image with no text layer at all.
Screenshot-style exports. Some "save as PDF" paths on mobile produce a single flattened image per page.

An independent June 2026 test published as the ATS Autopsy tried to quantify the damage: 48 synthetic resumes across nine layouts, each saved as both PDF and DOCX, read by three open-source extractors (pdfplumber, pdfminer.six, python-docx) for 600 scored field reads. The findings are worth quoting precisely:

Clean single-column resumes lost 0% of fields in every format.
A resume whose contact block was built as an image lost 37.5% of fields, dropping the name, email and phone on every single read.
By file type, PDFs lost 19.9% of field reads against 0% for Word — but nearly all of that loss came from a handful of designed layouts, not from the PDF format itself.

Treat the sample size as a floor on what can go wrong rather than a verdict on any named ATS — these extractors do not run OCR, and some commercial systems do. The directional lesson holds: layout and text layer dominate, extension barely registers.

Why Two-Column Layouts Break Parsers

Text extractors read a page by position, roughly top-to-bottom and left-to-right. In a single-column resume that is the same order a human reads, so everything lines up. In a two-column resume, the parser walks across both columns at the same vertical position and interleaves them.

The result looks like this in the parsed record:

Senior Product Manager | 2019–2021 Python, SQL, Tableau Led roadmap for 3 teams | SKILLS

Your title, dates, and achievements get fused with sidebar content into one unusable string. The recruiter sees a candidate with no readable job history — not because you lacked keywords, but because the reading order was wrong.

Sidebars are the worst offender because they invite you to put skills there. If a keyword list lives in a right-hand column, expect a portion of it to be attributed to the wrong employer or dropped entirely.

The 90-Second File Check

Run this before every upload. It catches roughly everything that matters.

Select all. Open the PDF and press Ctrl+A (Cmd+A on Mac). Text should highlight as flowing words. If whole blocks light up as rectangles, you have an image.
Copy a bullet. Paste one achievement into a plain text editor. If the words arrive in order and intact, extraction will likely succeed.
Search the file. Use Ctrl+F for your last name. If the viewer cannot find text that is visibly on the page, the text layer is missing or broken.
Check the file size. A text-only one-page resume typically lands between 40 KB and 120 KB. A one-page resume over 1 MB is carrying images.
Export from source. Always export to PDF from Word, Google Docs, or your editor's export function. Never re-save a PDF from a preview window or a screenshot.
Name it plainly. Firstname-Lastname-Resume.pdf — no version numbers, no symbols, no spaces if you can avoid them.
Re-open the exported file on a second device if it matters. Font substitution is a real cause of layout shifts between machines.

PDF vs DOCX by Portal Type

Modern SaaS ATS (Greenhouse, Lever, Ashby, iCIMS, Workday cloud): a text-based PDF is the safe default. These systems are built to ingest PDFs from day one.
Legacy enterprise systems, especially older Oracle Taleo versions: DOCX is the lower-risk upload. Several independent tests report Word parsing more cleanly on legacy stacks.
Government and public-sector portals: follow the instructions literally. Some accept only certain formats, and some have their own internal resume builders that ignore your file.
Portals that say "PDF only": send PDF. Do not test the parser's tolerance.
Portals that say "either": send PDF. Better visual fidelity for the human stage, equivalent parsing when built correctly.
Emailing a human directly: PDF. It renders identically on their machine.
"Paste your resume as text" fields: paste plain text and strip bullets and smart quotes. Formatting characters are noise in that box.

Keep a DOCX master of your resume. It takes seconds to save a second copy when a portal fights back, and it beats retyping your history into a web form.

Formatting Rules That Matter More Than the Extension

These rules come straight from the parse-failure lists vendors publish, and they matter far more than which container you pick.

One column, always. No sidebars, no two-column "modern" templates, no text boxes for the summary.
Standard section headings. Use "Professional Experience," "Education," "Skills." Creative headings like "My Journey" give the parser nothing to anchor on.
Contact details in the body of page one. Never in a document header, footer, or text box — those are frequently excluded from extraction.
No tables. Even single-column tables confuse field mapping in some parsers. Use plain paragraphs with hanging indents instead.
Full job titles, spelled out. "Senior Account Executive," not "Sr. Acct Exec." Keyword matching runs on the expanded form.
Consistent date format. Pick Month YYYY or MM/YYYY and use it in every single entry. Mixed formats break the employment timeline.
Standard fonts. Calibri, Arial, Georgia, Garamond, Times New Roman. Embedded or subset display fonts are a common extraction failure.
Real bullets only. Use the built-in list tool in your word processor, not typed hyphens with manual indents, and never use decorative glyphs from an icon font.

If you want the full formatting playbook, the ATS-friendly resume tips guide walks through layout decisions in detail, and the resume keyword guide covers what actually gets matched.

Common Mistakes That Cost Interviews

Sending a design-tool export. The portfolio looks great. The parser sees a picture. Recruiters report the attachment without readable fields.
Putting your contact block in a graphic. Banner-style name headers are the single fastest way to lose your phone number.
Using a two-column template from a free builder. It is the most common structural failure in circulation.
Building the resume in the header and footer. Page numbers and repetition across pages seem tidy; they are extraction dead zones.
Uploading a scan of a printed resume. No text layer, no fields, no interview.
Sending a password-protected or restricted PDF. Extraction fails outright. Remove protection before uploading.
Sending a DOCX with tracked changes or comments still in it. The reviewer opens it and sees every edit and every snarky note you left yourself. Accept all changes, delete comments, and check document properties.
Assuming the extension guarantees anything. .pdf is not a promise. The text layer is.

FAQ

Does DOCX parse better than PDF? Not on modern systems. Greenhouse accepts both, Workday accepts both, and clean single-column files of either type extract nearly perfectly. DOCX has an edge only on older legacy stacks and in a few vendor-specific cases.

Will an ATS automatically reject a PDF? Vendors do not document format-based auto-rejection; they document parse failures. In practice a failed parse means blank or partial fields, which pushes your record to the bottom of a ranked list a recruiter may never scroll through. The outcome feels like rejection, but the mechanism is missing data.

What about Google Docs, Pages, or LaTeX? Export to PDF from inside the app and download the file — never share a link, because the upload field will not follow it. Google Docs PDF exports parse cleanly. LaTeX PDFs parse well when they are single-column with conventional section headings; the risk is custom headers and heavy typography.

Should a designer send a creative resume? Send the creative version as a portfolio or website link, and send a conventional single-column PDF to the application form. You get both audiences what they need without gambling on the parser.

Does the file format change how I rank? Ranking runs on parsed fields matched against the requisition. If the fields are empty, there is nothing to rank — which is why an unreadable file is far more damaging than a missing keyword. For the fuller picture on where qualified candidates actually get lost, read the 2026 ATS rejection analysis.

Pre-Submit Checklist

The last item is the one that moves outcomes. A perfectly parsed resume that says "managed projects" will still lose to one that says "managed a $2.4M migration across 3 teams," because parsing only gets you into the ranking — relevance wins it. If you are still tightening that layer, skills-based hiring in 2026 explains how requirements are actually scored.

Check Your File Before You Apply

Two minutes of verification beats another fifty applications that never surface. Run your resume through the free analyzer at applystudio.app to see what an ATS extracts from your file, which keywords your target job description expects, and where your bullets are leaking matches. You will know within a minute whether your PDF is a document or a picture — and that single answer explains a surprising share of the silence in your inbox.