How ATS Works: Keywords, Parsing & Ranking

The pipeline: from file to shortlist

When you hit “submit”, the ATS takes your file through a series of automated steps before a human ever sees it. Understanding those steps — not as a “black box”, but as a pipeline with known mechanics — is what separates educated resume writing from superstition.

The typical pipeline is: (1) extract the text from your file, (2) normalize it into fields, (3) index it for search, (4) match it against the job’s requirements, and (5) rank it for a recruiter to review.

Extraction & normalization

Extraction is where resumes die silently. The parser reads your PDF or Word file and pulls text out of the layout. Anything the layout obfuscates gets lost: text inside tables, text boxes, images, headers and footers, decorative graphics. That is why single-column, text-based layouts parse reliably while visually clever ones don’t.

Normalization then maps the text to fields: contact information, work history, education, skills. Standard section names make mapping reliable; invented ones (“Career Journey”) often map to nothing. One honest note: every ATS vendor parses a little differently, and no resume can be perfect for all of them — the standard rules cover the vast majority.

Keyword search

Once extracted, the text becomes searchable. The system compares your resume against the job description and looks for overlapping terms — especially exact phrases and role-specific jargon. This is the step that powers the “match score” many portals show you. Coverage is a matter of degree, not pass/fail: more overlap ranks you higher, less overlap ranks you lower, but only rarely does a system outright reject based on content.

Ranking signals

Ranking is not one magic formula. Vendors weigh combinations of signals, and employers configure priorities per role. The signals that consistently matter, in rough order:

• Keyword coverage — overlap with the job description’s vocabulary.
• Field completeness — contact info, dates, sections present and parseable.
• Recency and relevance — recent roles and matching titles rank above older, unrelated ones.
• Tenure patterns — some employers weight stability; others weight growth.

That’s why a resume tailored per job (see the tailoring guide) outperforms one generic version: coverage, recency and relevance all improve together.

Knockout questions vs content

Where auto-rejection genuinely happens is not resume text but knockout questions: work authorization, location, required certifications, salary-band minimums. Be honest with these — they are hard filters, and misrepresenting them is both risky and useless, since the mismatch surfaces at the interview or onboarding stage anyway.

What we still don’t know

Honesty cuts both ways: the exact scoring weights of any given vendor are proprietary and change. That is why this guide — and every credible 2025–2026 analysis — lands on the same practical conclusion: you cannot game a system you can’t see, but you can stop losing points to parse failures and missing keywords. Coverage that is honest, structure that is standard, and content that is evidence-first will rank well in every ATS and, more importantly, in the hands of the human who reads the shortlist.

Run your resume against a job posting with the ATS Match Checker and the Format Validator — those two checks cover the mechanics this guide describes.

File formats the ATS actually reads

Parsing starts with the file itself, and format choice quietly influences everything downstream. The practical hierarchy for most modern systems: clean text-based PDF and plain .docx parse reliably; scanned images require OCR and lose accuracy; and web-form pasted text parses from whatever the portal captures. This is why the guides consistently recommend exporting from a tool rather than scanning paper — the difference is real text versus a picture of text. If you paste text into a portal, paste a plain-text version from the Plain Text Converter so formatting symbols don’t corrupt your fields.

What to do when the portal shows a low match score

Many career sites show you a “match score” after upload — and candidates routinely panic when it’s low. Before you rewrite anything, ask three questions:

• Is the scoring against a keyword list or a semantic model? Keyword-based scores rise fast when you mirror exact phrases; semantic scores care more about context and skills placement.
• Did you paste the full resume text? Scores computed from an uploaded PDF sometimes miss text the parser couldn’t read — re-export a clean PDF and re-upload.
• Is the score even the employer’s number? Often it’s the portal’s own estimate, not the employer’s ATS. Use it as a relative signal, not a verdict.

Treat any match score as directional feedback: rerun your resume through the ATS Match Checker with the actual job description, fix the missing keywords honestly, and move on — the score you control is coverage, not the employer’s final decision.