How Does ATS Work? The Filtering Logic Most Job Seekers Never See
FitWard AI Team
You rewrote your resume twice. You tailored it. You applied to 30 jobs. Nothing.
Eventually you Googled how ATS works and found the same advice recycled across a dozen articles: "use keywords," "avoid graphics," "save as PDF." Tips for a test you've already taken. What none of them explained was the actual mechanism — what happens to your file the moment you click Submit, which system touches it, in what order, and where it quietly disappears from consideration.
That's what this article covers: the three-phase mechanics behind ATS filtering, where formatting kills candidates before a human ever sees them, and how scoring actually weights your content — so you can diagnose your specific failure, not guess at fixes.
How Does ATS Work? The Three-Phase Process
ATS processes every resume in three sequential steps: parsing (extracting your text into structured database fields), keyword matching (comparing your content against the job description), and ranking (scoring you relative to every other applicant for that role). Most candidates disappear in step one — if parsing fails, the system never indexes your keywords, and you're invisible regardless of your qualifications.
Fortune 500 ATS adoption is near-universal — commonly cited estimates put it at 97.8% — and most mid-market employers follow the same approach. The technology isn't a filter reserved for large corporations. It's the entry point for nearly every corporate application you submit.
Phase 1 — Parsing: Your Resume Becomes a Database Record
When your file arrives, the ATS doesn't read it like a person. A parser scans the document and attempts to extract structured data into specific fields: name, phone, email, job titles, employers, dates, education, skills. Each field is stored separately in a candidate database.
Format matters here — not because clean design impresses anyone, but because the parser follows rigid rules. A two-column layout breaks it. The parser reads left-to-right, top-to-bottom, so it merges content from both columns into a single stream. Your job title from column one interleaves with a skill from column two, producing records like "Senior Engineer Python Deloitte 2021 Google." The recruiter searches for "Senior Engineer" and your record doesn't surface, because the title the system stored is garbled text.
A Canva-exported PDF is a total failure. Most design tools export resumes as flattened images — the file contains no extractable text at all. The parser stores a blank profile. Your application exists in the system; your data does not.
Phase 2 — Keyword Matching: How the Algorithm Reads Your Bullet Points
Once parsing succeeds, the ATS compares your stored text against the job description. Different platforms do this differently. Legacy systems like Taleo use exact string matching — "Python" and "Python development" are treated as distinct terms, and only the phrase that matches the recruiter's query scores. More modern platforms like Lever use semantic matching, where closely related terms carry partial credit .
What almost all platforms share: matching happens against the data that was parsed successfully. If your skills section didn't parse cleanly, your Python skill isn't in the "Skills" field of your database record. The algorithm checks the fields, not your original file.
Phase 3 — Ranking: How You're Scored Against Other Applicants
Ranking is relative, not absolute. A corporate job posting attracts an average of 250 applications . The ATS produces a ranked list; recruiters open the top results and stop when they have enough candidates to phone screen.
Your goal isn't to "beat" ATS — it's to rank in the top 10–25% for the specific role you're applying to. Not every role. This one. With this job description. That framing changes what you optimize for.
Why Parse Failures Are More Dangerous Than Keyword Gaps
A parse failure is silent. You receive no error, no rejection email, no indication that the system stored your work history as scrambled text. Your application status reads "Received."
Analysis of resume submissions found that 34.7% contain formatting issues that affect parsing accuracy . Among resumes reformatted to single-column layout with standard section headings, median ATS scores improved from 58 to 76 — an 18-point jump with no change to the underlying content .
A keyword gap costs you points in phase two. A parse failure kills phase two entirely. The order matters.
The 5 Most Common Parse Failures (and What Goes Wrong)
1. Multi-column layout. Parsers read linearly. Two-column layouts merge column content into a single stream, corrupting job titles, dates, and employer names in the stored record.
2. Canva or image-based PDFs. Design tools often export resumes as flattened images with no embedded text. The parser extracts nothing. Your profile is stored with empty fields.
3. Contact information placed in header or footer. Many parsers skip header and footer regions entirely. If your phone number and email exist only in the header, your contact record may be stored as blank — and you won't appear in searches filtered by contact completeness.
4. Tables used for work history. Table cells get read as a continuous stream: "Role Date Company Description Senior Engineer 2021–2024 Google Led platform migration." Dates attach to wrong employers; descriptions merge into job titles.
5. Non-standard section headings. "Career Highlights" doesn't map to the ATS field "Experience." "Technical Toolbox" doesn't map to "Skills." Sections the parser can't identify are skipped, leaving those fields empty in your record.
Any one of these failures means the system stores an incomplete or corrupted profile — and keyword matching runs against that corrupted data, not your actual resume.
