There's a myth that an applicant tracking system gives your CV a secret score and an algorithm decides your fate. The reality is more mundane and more useful to understand. Most ATS "scoring" is a keyword and criteria match against the job posting, and a recruiter still does the real deciding. Once you know what the score measures, you stop optimizing for ghosts.
What the score is actually counting
When an ATS ranks candidates, it's usually comparing your CV against the requirements the recruiter set up for that job. The main inputs are simple:
- Keyword match. Do the skills, tools, and titles from the posting appear in your CV? This is the biggest lever, and it's literal string matching, not meaning. "JavaScript" and "JS" can count as different terms.
- Knockout questions. Things like work authorization, years of experience, or a required certification. Miss a hard one and you can be filtered regardless of everything else.
- Title and experience signals. Whether your past titles and stated years line up with what the role expects.
That's most of it. The system is not reading your CV for quality, judging your prose, or detecting your potential. It's checking boxes a human told it to check.
What the score does not measure
This is where people waste effort. The ATS score generally does not assess:
- How well-written your CV is. Beautiful prose earns no points from the parser.
- Whether your experience is genuinely impressive in context. Nuance is invisible to a string match.
- Design and visual polish, beyond whether the file parses cleanly.
So two things follow. First, getting the keywords right matters more than agonizing over wording the machine can't read. Second, parse-ability is a precondition: if the system can't read your file because of columns, text in images, or a weird layout, you score near zero no matter how qualified you are. A clean, single-column, standard-heading CV isn't about looking plain, it's about being readable.
How to score well without keyword stuffing
The naive response to keyword matching is to cram every term from the posting into white text or a skills wall. Don't. Recruiters search these systems and a stuffed CV gets caught and looks desperate when a human opens it. The durable approach:
- Pull the genuine must-have skills and tools from the posting, the ones that repeat or sit in the requirements.
- Use the exact phrasing for hard requirements, since the match is literal. If they say "Kubernetes," write Kubernetes, not "container orchestration."
- Place those terms inside real bullets about work you did, not in a detached keyword list. A skill backed by a result beats the same word floating alone.
- Mirror the job title somewhere sensible if your real title differs but the role is the same.
That gets you a high match honestly, which is the only kind that survives the recruiter reading it afterward.
The score gets you seen, not hired
Keep the score in proportion. A strong match moves you into the pile a human reviews. It does not get you the job, and a perfect keyword match with a thin CV still loses to a real one. So spend the first effort on parse-ability and honest keyword alignment, then spend the rest on a CV a person wants to read.
If you're rebuilding the CV to pass cleanly, starting from a structured source helps you avoid the layout traps that wreck parsing. Some people generate the draft from their LinkedIn profile with a tool like Postulit, which gives a single-column, standard-section base that's easy for an ATS to read, then tailor the keywords per job from there.
Take one target posting, list the eight terms that repeat or sit in the requirements, and check each one appears in a real bullet on your CV. That's most of the score, done honestly.