---
title: What Is Resume Parsing? The 5-Stage ATS Process Explained
description: Resume parsing converts your file into ATS-readable data across 5 stages. See
  what breaks each stage and run an 8-point checklist to fix your resume.
type: article
url: https://www.foundrole.com/blog/what-is-resume-parsing
date: 2026-06-02T20:48:23Z
og_description: A machine reads your resume before any human does. Here is what breaks at each
  of the 5 parsing stages, and the 8 fixes that get you through.
og_image: https://www.foundrole.com/img/pages/pga0mn/what-is-resume-parsing.png?v=2
breadcrumbs:
  - label: Home
    url: https://www.foundrole.com/
  - label: Blog
    url: https://www.foundrole.com/blog
  - label: Resume & Cover Letters
    url: https://www.foundrole.com/blog/category/resume-cover-letters
---

**Author:** Jessica Baker
**Reading time:** 14 minutes
**Tags:** AI Career, Resume Writing, ATS Optimization

Marcus, a senior product manager with twelve years of experience, sent me his resume after eight weeks of silence. "I've applied to forty roles I'm clearly qualified for," he wrote. "Not one callback. What am I doing wrong?" His resume looked great. Clean two-column layout, built in Canva, a tidy skills sidebar running down the left. It was also, to the software reading it first, almost unreadable.

That software is everywhere now. **97.8% of Fortune 500 companies** run an [applicant tracking system](https://www.jobscan.co/blog/fortune-500-use-applicant-tracking-systems/), 489 of the 500, according to Jobscan's 2025 ATS Usage Report. And **79% of organizations** have wired [AI or automation directly into that system](https://www.selectsoftwarereviews.com/blog/applicant-tracking-system-statistics), per SelectSoftwareReviews. A machine reads your resume before any human does. If it can't read you cleanly, the recruiter never sees you.

Here's the part that trips up smart people. Your resume can look polished and still come out the other side as scrambled fragments. The qualifications are there. The format hides them.

I'm Jessica Baker, a career strategist, and I've reviewed thousands of resumes that died this exact way. The good news: resume parsing is a predictable five-stage process. Document Conversion, Layout Reconstruction, Section Segmentation, Field Extraction, Normalization. This article walks all five, names what breaks at each, and ends with an 8-point checklist you can run against your own file in under twenty minutes.

Want a first clue right now? Open your resume PDF and try to highlight a single line of body text. What happens next tells you a lot.

## What Is Resume Parsing — and Where It Fits Inside an ATS

Resume parsing is the automated process that converts your resume file into structured, searchable data. The parser strips your name, contact details, work history (company, title, dates, duties), education (degree, institution, year), and your skill terms out of the document and drops them into labeled fields. That structured profile is what the scoring engine actually reads. Not your nicely formatted PDF.

So parsing is not the whole system. Think of the ATS as an airport, and parsing as the scanner at the entrance. If the scanner can't read what you brought through, the rest of the airport has no idea what to do with you. Everything downstream (keyword matching, recruiter search, ranking) runs on the data the parser pulled out. Garbage in, invisible candidate out.

At FoundRole we map that scanner to a five-step sequence: the **FoundRole 5-Stage Resume Parsing Pipeline**. Document Conversion gets text out of your file. Layout Reconstruction puts that text back in reading order. Section Segmentation labels your headings. Field Extraction pulls the values from each section. Normalization standardizes those values so they compare cleanly against other candidates. A resume can pass four stages and still die on the fifth.

The pipeline below shows all five stages in order, with the specific thing that breaks each one.

One scope note before the walk-through. This article is about parsing, meaning how your file gets *read*. How the ATS then *scores and ranks* you against a job is a separate problem, and we'll point you to it at the end.

## Stage 1: Document Conversion — How the Parser Gets Text Out of Your File

Stage 1 decides whether the parser can read any of your words at all, and it comes down to one thing: does your file have a real text layer? A `.docx` is the easiest format here. Its underlying XML tells the parser exactly where every block of text lives, so nothing gets guessed. A text-based PDF (one exported digitally from Word or Google Docs) is the close second, with the text layer intact and a single column coming through cleanly.

The dangerous format is the scanned or image-only PDF. There's no text layer, so the parser falls back to OCR, reading pixels like a camera. Accuracy drops on low-resolution scans, decorative fonts, and stylized headings. The usual offenders: **Canva exports, photographed printouts, and PDFs made by scanning a paper copy**. They look identical to a clean PDF on screen. To the parser, one is text and the other is a photo of text.

