How ATS Resume Parsing Actually Works: Why Resumes Get Filtered (And What Doesn't)

Direct Answer: An Applicant Tracking System (ATS) is an enterprise recruitment database and workflow manager. It does not possess artificial consciousness, and it does not unilaterally reject 75% of resumes with an automated "reject" button. Instead, modern ATS platforms (such as Greenhouse, Workday, Lever, and iCIMS) parse uploaded documents into plain-text database fields, execute recruiter-defined knockout filters, and enable recruiters to filter and search candidates using keyword queries and attribute facets.

Resumes fail in an ATS primarily for two concrete reasons: text extraction failures (where complex layouts or corrupted fonts prevent the parser from extracting readable text) and keyword mismatch (where a qualified candidate uses non-standard phrasing that recruiters do not search for).

---

1. The Myth of the "75% Robot Rejection Rate"

One of the most persistent claims in career advice is that "75% of resumes are automatically rejected by ATS algorithms before a human ever sees them."

This claim originated as an unverified marketing soundbite from early keyword-scanning software vendors in the late 2000s. In reality, modern recruitment workflows operate differently:

• Knockout questions trigger automated filters: The only automated rejections configured in modern systems like Greenhouse or Workday are binary knockout questions answered during the application form (e.g., "Are you legally authorized to work in this country?", "Do you hold an active CPA license?", or "Do you meet the minimum 5 years of required medical residency?").
• Recruiters review candidates in lists: When you submit a resume, the software parses your document into a candidate profile. In platforms like Greenhouse and Lever, recruiters review applications in chronological or reverse-chronological order.
• Filtering happens via recruiter search: In large corporate requisitions receiving 500+ applicants, recruiters rarely read all 500. Instead, they apply filters: filtering by location, required degrees, or running Boolean keyword searches (such as "distributed systems" AND ("Go" OR "Rust")). If your resume lacks the searched term, you simply remain unviewed in the candidate pool.

Understanding that the ATS is an indexed search engine rather than an adversarial decision-maker transforms how you should write and structure your resume.

---

2. The Four Stages of ATS Processing

When you upload a PDF or DOCX file to an employer's job portal, the document passes through a standard four-stage ingestion pipeline:

Upload (PDF/DOCX)
       │
       ▼
1. Text & Layout Extraction  ──► (Converts visual document to plain text stream)
       │
       ▼
2. Semantic Tokenization     ──► (Maps sections: Contact, Experience, Skills, Education)
       │
       ▼
3. Entity Normalization      ──► (Standardizes job titles, company names, date ranges)
       │
       ▼
4. Database Indexing         ──► (Populates recruiter search fields & keyword facets)

Stage 1: Text & Layout Extraction

The parser's primary job is converting binary document data into clean UTF-8 text strings.

• Plain Text Reading Order: Parsers read documents horizontally, from left to right, top to bottom.
• The Multi-Column Trap: When resumes use multi-column templates (such as two-column designs built in Canva), parsers often read straight across the page. Line 1 of Column A gets concatenated with Line 1 of Column B, resulting in garbled sentences that make zero sense when parsed.
• Text Layers in PDFs: An ATS cannot read rasterized images or vector graphics. If your resume is saved as an image PDF without an embedded text layer, the extraction step yields empty text.

Stage 2: Section & Semantic Tokenization

The parser scans the extracted text stream to identify standard section headings. It looks for established anchors:

• Work Experience / Professional Experience / Employment History
• Education / Academic Background
• Skills / Technical Competencies
• Contact Information / Header

If you use non-standard headings like "Where I've Made Magic" or "My Journey", the tokenizer fails to classify the section. As a result, 5 years of relevant engineering experience may be categorized as miscellaneous notes rather than verifiable employment history.

