PDF or Word resume? What each ATS actually accepts in 2026

Send the PDF. Greenhouse, Workday, SmartRecruiters and Ashby all name PDF and Word among the file types they accept, and none of them lists either format among the things that break a parse — so the decision people spend twenty minutes on is not the decision that costs them interviews.

Two other things on the same upload screen do cost them: how big the file is, and whether the letters inside it are letters or pixels. Both have published numbers behind them. What follows is those numbers, the one place where "accepted" and "readable" are not the same thing, and the short list of moments when the Word file is genuinely the better send.

Does an ATS reject PDF resumes?

No. Greenhouse's supported-formats page lists .doc, .docx, .pdf, .rtf and .txt. Workday's HiredScore candidate profile — the parsed "smart resume" a recruiter reads instead of your original layout — supports "DOC, DOCX, PDF, RTF, and TXT". SmartRecruiters says "the accepted files are PDF, DOC(X), RTF, JPG, and PNG". Ashby accepts documents in "pdf, doc, docx" alongside images and video.

Four systems, four help pages, PDF on every one of them. The claim that an ATS "can't read PDFs" is old enough to have been true once, and it is not something any of these four vendors says about its own product today.

What each system says it accepts

Accepted resume file types and upload limits, four ATS platforms

From each vendor's own documentation as it reads in September 2026, listed in sources. Accepted means the upload succeeds — not that every field is read.

	Accepted file types	Maximum file size
Greenhouse	doc, docx, pdf, rtf, txt	100 MB (parses up to 2.5 MB)
Workday (HiredScore resume view)	DOC, DOCX, PDF, RTF, TXT	Not published
SmartRecruiters	PDF, DOC(X), RTF, JPG, PNG	2 MB per file
Ashby	pdf, doc, docx, jpg, jpeg, png, mp4	50 MB

The interesting column is the second one. A file that a paralegal uploads without a second thought to an Ashby careers page — 8 MB, because a scanned signature block sits in the header — is four times the SmartRecruiters ceiling and would be refused outright there. The accepted formats converge across vendors. The limits do not, and nobody puts them on the upload button.

One entry on three of the four lists deserves a mention precisely because nobody uses it. RTF is accepted by Greenhouse, by Workday's HiredScore resume view and by SmartRecruiters; Ashby's document list is the one exception. It has been the quiet safe format for a decade, plain enough that no parser trips on it and structured enough to keep bold headings. It is also, on the recruiter's screen, an unstyled document that looks like it was typed in 1998. That trade is a bad one for a marketing manager whose layout is part of the argument, which is why the format survives on help pages and not in job searches.

The limit that actually stops people is 2.5 MB

Greenhouse accepts uploads "up to 100 MB". It also says, on a different help page, that it "can't parse resumes larger than 2.5MB". Both are true. The gap between those two numbers is where applications quietly go wrong.

Here is the sequence. A UX designer exports a portfolio-styled resume at 6 MB because two case-study screenshots are embedded at full print resolution. The upload succeeds — it is well inside 100 MB — and the confirmation page says the application was received. But the parser never opened it, so the candidate record arriving in the recruiter's queue has no name pulled from the document, no email address, no current title. The file is attached. None of it is searchable.

That failure is worse than a rejected upload, because a rejected upload tells you something is wrong while you can still do something about it. From the applicant's side, SmartRecruiters' hard 2 MB cap is the friendlier design: you find out at the moment you can fix it. Under 2 MB clears every limit in the table above and sits well below the Greenhouse parse ceiling, which is why it is the only size target worth memorising.

Why does the same file parse well on one career site and badly on the next?

Because the parsers are different products looking for different fields, and because each employer configures what its form will accept from them. Workday's course manual for recruiting administrators says the parsed data includes personal and contact information, work experience and education. Its administrator guide then adds that resume parsing does not auto-fill Languages, Skills, or any field an employer has configured as hidden.

Read those two pages side by side and they do not quite agree — the course manual lists Languages among the parsed categories, the administrator guide lists it among the fields parsing leaves alone. That is not a gotcha aimed at Workday. It is the clearest evidence available that "what gets filled in" is a per-tenant setting rather than a property of your file, which is why no amount of reformatting will make the Languages box populate at an employer that has it switched off.

So an accountant applying to two companies that both run Workday can watch the same PDF pre-fill a clean employment history at one and leave half the form blank at the other. Workday states the underlying caveat plainly: "Resume parsing results can vary based on resume format and order of words." Nothing about the file changed between those two applications.

The practical consequence is a test you should stop running. A good pre-fill is not proof your file is healthy, and a bad one is not proof it is broken — both outcomes are partly about a configuration you cannot see. The copy-paste check is the only one that tells you something about the file itself.

The format that really fails is neither PDF nor Word

It is a picture. Greenhouse names "resumes that are uploaded as an image, rather than a document (such as a .docx or .pdf)" among the causes of an unsuccessful parse, and Workday's own parsing guidance is blunt: "For best results, use resumes that don't have images or image-based styles."