There's a quieter trap here too: ligatures. Some fonts merge "fi" or "fl" into a single glyph, and the parser reads "field" as "eld" or drops the letters entirely. You won't see it. The fix is boring and reliable: stick with Arial, Calibri, Helvetica, or Garamond.

So which format should you submit? **`.docx` is the safest universal bet**, a single-column text-based PDF is fine, and image PDFs should be avoided entirely. For the deeper breakdown of when each format wins, our guide to [ATS-safe resume formats](https://www.foundrole.com/blog/resume-formats-which-one-is-right-for-you) walks through the tradeoffs.

Run the 10-second test now: open your resume PDF, try to highlight a line of body text, and paste it into Notepad. If it pastes as real text, Stage 1 passes. If you can only grab a pixel block and the paste comes back empty or garbled, it's image-only. Convert it to a text-based PDF or save as `.docx` before your next application.

## Stage 2: Layout Reconstruction — Reading Order and the Multi-Column Problem

Stage 2 is where two-column resumes go to die, because the parser has to decide what order to read your words in. A single-column layout is unambiguous: the parser reads top to bottom, left to right, exactly the way you do.

A two-column or sidebar layout breaks that. A PDF doesn't store sentences. It stores text as spatial coordinates, fragments placed at x/y positions on the page. So the parser reads straight across the page at each vertical level, and your sidebar heading collides with the job line next to it. The skill term and the work duty get spliced into one line that means nothing.

This isn't rare. About **20% of real-world resumes use non-linear, multi-column layouts** that break the standard reading flow, according to [layout-aware parsing research on arXiv](https://arxiv.org/html/2510.09722v1) drawn from an Alibaba HR platform corpus. One in five resumes is quietly fighting its own parser. Invisible tables are the same problem in disguise: a borderless table in Word looks like clean columns to you, but the parser reads it cell by cell, interleaving left and right.

### Before/After plain-text parse example

Here's what Marcus's two-column resume produced when parsed:

> Skills Led product launch for Python a 200,000-user platform SQL Cut onboarding time Tableau by 40 percent

Now the same content in a single-column layout:

> **Work Experience: Product Manager, Acme Corp.**
> Led product launch for a 200,000-user platform. Cut onboarding time by 40 percent.
> **Skills:** Python, SQL, Tableau.

Same words. Same accomplishments. One is a strong PM. The other is gibberish the ATS will never match to a job.

The annotated comparison below shows why each version parses the way it does.

Paste your entire resume into a plain-text editor (Notepad on Windows, TextEdit on Mac) and read it top to bottom. If your sidebar skills show up mid-sentence inside your work history, your layout is scrambling the output. The fix is to go single-column.

## Stage 3: Section Segmentation — How the Parser Finds Your Headings

Stage 3 is where the parser figures out which part of your resume is which, and it does that by reading your section headings. Once Stage 2 has the text in order, the parser hunts for labels to mark off zones: contact, summary, work experience, education, skills. Get the labels right and every later stage knows where to look.

Some headings are recognized instantly. **"Work Experience", "Professional Experience", "Education", "Skills", "Technical Skills", "Certifications", and "Professional Summary"** all map cleanly. The classifier has seen them a million times.

Creative headings are where it falls apart. "My Journey", "Where I've Been", "Things I'm Good At", even the popular "Core Competencies", frequently come back as unknown. When that happens, your content gets dumped into the previous section or routed to an "unknown" bucket the ATS ignores. Your skills don't get misread. They get filed where no recruiter search will ever find them.

Here's the safe-vs-risky map. Pick from the left column:

| Recognized heading | Risky alternative to avoid |
|---|---|
| Work Experience / Professional Experience | My Journey / Where I've Been |
| Education | Academic Adventures |
| Skills / Technical Skills | Things I'm Good At |
| Certifications | Credentials & Extras |
| Summary / Professional Summary | About Me / My Story |

Consistent date and job-title formatting inside each entry reinforces the classification too, so even a borderline heading lands right when the content under it looks standard.

Worried this makes your resume sound robotic? You can keep your voice. Use the standard heading on its own line, then add a subtitle underneath. Put **"Work Experience"** on top, and below it, in smaller text, "Where I've built things that mattered." The parser reads the line it recognizes. The human reads the line with personality.

Audit your headings against this five-item list: Work Experience, Education, Skills, Certifications, Summary. Rename anything that isn't on it or a recognized synonym.