Stage 3: Entity Normalization & Date Calculation

Once sections are bounded, the parser extracts structured entities:

• Dates & Tenure: The system identifies dates (MM/YYYY or Month YYYY) to calculate total years of experience. Inconsistent formats (e.g., listing "Spring 2022" or omitting dates) cause the tenure counter to report 0 years.
• Job Titles: Parsers map custom titles to standardized taxonomy dictionaries (such as O*NET or SOC codes). A title like "Chief Problem Solver" fails to map, whereas "Senior Operations Analyst" matches recruiter filters immediately.
• Company & Location: Normalized against corporate entity directories to support geographic filtering.

Stage 4: Database Search Indexing

All normalized tokens and raw text are committed to an inverted search index (similar to Elasticsearch). When a hiring manager tells a recruiter, "Find candidates with Kubernetes and Terraform experience", the recruiter types those terms into the ATS search bar. The system ranks candidates whose indexed profiles contain those specific tokens.

---

3. Which Resume Elements Cause Parsing Failures?

Based on empirical testing across enterprise parsers (including Sovren, Textkernel, and Workday Parser), the following layout choices consistently trigger data loss:

You can test whether your current resume extracts cleanly by running a baseline diagnostic on our Free ATS Resume Checker. It inspects text extractability, identifies section boundaries, and flags formatting anomalies without requiring an account.

---

4. How Keywords Actually Function in Recruitment

A common misconception is that applicant tracking systems look for a single "secret keyword" or that repeating a word 20 times increases rank.

In enterprise environments:

Contextual Frequency Matters

Modern search engines evaluate keyword relevance using term weighting:

• A skill mentioned once in a skills list proves familiarity.
• A skill mentioned within an experience bullet with an associated outcome demonstrates applied competency.
• Example: Compare "Python" in a list versus "Engineered automated ETL pipelines in Python, processing 2M daily records". The latter provides context that recruiter searches evaluate favorably during human screening.

Exact Matches vs. Synonyms

While some enterprise parsers have semantic synonym expansion (e.g., knowing that "GCP" relates to "Google Cloud Platform"), recruiters frequently search for exact acronyms or exact phrases found in the job specification.

If a job posting asks for "Search Engine Optimization (SEO)", including both the spelled-out term and the standard acronym ensures your profile matches whether the recruiter searches for the full name or the shorthand.

---

5. The Truth About File Formats: PDF vs. DOCX

Should you submit a PDF or a Word document (.docx)?

• DOCX (Microsoft Word): Historically the safest format for legacy parsers because the underlying document is structured XML. However, DOCX files can render inconsistently across different operating systems when viewed by human recruiters.
• PDF (Portable Document Format): The preferred format for human visual presentation because layout, typography, and margins are preserved across devices.
• The Rule for PDFs: A PDF is completely safe for ATS parsing provided it is generated digitally (via Word, Google Docs, or LaTeX) and contains a selectable text layer. Never scan a printed resume to PDF, as scanned images require Optical Character Recognition (OCR), which introduces high error rates in technical terms and acronyms.

---

6. How to Verify Your Resume Before Applying

Follow this 5-minute pre-submission audit before submitting your application:

1. The Plain Text Copy-Paste Test: Open your PDF or DOCX file, press Ctrl+A (or Cmd+A) to select all, copy the text, and paste it into a plain text editor (like Notepad). Read through the text:
   • Does the reading order make chronological sense?
   • Did any section disappear?
   • Did two columns merge into scrambled sentences?
2. Section Header Check: Confirm your major sections use standard nomenclature: Experience, Education, Skills, Projects.
3. Contact Details Check: Confirm your email address, phone number, and location (City, State/Country) appear in the main body text, not tucked into an isolated header box.
4. Keyword Relevance Check: Compare your experience section against the target job posting. Ensure required hard skills appear naturally in the context of your achievements.
5. Run a Deterministic Diagnostic: Use our Free ATS Resume Checker to test your document against 20+ parsing standards. For details on how we measure parsing compliance, read our complete methodology in How We Score ATS.