What an ATS Actually Sees When It Reads Your Resume
Your Resume Is Not What You Think It Is
You spent hours perfecting your resume's typography, alignment, and visual hierarchy. But the first entity to evaluate that document will not be a human — it will be a parser that reduces your carefully designed layout to a flat data structure.
The 5-Stage ATS Parsing Pipeline
Stage 1: File Upload and Format Detection
The ATS identifies the file type by reading magic bytes. A PDF starts with %PDF-1., a DOCX is a ZIP archive starting with PK. Renaming a .pages file to .docx will fail at this stage.
Stage 2: Text Extraction
For DOCX files, the parser reads the underlying XML. For PDFs, it uses a library like Apache PDFBox. Text-based PDFs parse reliably. Image-based PDFs require OCR with 5-15% error rate. Tagged PDFs provide structural hints but are rare. At this stage, all visual formatting is discarded.
Stage 3: Section Detection
The parser segments raw text into semantic sections using regex and ML classifiers. A line that reads "PROFESSIONAL EXPERIENCE" is a strong signal. But "Where I've Added Value" gives the parser nothing to work with.
Stage 4: Field Mapping (Entity Extraction)
Within each section, the parser extracts structured entities. Here's what it tries to produce from your Experience section:
{ "experience": [{ "company": "Stripe", "title": "Senior Software Engineer", "start_date": "2022-03", "end_date": "present", "bullets": ["Led migration of payment processing pipeline..."] }] }Now here's what happens with a two-column template — the parser, reading left-to-right across the full page width, produces garbled output where every field is wrong.
Stage 5: Keyword Scoring and Ranking
After extraction, the ATS compares structured data against the job requisition. At high-volume companies, only the top 20-30% of scored applications reach human review.
What Gets Lost
- All visual design: colors, fonts, sizes, alignment, margins
- Images and graphics: photos, logos, skill bars, charts, icons
- Headers and footers content
- Hyperlink URLs (display text preserved, URL often discarded)
- Non-standard characters and emojis
How ResumeAgentics Helps
ResumeAgentics shows you both views simultaneously: your designed resume on one side, and the raw parsed output on the other. Every field is color-coded — green for correct extraction, red for failures — and each failure includes a one-click fix.