## Stage 4: Field Extraction — What the Parser Pulls From Each Section

Stage 4 is where the parser pulls the actual values out of each labeled zone, and where a handful of formatting choices silently delete information you swore was on the page. The parser uses two methods: rule-based pattern matching for predictable shapes (dates, emails, phone numbers), and NLP entity recognition for the messier stuff (job titles, company names, skill phrases). Well-tuned systems like Workday and iCIMS extract standard formats accurately, but push them off that path and they fall back to less reliable methods where things start dropping.

Five things reliably break extraction here:

1. **Graphics, icons, and logos.** The parser skips non-text entirely. That skill-rating bar graphic? Invisible.
2. **Text boxes and shape containers in Word.** They sit outside the main document flow, so the parser often never enters them.
3. **Word headers and footers.** Many parsers skip them completely. Put your name, email, and phone in the Word header and your application can submit with no contact info attached.
4. **Tables used for layout.** Same interleaving problem from Stage 2, now eating your individual fields.
5. **Slash-combined skill lists.** "Python/SQL/Tableau" parses as one unrecognized skill instead of three real ones.

Skills extraction has its own quirk: the parser matches your terms against an internal taxonomy. "Python", "project management", and "SQL" land cleanly because they're in the dictionary. "Creative problem-solving rockstar" matches nothing and gets dropped. As AI gets woven deeper into screening, clean Stage 4 output is what feeds the model. Our breakdown of [AI resume screening mechanics](https://www.foundrole.com/blog/ai-in-job-search-why-your-resume-gets-filtered-out-and-what-to-do-next) covers what happens after the data is extracted.

The reference table below lays out what extracts cleanly versus what breaks, side by side, so you can audit your file in about two minutes.

Copy just the top section of your resume into plain text. If your name, email, and phone appear, good. If they don't, they're probably trapped in a Word header the parser skipped. Move them into the document body.

## Stage 5: Normalization — How the Parser Standardizes What It Extracted

Stage 5 is the last step before a human is even in the picture. The parser takes the raw text it extracted and maps it to standardized, canonical values so it can compare you against every other candidate. "Sr. Software Eng.", "Senior Software Engineer", and "Software Eng - Senior" all have to resolve to one entry, or recruiter search simply doesn't work. Normalization is what makes your resume comparable, not just readable.

Three things tend to break here, and each has one simple rule.

**Dates.** "Jan 2023 – Mar 2024" and "01/2023 – 03/2024" both parse cleanly. Mix formats across entries, or use "Spring 2023" or bare numeric short years, and the parser can miscalculate your duration, sometimes inventing a two-year gap that never happened. The rule: pick one format and use it on every entry. "Month YYYY" works universally.

**Job titles.** Standard titles like "Marketing Manager" resolve cleanly against the taxonomy. Creative ones, such as "Growth Ninja" or "Chief Vibe Officer", may map to nothing, and a title that maps to nothing can't be matched to a posted role. The fix: keep your official title in your work history, then mirror standard job-title language in your summary where it fits naturally.

**Skills.** "MS Excel", "Microsoft Excel", and "Excel" should all collapse into one skill. A slash-combined entry like "Tableau/PowerBI", the same trap from Stage 4, sometimes parses as a single unrecognized entity. The practical rule: use the job description's own wording wherever you honestly can.

Once normalization finishes, your profile is done. It gets handed to the scoring engine, and the parsing job is over.

Scan your work history for mixed date formats and standardize to one. "Jan 2023" (Month YYYY) is the most reliably parsed format across the major systems.

## What Breaks Resume Parsers: The Most Common Failure Modes

Here's every failure mode from the five stages in one place, grouped by where it happens. The most common resume parsing failures, by stage:

1. **Stage 1:** Scanned or image-only PDF (no text layer, OCR guesses).
2. **Stage 2:** Two-column or sidebar layout, plus invisible layout tables.
3. **Stage 3:** Creative section headings the classifier can't recognize.
4. **Stage 4:** Contact info trapped in a Word header, text boxes, graphics, slash-combined skills.
5. **Stage 5:** Mixed date formats and creative job titles that map to nothing.

The stakes are real, and recruiters know it. **88% of employers believe they're losing highly qualified candidates** to ATS formatting failures, according to [SelectSoftwareReviews](https://www.selectsoftwarereviews.com/blog/applicant-tracking-system-statistics). Pair that with **97.8% of Fortune 500 companies** running an [ATS](https://www.jobscan.co/blog/fortune-500-use-applicant-tracking-systems/) and the **20% of resumes** using [multi-column layouts](https://arxiv.org/html/2510.09722v1) that break parsing, and you see the size of the leak. The candidates getting dropped aren't unqualified. Their formatting is.