The genuinely risky detail in the format lists above is that two of them accept image files at all. SmartRecruiters accepts JPG and PNG; Ashby accepts jpg, jpeg and png. So a registered nurse who photographs a printed resume on a phone and attaches the JPG gets a green checkmark and a submitted application, and every word on that page is invisible to the software behind it. The vendors are not doing anything wrong here — allowing an image attachment is sensible for a portfolio piece or a licence certificate. But "the form took my file" is not evidence that anything read it, and that is the assumption the checkmark encourages.

Those rows are not ranked by quality. They are ranked by whether the file contains characters at all, and the bottom two do not — the only distinction that matters on this screen.

Here is what the recruiter sees when a financial analyst sends the scan. Before:

Name: (blank) Email: (blank) Current title: (blank)

The same person, the same words, exported from the document instead of photographed:

Name: Dana Okoro Email: dana.okoro@example.com Current title: Financial Analyst

There is a thirty-second test that separates the two. Open the file, select all, copy, then paste into a plain text box. If your words appear, you have a document. If nothing appears, you have a picture of one.

Is there ever a reason to send the Word file?

Three, and they are narrower than most advice suggests. The first is that the form asks for it — some application fields accept only .doc or .docx, and arguing with a file picker is not a strategy. The second is that a recruiter has told you they will edit the document before passing it on, which is normal on agency submissions and internal referrals, and a locked PDF makes their job harder. The third is that your PDF export is broken in a way you have not checked yet, which the copy-paste test settles in half a minute.

Outside those three, keep the PDF. A DOCX re-flows on the reader's machine according to whichever fonts and page setup they happen to have, so the two-line summary you fitted carefully at the top of page one can land as three lines that push your most recent role below the fold. The parsed text is identical either way. What a human eye lands on first is not, and on a screening pass that is most of the document's job.

There is a fourth format worth knowing about, though it serves a narrow group. Greenhouse accepts txt and Workday's HiredScore view supports TXT, so a plain-text resume parses cleanly wherever it is allowed. Ashby is the exception in both directions: its candidate-files documentation says "certain text based formats (such as md) are not supported", while a product update dated 21 August 2026 added .txt and .md to application-form file upload fields. A technical writer who keeps a resume in Markdown should convert it before applying rather than find out which of those two rules the form is running.

What none of these pages will tell you

None of the four vendors publishes a comparison of how accurately its parser reads a PDF versus a DOCX. They publish what they accept and what makes a parse fail; they do not publish accuracy rates by format, and there is no public benchmark underneath the confident percentages you will see quoted on this topic elsewhere. The closest any of them comes is Workday's acknowledgement that "resume parsing results can vary based on resume format and order of words" — layout matters, with no number attached to how much.

So treat any claim that one extension parses "X percent better" as unsourced, however firmly it is stated. What the documentation actually supports is narrower and more useful than a percentage: both formats are accepted everywhere, images are not readable anywhere, and the size limits are real and unadvertised.

It is also worth remembering who is at the other end of the upload. The Bureau of Labor Statistics counts 939,700 human resources specialists employed in 2025, puts their median annual wage at $75,940 as of May 2025, projects 6 percent growth in the occupation from 2025 to 2035, and describes the job as recruiting, screening and interviewing applicants and placing new hires. The parser decides what your record looks like in their list. A person decides what happens to it.

How to export a file that survives the upload

Five minutes, once, covering every application after this one

1. 1
   Export, do not scan

   Use File > Export or Save as PDF from Word, Pages or Google Docs. Never photograph or scan a printout: the result is a picture of your resume, which Greenhouse names as a cause of failed parsing.

2. 2
   Check the size

   Right-click the file and read its size. Under 2 MB clears every limit in this post, including the 2.5 MB ceiling above which Greenhouse stops parsing.

3. 3
   Find the megabytes

   A resume over 2 MB almost always holds a headshot, a company logo or a scanned signature. Remove them, or re-insert the image at screen resolution instead of print resolution.

4. 4
   Run the copy-paste test

   Open the PDF, select all, copy, paste into a plain text editor. Your words should appear. Blank output means the file has no text layer and no parser will read it.

5. 5
   Keep the source file

   Keep the editable DOCX beside the PDF, named the same way. When a form insists on Word, you send it in five seconds instead of rebuilding the document from memory.

Do this once and it holds until you next rewrite the resume itself. The step people skip is the third, and it is the one that decides whether the second passes.

What to do this week

Check the size of the file you are sending right now. If it is over 2 MB, open it, find the image, and take it out — that single edit moves you from "attached but unparsed" to "attached and readable" on the strictest system in this post, and it takes about a minute.

Then run the copy-paste test on that file and on the older version still sitting in your downloads folder, because scanned resumes survive in a job search for months without anyone noticing. Last, keep the PDF as your default send and the DOCX as the backup for forms that demand Word. Two files, one folder, no further decisions to make on a Tuesday night.

If the part you are tired of is retyping the same six fields into the form after you have uploaded that file, the platform can fill the application from the resume you just fixed — the same name, email and title the parsers in this post are reaching for.