Short answer: follow the posting. If the portal says Word, send a .docx. If it takes either, send a PDF that you saved from Word or Google Docs, because it looks identical on every screen and modern applicant tracking systems read its text. What actually breaks parsing is the way a file was made: a scanned page, contact details tucked into Word's header, a layout built from a table. We exported one resume as both PDF and Word and ran every version through our own parser to see where each one fails. Our parser did worse with Word, and we explain below why that is partly our parser's fault.
| Word (.docx) | ||
|---|---|---|
| How the text is stored | Characters placed at positions on a page | Paragraphs in reading order |
| Looks the same on every screen | ✓ | ✕ |
| Easy for a recruiter to edit | ✕ | ✓ |
| Page header and footer | Part of the page like any other text | Stored apart from the body; our reader skipped it |
| Two-column layouts | Reader rebuilds columns from positions | Usually built as a table, which gets flattened |
| Can end up with no text at all | Yes, if scanned or saved as an image | Rarely |
Format facts plus what we saw in our own parser, Nov 7, 2026.
What happened when we exported the same resume both ways
Our app exports every resume as PDF and as Word from the same data, so we could make a fair pair: identical content, two formats, no copy-pasting in between. We did it for a one-column layout and a two-column sidebar layout, then ran all four files through the parser our app uses when someone imports a resume, with its AI clean-up switched off. We checked 20 details: name, email, phone, the number of jobs, each job's employer, title, start date and bullet count, the school, skills and languages.
Details placed in the right field, out of 20, same resume
| PDF, one column | 20 |
| Word, one column | 7 |
| PDF, two columns | 20 |
| Word, two columns | 3 |
All four files gave back 100% of their text. The difference is whether that text landed in the right fields.
Source: RezPolish parse lab, Nov 7, 2026: one resume exported four ways, our own parser with its AI clean-up off
Every word survived in all four files. The structure didn't. From the one-column Word file our parser found six jobs instead of three, left the employer fields empty and filed one company name as a job title. From the two-column Word file, where the sidebar is a borderless table, it found no jobs at all, because every cell's text came out as one long run.
That says more about our parser than about Word. Ours was built PDF-first: it reads font sizes, gaps and column positions off the page to work out what belongs together. A Word file arrives without page positions, so it has less to go on. A parser that reads Word's paragraph structure directly can show the opposite pattern, and the numbers below suggest at least one does.
One more caveat in Word's favour. In the app, an import our rules are unsure about goes through an AI clean-up step that fills in sections the rules left empty, and it flags the fields you should check. We switched that step off here to see what the file format alone does, so a real import of these Word files may come out better. We have not measured by how much. What it cannot do is bring back text it never received, like a Word header.
| Measure | PDF (7,151 files) | Word .docx (1,069 files) |
|---|---|---|
| No reading problems | 32.4% | 51.8% |
| Words glued together | 24.1% | 7.8% |
| Missing standard headings | 13.2% | 5.4% |
| No readable text at all | 3.5% | 0.5% |
Their caveats, which we agree with: these are people who chose to check a resume, and their scanner is one parser, not an employer's system.
Source: ATSVerification, uploads to its scanner, Oct 18 to 31, 2026
Where PDF resumes go wrong
A PDF can be perfect or useless, and they look identical on screen. The failures we saw:
- Scans and photos. We saved a resume page as an image inside a PDF. There was no text to read, and our parser refused it. A phone photo of a printout converted to PDF does the same.
- A header saved as a picture. When the name and contact line was an image, the rest parsed fine and the name, email and phone were simply gone.
- Tight columns. With a 4pt gap between sidebar and main column, lines from both sides were read across as one sentence.
None of these depend on the PDF format itself. They depend on the tool and settings that made the file. A PDF saved from Word, Google Docs or a resume builder that writes real text avoids all three.
Where Word resumes go wrong
Word's problems are more hidden, because the document looks normal while you edit it.