The numbers below are worth screenshotting before you open your own resume.

Now the good news. Every item on that list is fixable, most in under thirty minutes. You don't need a new degree or a better job history. **The problem is the formatting, not the qualifications.** If your resume keeps disappearing into the void, this list is the first place to look before you rewrite a single bullet point.

## Make Your Resume Parser-Ready: 8-Point Checklist

Here's a self-contained audit you can run against your own file before your next application. No rewriting your accomplishments, only fixing formatting. Each point maps to one of the five stages, so a clean run means your resume should clear all of them.

**8-Point Parser-Readiness Checklist:**

1. **Text-layer PDF or `.docx`** (run the Stage 1 copy-paste test).
2. **Single-column layout** (paste to plain text, reading order holds).
3. **Standard section headings** (Work Experience, Education, Skills, Certifications, Summary).
4. **Contact info in the document body** (not only in a Word header or footer).
5. **No graphics, icons, or logos** inside the document.
6. **No text boxes or shape containers.**
7. **Consistent date format** ("Jan 2023" on every entry).
8. **Skills comma-separated, not slash-combined** ("Python, SQL" not "Python/SQL").

Picture the before and after. Marcus's original: a Canva two-column image PDF, creative headings, contact info stuck in the header, slash-combined skills, mixed date formats. The fixed version: a single-column `.docx`, standard headings, contact in the body, comma-separated skills, "Month YYYY" dates throughout. **The "after" version isn't fancier. It's just legible to the machine.**

The interactive checklist below lets you tick off each item as you go and copy the whole list for later.

Once the formatting is clean, the words still have to earn the interview. Our [expert resume writing tips](https://www.foundrole.com/blog/resume-writing-tips) cover the bullet-level rewrites that turn a parseable resume into a compelling one. When you are ready to put a fixed resume to work, you can [search ATS-listed job openings](https://www.foundrole.com/jobs?utm_source=blog&utm_medium=article&utm_campaign=what-is-resume-parsing&utm_content=cta-conclusion) and apply knowing the parser can read you.

Run the 8-point checklist against your resume before your next application. If any check fails, fix that one item before you hit submit.

## What Happens After Parsing: The Scoring Step

Passing parsing cleanly is necessary, but it isn't the finish line. It just gets you into the race. Once your profile is structured, it goes to a scoring engine that ranks you against the job's requirements: keyword match, years of experience, required skills, and sometimes an AI fit score on top.

Keep the two jobs separate in your head. **Parsing is being read. Scoring is being ranked.** Clean parsing without the right keywords gets you a low match score. The right keywords trapped inside a broken layout never get parsed at all, so the scoring engine never sees them. That's why the order matters: fix parsing first, then keyword alignment. Doing it the other way around is polishing words the machine can't read.

One caveat about scale. Roughly **70% of large companies use an ATS versus about 20% of small and mid-sized businesses**, per [SelectSoftwareReviews](https://www.selectsoftwarereviews.com/blog/applicant-tracking-system-statistics). If you're applying to a smaller employer, a human may read your resume directly. Parser-friendly formatting costs nothing there either, since a clean single-column resume is easier for a person to scan too.

This article stops at parsing mechanics. Keyword alignment, structuring your bullets, and ranking well once you've parsed cleanly is the next problem. Our guide to [ATS optimization strategies for 2026](https://www.foundrole.com/blog/ats-optimization-in-2026-how-to-beat-the-ai-resume-screeners) picks up exactly where this one ends.

The tool below routes you to your next step based on one question: have you confirmed your resume parses cleanly?

Once your resume parses cleanly, move to keyword alignment. Run it against a specific job description through an ATS scanner and see where your match score lands.

## Your Resume Has to Pass a Machine Before It Reaches a Human

Five stages, one sentence: your resume gets converted, reconstructed, segmented, extracted, and normalized, all before a recruiter sees a single bullet point. Miss any one of them and the rest doesn't matter.

Remember Marcus? He didn't have an experience problem. He had a two-column Canva problem. One weekend, one `.docx` rebuild (single column, standard headings, contact in the body, consistent dates), and the callbacks started inside two weeks. Same career. Same accomplishments. A resume the machine could finally read.

