cvharbor editorial

Resume file formats: what actually parses cleanly

PDF and DOCX are not two versions of the same thing. They store text differently, and that difference is exactly where CVs go missing. Here is what each format contains, what three major ATS platforms do with it, and which export settings quietly break parsing.

A .docx file is a zip archive of XML documents. Open one in any archive tool and you will find a folder structure: document.xml holds the text, in reading order, tagged with the styles applied to it. A .pdf is a page-description format: it records where to draw each glyph on a fixed canvas, and text order is a side effect of how the exporting program wrote those drawing instructions, not a guarantee. That single difference explains most of what follows.

What each format actually contains

DOCX stores your CV as structured text. A parser opens the archive, reads document.xml, and gets your words back in the order Word wrote them, along with which parts were bold, which were a heading, and which sat in a table cell. There is no ambiguity about what counts as text, because everything in that XML is text by definition.

PDF stores your CV as a set of positioned marks. A well-exported PDF from Word or Google Docs includes an invisible text layer alongside those marks, and a parser reads that layer. A scanned or photographed PDF has no text layer at all — only an image of marks — and a parser reading it gets nothing back unless it runs optical character recognition, which most ATS ingestion pipelines do not do by default.

What Workday, Greenhouse and Taleo do with each format

These three systems account for a large share of corporate hiring, and their public developer documentation and support articles describe the same ingestion pattern with different tolerances.

• Workday parses both PDF and DOCX using an internal resume-parsing service, then maps extracted fields — name, employer, dates, education — into structured application fields the candidate is shown and asked to confirm before submitting. Because the candidate sees and can correct the parsed fields, a Workday-specific failure mode is a blank or wrong field the applicant does not notice before submitting, rather than a silent rejection.
• Greenhouse extracts text from the uploaded file and stores both the file and the extracted text against the candidate record; recruiters search that extracted text. Greenhouse's own support documentation recommends PDF or DOCX and single-column layouts, and notes that scanned files and files built as images will not extract usable text.
• Taleo (Oracle Recruiting Cloud) is the oldest of the three platforms still in wide use and has the least forgiving parser of the group in practice — older deployments in particular are more likely to mis-order multi-column layouts and to drop text placed in headers, footers, or text boxes than the other two.

None of the three penalizes a plain, single-column CV in either format. The differences between them show up only at the edges: complex layouts, non-standard section names, and files with no text layer at all.

Word export settings that break parsing

A CV written in Word is not automatically safe once exported. Three settings determine whether the exported file keeps its text structure intact.

1. File > Export > Create PDF/XPS, not Print > Save as PDF. The export path embeds a proper text layer with the document's structure; some print-to-PDF drivers, particularly older or third-party ones, rasterize part or all of the page instead.
2. Headers and footers. Word headers and footers are a separate XML part from the body, and several parsers do not read that part at all. Contact details placed there can be invisible to the parser even though the DOCX is otherwise perfectly structured.
3. Text boxes and SmartArt. Both are stored as floating objects anchored to a position on the page rather than as part of the main text flow. A job title or skills list placed in one of these can be skipped entirely by a parser that reads the body text stream in order.

Google Docs export settings that break parsing

Google Docs exports a clean text-layer PDF by default through File > Download > PDF, which is the safer of its two export paths for this purpose.

• Downloading as DOCX (File > Download > Microsoft Word) converts Google Docs' internal format to OOXML on the fly, and tables built with Google's table tool sometimes convert with extra nested cells that a parser reads as empty fields between your actual content.
• Multi-column layouts built with Google Docs' native "Columns" feature (Format > Columns) export as true PDF columns, which read left-to-right per column — the same interleaving risk described above applies here, and is more common in Docs' column feature than in a Word table-based two-column layout.
• Inserted images of logos or icons sitting close to a job title can shift the text-drawing coordinates around them in the exported PDF, occasionally pushing a line out of its expected reading order.

LaTeX: clean text, unusual structure

A LaTeX-compiled PDF from pdflatex or xelatex embeds a full text layer by default, so the "is there text at all" failure mode common to scanned PDFs does not apply here. The risk with LaTeX resumes is different: many popular resume templates (moderncv, altacv, and similar) build the layout using absolute positioning through packages like tikz or textpos, or use multi-column environments. Those approaches produce a PDF whose internal text order can differ substantially from the visual reading order, because LaTeX is drawing marks at coordinates the same way any other program can, and a text-heavy template with a simple single-column body article-class structure produces a far more reliable text order than a two-column CV template with a sidebar.

Layout choices that help parsers and humans: ATS-optimised resume format for 2025

Our ATS check extracts the text layer from your file the way an ATS would and shows you the order it comes out in. Check your CV

Which format to actually send

Send PDF by default, exported through your editor's dedicated export path rather than a print driver. It fixes your layout everywhere it is opened, and a correctly exported PDF parses as reliably as DOCX in all three platforms described above. Send DOCX only when the employer's application form explicitly asks for it — some older Taleo deployments and some recruiter-facing workflows still request Word specifically because they route the file into a redline or annotation step downstream, not because DOCX parses better.