How ATS parses your resume: what happens the second you hit submit
Most job seekers think ATS "reads" their resume the way a human does — top to bottom, picking up the highlights. It doesn't. ATS parsing is a structured data extraction process, and anything that disrupts that process means your information never makes it into the system correctly.
Understanding exactly how ATS parses your resume is the most useful thing you can know about the job application process. Not because it lets you "trick" the system — but because it explains why specific formatting choices cause real problems, and what to do instead.
Step 1: File ingestion — what format you submit matters immediately
The moment you upload your resume, the ATS performs its first operation: converting your file into raw text it can process. How cleanly this happens depends almost entirely on the file type.
PDF (text-based): The safest option for most modern ATS platforms. A text-based PDF preserves your content in a readable format that parsers can extract directly. The key word is text-based — a PDF created by saving a Word document is text-based. A scanned PDF is an image.
DOCX (Word): Also widely supported. Most ATS platforms handle DOCX well. The risk is formatting — complex tables, text boxes, and multi-column layouts in Word documents frequently break parsing even when they look fine on screen.
Scanned PDF or image-based PDF: These require OCR (Optical Character Recognition) to convert the image into text. OCR accuracy is imperfect, and older ATS platforms often skip OCR entirely, meaning a scanned resume may be ingested as a blank document.
Rule: Submit a text-based PDF unless the job posting specifically requests a Word document. Never submit a JPG, PNG, or scanned copy of a physical resume.
Step 2: Text extraction — where most formatting breaks happen
Once the file is ingested, the ATS extracts all text from the document. This is where layout-heavy resumes fall apart.
What the parser sees vs. what you see:
What you designed:
Skills Experience
• Python Software Engineer, Acme Corp
• SQL 2021–2024
• Tableau
What the parser extracts from a two-column layout:
Skills Experience • Python Software Engineer, Acme Corp • SQL 2021–2024 • Tableau
The parser reads left to right, row by row, across the entire page width. Columns get merged into a single stream of scrambled text. The ATS then tries to assign this jumbled output to the correct resume fields — and frequently fails.
Other formatting elements that break text extraction:
- Tables: Cell contents are concatenated in unpredictable order
- Text boxes: Often extracted separately, out of order, or skipped entirely
- Headers and footers: Most parsers ignore the header/footer regions of a PDF — meaning your name and contact information placed there may never be read
- Graphics and icons: Ignored completely. A phone icon next to your number is fine. Your phone number inside a graphic object is invisible
- Fancy bullet characters: Non-standard Unicode bullets (❖ ✦ ▸) occasionally cause character encoding errors that corrupt surrounding text
Step 3: Section labelling — how the ATS knows what's what
After extracting raw text, the ATS needs to map each piece of information to a structured field: name, email, phone, work history, education, skills, and so on.
This section-labelling step relies on recognising your section headings. The parser looks for known heading patterns and assigns everything that follows to the corresponding field — until it encounters the next recognised heading.
Headings that parse reliably:
- Work Experience / Professional Experience / Experience
- Education
- Skills / Technical Skills / Core Competencies
- Certifications / Licences
- Summary / Professional Summary / Profile
Headings that confuse parsers:
- "Where I've Been" (instead of Experience)
- "My Toolkit" (instead of Skills)
- "Academic Background" (instead of Education)
- "What I Bring" (instead of Summary)
When the parser doesn't recognise a heading, it either ignores the section entirely or dumps its contents into a catch-all "other" field that recruiters rarely search. Your skills end up invisible.
Use standard heading labels. There is no SEO benefit to being creative with resume section names — only parsing risk.
Step 4: Field extraction — pulling your actual data
Within each labelled section, the ATS extracts specific data points using pattern matching and NLP (Natural Language Processing):
Contact information:
- Name: Usually extracted from the largest text element near the top
- Email: Matched by regex pattern (text@domain.extension)
- Phone: Matched by number pattern
- LinkedIn / URL: Matched by URL pattern
- Location: City and state extracted by geographic name recognition
Work history (per role):
- Job title
- Employer name
- Start date and end date
- Duration (calculated)
- Responsibilities/achievements (full text stored for keyword search)
Education (per qualification):
- Degree name
- Institution
- Graduation year
- GPA (if present)
Skills:
- Extracted as a list from the Skills section
- Also identified from work experience bullets (contextual skill extraction)
Where this goes wrong: Dates in unusual formats (e.g. "Summer 2022" or "Q3 2021") often fail to parse into start/end date fields, which breaks tenure calculations. Use Month YYYY format (June 2022) or MM/YYYY.
Step 5: Keyword indexing and ranking
Once your resume data is structured in the ATS database, the system indexes it for search and scoring.
When a recruiter searches for candidates — or when the ATS auto-ranks applicants against a job posting — it runs queries against this indexed data. The match depends on:
Exact keyword matches: "Python" matches "Python." "Python programming" may not match "Python" depending on the system.
Semantic/NLP matching (modern ATS): More sophisticated platforms use NLP to understand that "machine learning" and "ML" refer to the same thing, or that a "software engineer" and "developer" share relevant skills. Older systems rely purely on exact matching.
Weighted fields: Most ATS platforms weight job title and skills section matches more heavily than matches found only in the body of a bullet point. A keyword appearing in your Skills section and your experience bullets scores higher than one appearing only once.
Recency: Recent roles are weighted more heavily. A keyword that appears in your most recent job scores higher than the same keyword from a role 8 years ago.
What this means for how you write your resume
Knowing how parsing works changes specific decisions:
| Decision | Why it matters |
|---|---|
| Single-column layout | Prevents text extraction from scrambling columns |
| Standard section headings | Ensures sections are labelled correctly |
| Name and contact in the body, not the header | Header/footer regions are often skipped |
| Text-based PDF | Avoids OCR errors |
| Month YYYY date format | Enables accurate tenure calculation |
| Standard bullet characters (•) | Avoids encoding errors |
| Keywords in both Skills and experience | Increases weighted match score |
| Full terms + abbreviations ("Python" not just "PY") | Covers both exact and semantic matching |
Which ATS platforms parse best and worst
Not all ATS systems parse equally well. Platforms that are known to handle complex formatting better include Greenhouse and Lever (both built more recently). Platforms with older parsing engines — Taleo, some legacy iCIMS configurations — are stricter and less forgiving of non-standard formatting.
The practical implication: unless you know which ATS the company uses, write for the strictest parser. A clean, single-column, standard-heading resume passes every system. A designed resume may pass modern ATS but fail older ones.
Frequently asked questions about ATS parsing
Does ATS read the entire resume or just scan it? The entire text is extracted and indexed. However, keyword weighting means some sections (title, skills) carry more influence over ranking than others (responsibilities from 10 years ago).
Can ATS read graphics or icons? No. Text within images, icon-based contact information, and graphical skill bars are invisible to parsers. All meaningful content must be in actual text.
Does the order of sections matter? For ranking purposes, no — content is indexed regardless of where it appears. For readability by the human reviewer who follows, yes. Put your most relevant information early.
What file name should I use for my resume? Use FirstnameLastname-Resume.pdf. Some ATS systems log the file name and display it to recruiters. "FinalFinalResume_v3.pdf" looks careless.
Does ATS parse LinkedIn profiles differently from resumes? LinkedIn profiles submitted via "Easy Apply" are parsed through LinkedIn's own system, which then exports structured data to the ATS. The same data quality rules apply, but LinkedIn's own parser handles the extraction.
