PricingHelpSign in

The ATS resume guide: what actually happens to your PDF

How applicant tracking systems parse a resume, which formatting choices survive, and which ones quietly delete your text. Written from a render pipeline we run ourselves.

Last reviewed

An applicant tracking system is a database with a resume parser in front of it. When you apply, your file is turned into text, the text is cut into fields — name, employers, dates, titles, skills — and a recruiter then searches that database. Nearly everything people call “ATS-friendly” is really about one thing: whether the text your PDF gives back is complete and in the right order.

This guide is written from a render pipeline we run ourselves. Every claim here was measured by rendering resumes to PDF and reading the text back out, not inferred from a vendor’s marketing page.

The five things that actually matter

  1. Real text, not an image. A scanned or exported-as-image resume has no text layer. The parse is empty and no amount of good writing survives it. This is the single most common total failure.
  2. Conventional section headings. Parsers find sections by matching expected words on their own line. “Experience” works; “My Journey” leaves your jobs unassigned to any field.
  3. Canonical order. Summary → Experience → Education → Skills. Systems are calibrated against it, and a resume that extracts in a different order is one they interpret worse — even when nothing is missing.
  4. Dates a machine can read. Mar 2022 – Present beats 3/22–now. Month and year, consistently, on every entry.
  5. Words from the job posting, honestly used. Search is how a recruiter finds you in the database. If the posting says “Kubernetes” and your resume says “container orchestration”, you are not in the result set.

What does not matter as much as you have been told

  • Colour and design. A coloured heading is a drawing instruction; it has no effect on extracted text. What breaks resumes isstructure faked with layout tricks, not decoration.
  • Two columns, categorically. A side column is safe or unsafe depending on how it is built — see two-column resumes and ATS, which concedes the one real cost.
  • “ATS scores” out of 100 from tools that never read your PDF. A score computed from a pasted plain-text box cannot tell you anything about your file, because your file is the thing that might be broken.
  • Keyword density. No mainstream ATS ranks by repetition. Stuffing a term five times does not beat using it once in a sentence a human will read.

The failures you cannot see

This is where most advice stops and where the interesting part starts. There is a class of defect that renders perfectly and extracts wrongly: a CSS property that deletes every digit from your dates, a bullet style that reorders your entire employment history, letter-spacing that splits a heading into individual characters. We hit all three while building this product, and each one now has a test holding it shut. What an ATS reads from your PDF shows the actual extracted output beside the page it came from.

Format, file type and fonts

Send a PDF unless the posting asks for something else — and the reason is not that PDFs are “safer”, it is that a PDF is the only format where what you see is what the parser gets; PDF or Word goes through the trade properly. Fonts matter more than expected, and not for readability: fonts that break resume parsing documents a typeface we had to remove from this product because its embedded subset dropped six digits and three letters out of the preview our own buyers were proofing.

A short checklist

  • One file, PDF, text-based, under about 1 MB.
  • Standard section headings, in the standard order.
  • Month-and-year dates on every role and degree.
  • No text inside images, headers, footers or text boxes.
  • Bullets that are bullets — no pasted •, no tabs used as columns.
  • One page for most people, two once you have roughly eight years of relevant work.
  • Terms from the posting, in sentences that are true. The gap between your resume and a specific posting is a solvable, boring problem.

In this guide

  • What actually breaks resume parsing: a controlled experiment

    We took 33 resume templates, changed one CSS declaration at a time, and extracted the text back out of every render. Which formatting choices destroy text, which silently reorder it, and which turn out to be harmless.

  • What an ATS reads from your PDF

    The same resume, shown twice: the page as it prints, and the literal text a parser pulls out of it, in order. Four ways that text comes out wrong while the page looks perfect.

  • Are two-column resumes ATS-safe?

    Mostly yes, with one real cost nobody mentions. What decides whether a side column extracts in the right order, and what a column does to a two-page resume.

  • Fonts that break resume parsing

    A font can embed correctly, print perfectly, and still lose every digit in your dates. Three measured failures, including one that cost us a shipped typeface.

  • PDF or Word: which should you send?

    PDF, unless the posting says otherwise — and the reason is not the one usually given. What each format actually guarantees about the text a parser receives.

  • Turning your LinkedIn profile into a resume

    LinkedIn will hand you a PDF of your own profile. What that file actually contains, the three things in it that are the wrong shape for a resume, and what to fix first after importing it.