How an ATS actually reads your resume
An ATS parses your PDF or DOCX into a structured record: name, contact, work history, skills, education, certifications. It does this with a combination of layout heuristics and named-entity recognition. The cleaner and more predictable your structure, the higher the parser's confidence. Confidence drives ranking.
The parser builds a skill graph from your bullets and matches it against the role's required skill graph. Then it scores the overlap. The recruiter sees the top N matches, sorted by overlap score. Most ATS interfaces do not show resumes that fall below the fold unless the recruiter specifically widens the search.
What the ATS quietly rejects
- Two-column layouts — the parser reads left-to-right and merges columns badly.
- Headers and footers — most parsers ignore them entirely.
- Text inside images or graphics — invisible to the parser.
- Skills only listed in a 'Skills' section without supporting bullets.
- Job titles that don't match common titles for the role you're targeting.
- Date formats the parser doesn't recognize ('Summer 2023', 'present').
Run the check yourself
Open your resume in a plain-text reader (most operating systems have one built in). If the order of text is not the order a recruiter would read it in, the ATS is reading it wrong too. That alone explains roughly 30% of "I'm not hearing back" cases.
Next, run the Behind Offer analyzer on the homepage with a real job description. The fit probability that comes back is the same signal the ATS is computing internally — overlap of evidence against requirements, with confidence calibrated to how clear both sides are. If the probability is low, the gap report tells you which specific bullets to rewrite first.
The system is probabilistic, not deterministic. A 0.45 probability means "consider applying, with caveats" — not "you will not get this role". Use it as guidance for where to invest your application time.