How ATS Scores Your Resume — The Weighting Breakdown
After a successful parse, the ATS produces a score. Exact weights vary by platform, but the components are consistent enough to optimize for:
| Component | Approximate Weight | What It Measures |
|---|---|---|
| Keyword match | ~40% | How many required terms from the job description appear in your resume |
| Placement multiplier | ~15% | Where keywords appear: job title, summary, skills, or body bullets |
| Section completeness | ~20% | Whether all expected sections are present and filled |
| Parse success | ~15% | How cleanly structured data was extracted from your file |
| Relevance boost | ~10% | Job title alignment, seniority level, industry match |
The placement multiplier is what most guides overlook. A keyword in your current job title carries roughly three times the weight of the same term buried in a bullet point from a role you held five years ago . If "Product Manager" is the role you're targeting, it should appear in your summary and current title — not only in a responsibilities list from a previous job.
Platform behavior also differs. Workday and Taleo favor exact keyword matching — using "data analysis" when the JD says "data analytics" may score zero on those systems. Lever uses semantic matching, so related terms carry partial weight. The same resume can produce meaningfully different scores across platforms .
Check your resume's ATS score free with FitWard AI — the 4-dimensional breakdown (completeness, impact, formatting, clarity) shows which component is costing you the most.
Before and After — A Resume Bullet That Fails vs. Passes ATS
Abstract rules are easier to apply once you see them on an actual resume line. Here's the same work described two ways:
FAILING:
"Helped with various projects and contributed to team success across multiple departments."
This bullet fails in phase two. It contains no searchable keywords — no domain terms, no role-specific verbs, no quantified results. A recruiter searching for "project management," "cross-functional," or any standard skill returns nothing from this line. Even if parsing worked perfectly, there's no content for the matching algorithm to score.
PASSING:
"Led 3 cross-functional product launches, reducing time-to-market by 20% and delivering $2M in first-year revenue."
This bullet clears phase two for three reasons. "Led" is an action verb present in most mid-to-senior job descriptions. "Cross-functional" and "product launches" are indexable terms that match common search queries. The quantified outcomes (3 launches, 20%, $2M) score directly against the impact dimension of a 4-dimensional ATS check — the one that measures whether your experience section contains measurable achievements.
Both bullets describe the same work. One describes it in language the system and the recruiter can use. The difference isn't about gaming anything — it's about saying specifically what you did.
What "Getting Past ATS" Actually Means
Harvard Business School's "Hidden Workers" report found that 88% of employers believe qualified high-skill candidates are filtered before a human ever reviews them. ATS wasn't built to identify the best candidates — it was built to manage volume. When it filters out qualified people, it's working as designed, just with a flawed input.
The "75% of resumes rejected by ATS" figure you've probably encountered isn't backed by primary research. Most ATS platforms don't reject — they rank. Recruiters open the top results and move on when they've found enough candidates to interview. Your resume isn't rejected. It simply wasn't retrieved.
That distinction matters. Rejection implies a failed test. Not being retrieved means your resume didn't match the specific search terms a recruiter used on a specific day in a specific hiring cycle. You have more control over that than a rejection message suggests.
The concrete goal: parse cleanly, match the language from the job description, and land in the top 10–25% of the ranking for this role. Not a universal score. Not a single magic threshold. This job description, these keywords, in the right positions.
FitWard AI is free to start — upload your resume and get your ATS score in under a minute at fitward.ai.
Frequently Asked Questions
Q: Does ATS automatically reject resumes?
Most ATS systems don't automatically reject. They rank. Recruiters then work through the highest-ranked results and stop when they have enough candidates to interview. The practical experience feels like rejection because you hear nothing, but your application is still in the system — it just ranked below the threshold where a recruiter looked.
Q: What score do I need to pass ATS?
There's no universal threshold, because ATS ranking is relative to other applicants for a specific role. A frequently cited benchmark is a 70–80% keyword match to be competitive , but what determines whether you get a call is whether you rank in the top results for that posting — which depends on both your score and everyone else's. A score that works for a niche role may not be enough for a high-volume posting at a Fortune 500 company.
Q: Do all companies use ATS?
Large and mid-market employers nearly always do. Fortune 500 ATS adoption is near-universal at 97.8% , and most companies with 250–5,000 employees follow the same approach. Smaller businesses may still manage applications manually or by email, but any organization posting on major job boards (LinkedIn, Indeed, Glassdoor) almost certainly routes applications through an ATS before a recruiter opens anything.
Q: Does ATS read PDFs?
It depends on how the PDF was created. Text-based PDFs exported from Word or Google Docs parse correctly on modern ATS platforms. Image-based PDFs — including most exports from Canva and graphic design tools — contain no extractable text. The parser stores a blank record. If you're unsure how your PDF was created, open it in a browser and try to select the text. If you can't highlight your name, the file will fail to parse.
Q: How do I know if my resume is ATS-friendly?
The fastest manual test: copy-paste your resume text into a plain text editor. If the result is scrambled — job titles mixed with bullet points, dates in wrong positions, contact information missing — your format is causing parse failures. A dedicated ATS scorer catches these issues automatically. FitWard AI's 4-dimensional check (completeness, impact, formatting, clarity) flags specific problems with a single upload, so you know which issues to fix rather than rewriting the entire document.