You don't have to fix everything to clear most of the failure surface. Four root fixes do the heavy lifting: **single-column layout, standard section headings, a text-based PDF or `.docx` file, and one consistent date format.** Handle those four and you've cleared the majority of what kills resumes at the parsing stage.

When your resume can be read, put it to work. You can [browse ATS-indexed job listings](https://www.foundrole.com/jobs?utm_source=blog&utm_medium=article&utm_campaign=what-is-resume-parsing&utm_content=cta-conclusion) on FoundRole, and once you start applying, [track your job applications](https://www.foundrole.com/job-tracker?utm_source=blog&utm_medium=article&utm_campaign=what-is-resume-parsing&utm_content=cta-tracker) in one place so nothing slips after it clears the ATS.

The machine reads first. Your job is to make sure the human gets a chance to read next.
## Latest Articles

- [ATS Resume: How to Get Past the Bot and the Recruiter](https://www.foundrole.com/blog/how-to-get-your-resume-past-a-recruiter)
- [ATS Optimization in 2026: Beat AI Resume Screeners](https://www.foundrole.com/blog/ats-optimization-in-2026-how-to-beat-the-ai-resume-screeners)
- [AI in Job Search: Why Your Resume Gets Filtered Out](https://www.foundrole.com/blog/ai-in-job-search-why-your-resume-gets-filtered-out-and-what-to-do-next)
- [How Recruiters Read a Resume: What They See in 7 Seconds](https://www.foundrole.com/blog/how-recruiters-read-a-resume-what-they-look-for-in-6-seconds)
- [Resume Writing Tips: 25 Expert Tips to Stand Out in 2026](https://www.foundrole.com/blog/resume-writing-tips)


## Frequently Asked Questions

### Does every job application go through resume parsing?

Not every application. Roughly 70% of large companies use an ATS, while only about 20% of small and mid-sized businesses do, per SelectSoftwareReviews. If a job board or company career page asks you to upload your resume, assume parsing is happening, even at smaller firms running modern HR tools. Resumes emailed directly to a very small employer are more likely read by a human first.
### Is a PDF or Word file better for ATS parsing?

Both work well when formatted correctly. A text-based PDF (exported digitally, not scanned) and a .docx file both parse cleanly. The .docx has a slight edge because its underlying XML tells the parser exactly where each block of text lives, so nothing gets guessed; a single-column text-based PDF is a close second. The format to avoid is a scanned or image-only PDF, which forces OCR and drops accuracy. Canva exports often produce image-only PDFs.
### Can a well-formatted resume still parse incorrectly?

Yes. Well-tuned systems like Workday and iCIMS extract standard formats accurately, but push a resume off that path and the parser falls back to less reliable methods where fields start dropping. Edge cases such as uncommon certifications, abbreviations, or slash-combined skills are where divergence shows up. The plain-text paste test matters here: it exposes what any parser is likely to hit in your file, regardless of which ATS the employer runs.
### What happens to information the parser can't read?

It gets dropped silently. The ATS profile simply has an empty field, and recruiter search or the scoring engine treats it as absent. Text inside graphics, image-only PDFs, Word headers, and text boxes is the most commonly lost content. You will not see an error message; the resume still submits successfully. The failure stays invisible until your profile fails to surface in a skills search you should have matched.
### Does resume parsing use AI in 2026?

Yes. 79% of organizations have integrated AI or automation directly into their ATS, according to SelectSoftwareReviews. AI mainly augments the NLP field-extraction and normalization stages, improving accuracy on non-standard titles and abbreviations. The catch: AI-augmented parsing is trained on well-structured resumes, so a poorly formatted file gets a lower-confidence extraction and the clean Stage 4 output is what feeds the model.
### Will a creative job title hurt my chances with an ATS?

It can. The parser normalizes job titles against an internal taxonomy, so a title like 'Growth Ninja' or 'Chief Vibe Officer' may map to nothing, and a title that maps to nothing can't be matched to a posted role. The fix: keep your official title as-is in your work history, then mirror standard job-title language, like 'Marketing Manager', in your summary where it fits naturally.
### How do I know if my resume is parsing correctly before I apply?

Run the plain-text test: paste your entire resume into a plain-text editor and read it top to bottom. If the content is in the right order with no interleaving (your sidebar skills landing mid-sentence inside your work history), your structure is likely parsing cleanly. Then run the copy-paste test on your PDF: if text highlights and pastes, it has a real text layer; if you can only grab a pixel block, it's image-only and needs converting.
---

[Browse all articles](https://www.foundrole.com/blog)