- Contact details in the header. Many Word templates put your name and phone in the page header (double-click the top margin to see it). Inside a .docx file the header is stored separately from the body. Our reader skipped it completely: name, email, phone and city all missing, and the degree line was taken as the name. Some parsers do read headers. You cannot tell which kind an employer runs, so keep contact details in the body.
- Layouts built from tables or text boxes. Our two-column Word file was a table with invisible borders, the usual way to fake columns in Word, and it flattened into one run of text.
- Old .doc files. Our own app refuses them and asks for .docx or PDF; save in the newer format.
- It reflows. Open your .docx on a phone or in a different version of Word and the line breaks move. The one-page resume you made may open as two.
Which format to send, situation by situation
| Situation | Send |
|---|---|
| The posting or portal asks for Word | .docx |
| The portal takes both and you wrote the resume in Word or Google Docs | PDF saved from that app |
| The portal's autofill scrambled your PDF | Try the .docx on the same portal and compare the pre-filled fields |
| Emailing a recruiter or hiring manager directly | |
| A recruitment agency that edits CVs before sending them on | .docx |
| You made it in Canva or another design tool | PDF, but only after the copy test below passes |
| Your only copy is a scan or a photo | Neither. Retype it in a document first |
The autofill row is the most useful trick here. Big portals such as Workday offer to fill the application form from your resume, and the result shows you exactly what their parser made of your file. If your job titles end up in the company box, upload the other format and see if it does better. That's a test of the employer's real system, which no checker website can give you.
Check your file before you upload it
- 1
Copy everything out of the file
Open the PDF in your browser, press Ctrl+A then Ctrl+C, and paste into Notepad. For a .docx, use File, Save As, Plain Text. If nothing comes out, the file is an image.
- 2
Read the pasted text top to bottom
Name, email and phone should be in the first lines. Each job should read title, company, dates, bullets, in that order, with no lines from a sidebar woven in.
- 3
Look in the Word header
Double-click the top margin in Word. If your contact details live there, move them into the body of the page.
- 4
Name the file properly
Use your name and the role, like the examples below. Recruiters download dozens of files called resume.pdf.
- 5
Check the form after you upload
If the portal pre-fills fields from your resume, read every one before you submit and fix what came out wrong.
Priya-Sharma-Data-Analyst-Resume.pdf Rahul-Verma-Resume-Civil-Engineer.docx Anna-Schneider-CV-2026.pdf
No spaces or symbols beyond hyphens, and no "final_v3".
Most of what decides whether a file parses well is the layout, not the extension. Our ATS-friendly template tests cover headings, columns and dates in detail, what is an ATS explains what happens to the file after upload, and the ATS resume checker page has a 10-point check you can run on either format.
Questions people ask
Should my resume be a PDF or a Word document?
Whatever the application asks for. When it accepts both, a text-based PDF from Word or Google Docs is the safe default because it looks the same everywhere. Keep a .docx copy for portals and agencies that want one.
Can an ATS read a PDF resume?
Yes, if the PDF contains real text, which you can check by selecting a word with your mouse. The exceptions are scans, photos and PDFs exported as images: those have no text, and no parser can read them without OCR.
Is DOCX better than PDF for ATS?
It depends on the parser. ATSVerification's scanner found fewer reading problems in .docx files; our own parser placed fields better from PDFs. Both formats work when the file is cleanly made, so check your file rather than trusting a rule.
Should I convert my PDF resume to Word?
Only if you have to. PDF-to-Word converters rebuild the page with their own guesses, and the result can contain text boxes, tables and odd line breaks. If you need a .docx, export it from the app or document you wrote the resume in.
Is a PDF downloaded from Google Docs ATS friendly?
Yes, Google Docs' Download as PDF writes real, selectable text. The usual caveats still apply: plain headings, contact details in the body, and no tables pretending to be columns.
What about old .doc files?
Save them as .docx or PDF. Some portals still accept .doc, but it's an older format that fewer tools read well; our own app asks people to convert it before importing.
