How ATS Resume Scanners Work (Parsing, Keyword Matching and Scoring)

Short answer: A resume scanner does three things in order. It extracts text from your file, structures that text into fields (name, employers, dates, titles, skills), and then scores what it found — usually keyword coverage against a role, plus formatting and completeness checks. Every "ATS score" you've seen is a score of that third step. If step one or two goes wrong, the score is meaningless, because the tool is grading a mangled copy of your resume.

Below is what actually happens at each stage, and how to use that knowledge instead of guessing.

Stage 1 — Text extraction

Your PDF or DOCX is not text. It's a set of drawing instructions that happen to look like text. A parser has to reconstruct the reading order.

File type	How it's read	Where it breaks
PDF with a text layer	Glyphs plus coordinates are re-assembled into lines	Two-column layouts interleave the columns into nonsense
Scanned or image PDF	Nothing extractable without OCR	Usually returns an empty or near-empty resume
DOCX	XML with real paragraph structure — the cleanest input	Text boxes and SmartArt are often skipped entirely
Pages / ODT / RTF	Converted first, with losses	Bullets, dates and special characters get scrambled

Two things matter more than anything else at this stage: your file must contain a real text layer, and your layout must be single-column. A visually beautiful two-column CV can lose a third of its content before any scoring begins — and neither you nor the recruiter sees it happen.

Test it yourself: open your PDF, select all, copy, and paste into a plain text editor. What you see is roughly what the parser sees. If the order is jumbled or content is missing, fix the layout before you worry about keywords.

Stage 2 — Structuring the text into fields

Extracted text is still a wall of characters. The parser now has to decide which parts are your job titles, which are employers, and which are dates.

It does this with a mix of:

• Section header matching — it looks for conventional labels: Experience, Work Experience, Education, Skills, Certifications. Creative headings like "My Journey" or "What I'm Good At" often map to nothing, and the content underneath is discarded or dumped into a generic bucket.
• Date pattern recognition — Jan 2021 – Mar 2024 is parseable. 2021-24, '21 to '24, or a date sitting in a sidebar column often is not. Broken dates knock you out of "5+ years of experience" filters even when you have eight.
• Positional heuristics — the line above a date range is probably a title or employer. This is why an inconsistent job block format hurts: change the order between roles and the parser assigns the wrong field.
• Contact extraction — regex for email and phone. Contact details placed in the document header or footer are the single most common cause of a resume with no reachable contact info, because several parsers ignore those regions.

At the end of this stage the system holds a structured record. That record — not your PDF — is what gets searched, filtered and scored.

Stage 3 — Keyword matching

Recruiters rarely browse. They search the structured records with filters and keywords taken from the job description.

Matching is usually some combination of:

• Exact and stemmed matching — "manage", "managed", "managing" collapse to one root. "Kubernetes" and "K8s" usually do not, unless someone configured a synonym list.
• Acronym pairs — write both forms once: Search Engine Optimization (SEO). This costs you nothing and doubles your match surface.
• Section weighting — a skill named inside a real experience bullet carries more weight in most scoring models than the same word listed in a skills block, because it is evidenced.
• Frequency with diminishing returns — the second mention adds a little, the tenth adds nothing. Keyword stuffing is detectable and penalised by modern scoring, and it reads badly to the human who eventually opens the file.

The honest limitation: no scanner knows the exact keyword list a specific employer configured. Any tool claiming a guaranteed "pass rate" for a named ATS is guessing. What a good scanner can do is measure coverage against the job description you paste in, which is a real and useful proxy.

Stage 4 — Scoring

A score is a model, not a verdict. Reasonable models weight roughly four things:

Pillar	What it measures	Typical weight
Parsing and structure	Can the file be read and mapped to fields at all	Highest — it gates everything else
Keyword and role fit	Coverage of the terminology for the target role	High when a job description is supplied
Impact	Quantified outcomes rather than duty lists	Medium
Clarity	Bullet length, tense consistency, date hygiene	Medium

Two rules make scores useful instead of anxiety-inducing:

1. A critical parsing issue should cap the score. A resume with unreadable structure cannot be an 85 because the prose is elegant. DexterCV's scanner caps resumes carrying unresolved critical or warning issues, so a high score means genuinely clean, not merely wordy.
2. Compare against yourself, not others. The value is the delta between your first scan and your scan after fixing findings — not a leaderboard.

What a resume scanner cannot tell you

• Whether a specific employer's ATS will reject you. Configuration varies per company.
• Whether you're qualified. Scoring reads structure and language, not capability.
• Whether a human will like your resume. That's the interview-rate signal, and it needs a real recruiter's read.

Treat the score as a pre-flight check: it catches the mechanical failures that silently cost you interviews, then hands the judgement back to you.

A practical 15-minute pass

1. Copy-paste your PDF into a plain text editor and confirm nothing is missing or out of order.
2. Move contact details out of the page header into the body.
3. Rename every section to a conventional label.
4. Make every date Mon YYYY – Mon YYYY, consistently.
5. Spell out each important acronym once alongside its short form.
6. Add a number to at least half your bullets.
7. Re-scan and compare the two reports.

Try it on your own resume

Run a free scan — upload a PDF or DOCX and you'll get the parsed structure, the findings ranked by severity, and a score broken into the pillars above. If you want coverage against a specific role, paste the job description into the Resume Customizer instead, which scores match against that exact posting.

Related reading: what an ATS resume checker actually checks, single-column vs two-column resumes, and ATS keywords by industry.