Summary
An applicant tracking system parses your resume into database fields, records your answers to application questions, and puts you in a queue recruiters search. No major ATS auto-rejects based on resume content. The only automatic rejections come from knockout questions, and a Workday analysis of 50,000 resumes found those eliminate more candidates than formatting ever does. The second filter is boolean search: if your resume lacks the term a recruiter searches, you are invisible rather than rejected.
Ask how an ATS works and you will usually get a description of a robot with opinions — software that reads your resume, assigns a secret score out of 100, and deletes the bottom half.
That picture is wrong in a way that matters, because it sends people to fight the wrong opponent. A recruiter at Microsoft put it more accurately: an applicant tracking system is a filing cabinet, and filing cabinets do not make decisions.
So here is how an ATS works in practice: the actual pipeline, the two mechanisms that really do eliminate candidates, and what each of the major platforms does differently.
How does an ATS work, step by step?
An applicant tracking system does three things when you apply: it stores your resume file, records your answers to application questions, and places you in a queue for a recruiter to review. It parses your file into structured database fields so recruiters can search and filter the pile.
How an ATS works is the same across every major platform, even though the implementations differ:
1. Submission. Your file and your form answers arrive together. The form answers matter more than most people realise.
2. Parsing. The system extracts your name, contact details, employers, job titles, dates, education and skills into structured fields. Your resume stops being a document and becomes a database record — and the parsed record, not your carefully designed PDF, is often what the recruiter actually sees.
3. Knockout filtering. Answers that fail a hard requirement disqualify the application immediately.
4. Search and sort. Recruiters query the pool. In most configurations the candidate list opens sorted by application date and pipeline stage, not by any score.
5. Human review. A person decides who advances. Every advance and reject decision is made by a human.
Nothing in that sequence involves software forming a judgement about whether you are good.
Does an ATS automatically reject resumes?
Not on resume content. No major applicant tracking system auto-rejects based on what your resume says. The only automatic rejections come from knockout questions on the application form, which trigger on your answers rather than your resume text.
In fact, this is the single most persistent misconception in the category, and correcting it changes what you spend time on.
Workday, Greenhouse, Lever, Ashby and the rest do not scan your bullet points and bin you for insufficient keyword density. What they do is let a recruiter configure hard requirements as application questions — work authorisation, a required licence, minimum years of experience, willingness to relocate. Answer in a disqualifying way and the rejection is binary and immediate, with no human review.
Moreover, the scale of this is badly underappreciated. An analysis of 50,000 Workday applications found knockout questions eliminate more candidates than formatting ever does. No amount of layout optimisation recovers from a disqualifying answer.
Read the application questions before the job description
Knockouts are usually three to six questions on the form, and they are the only part of the process that can end your application in one click. Read them carefully and answer honestly u2014 a wrong answer to "do you have a valid certification" is unrecoverable, and a dishonest one is worse. If you fail a genuine hard requirement, that role was never available to you, and finding out in thirty seconds is a gift.
How do recruiters actually find your resume?
By searching, not scrolling. Recruiters build boolean queries from the job requirements u2014 specific tools, certifications, job titles u2014 and only matching profiles appear. If your resume lacks the exact term being searched, you are not rejected. You are invisible, which has the same effect.
Meanwhile, this is the second real filter, and it is the one worth optimising for.
A recruiter facing four hundred applications does not read four hundred resumes. They open the candidate list and search it like a database: "Kubernetes" AND "AWS" AND "site reliability". Profiles containing those terms surface. Profiles without them stay in the pile, unseen.
As a result, that mechanism explains why vocabulary matching matters so much, and why it is not the same thing as keyword stuffing. You need the term present once, in a true context. Repeating it six times does nothing for a boolean search — a field either contains the string or it does not — while a human reading the shortlist notices the repetition immediately.
Use the exact nouns from the posting for tools, certifications and titles you genuinely have u2014 "Kubernetes", not "container orchestration"; "CPA", not "accounting qualification". Boolean search is literal about nouns.
Assume a synonym will be found. Semantic matching helps in the ranking layer, but recruiter search is often plain text against specific fields, and "program management" does not surface in a search for "project management".
Do all applicant tracking systems work the same way?
The pipeline is identical everywhere, but the mechanics differ. Workday uses OCR extraction with a machine-learning relevance layer and notably strict knockouts. Greenhouse uses structured boolean field search with human scorecards. Lever weights full-text search by section. Ashby filters most aggressively at field level.
Workday dominates at scale: verified data across more than 700 large employers puts it at roughly 39% of large employers, with Greenhouse around 13%, followed by SAP SuccessFactors, Oracle Cloud, iCIMS and Taleo.
| Platform | How it reads you | What matters most |
|---|---|---|
| Workday | OCR text extraction, then an NLP layer that reads meaning | Answer knockouts correctly; its knockouts are unusually strict |
| Greenhouse | Structured boolean search across profile fields, plus human scorecards | Exact terms in the right fields; evidence for each scorecard attribute |
| Lever | Full-text search weighted by section | Recent, explicitly named skills |
| Ashby | The strictest field-level boolean filtering | Precise field data; least tolerance for parse errors |
| Taleo | Older keyword-based matching | Conventional structure; the least semantic tolerance |
Knowing how each ATS works is interesting, but the practical takeaway is not to optimise per platform — you rarely know which one you are facing. It is that a resume which parses cleanly and names things precisely works everywhere, while one that relies on a parser being clever fails on the strictest system in the set.
Is there really an ATS score?
Not one the recruiter sees. Workday, Greenhouse, Lever and Ashby do not display a universal match score out of 100 by default. The percentages shown on consumer resume-checker sites are calculated by those sites as a proxy for keyword coverage and parse quality u2014 useful as feedback, but not the number on a recruiter's screen.
Frankly, this deserves saying plainly, including by anyone who builds these tools.
The recruiter’s default view is a list sorted by application date and pipeline stage. There is no secret ranking your resume is locked into. Some platforms add optional AI scoring layers that surface strong candidates near the top, but even then a recruiter must make every advance decision, and not all recruiters turn those features on.
So what is a checker score actually good for? It measures the two things that genuinely determine whether you are findable: whether your fields extracted correctly, and whether your language overlaps the posting. Both are real gates. The number is a proxy for them, and it is a useful proxy — but treat the gap list as the output that matters, since it names specific missing terms you can act on. What an AI resume checker actually measures covers how to read those numbers without over-reading them.
An ATS is a filing cabinet. That's it. It doesn't make decisions.
What should you actually do about it?
Answer knockout questions carefully, make the file parse cleanly, and use the posting's exact nouns for things you have genuinely done. Those three cover every mechanism that removes candidates. Everything after that is written for the human who opens your file.
Four moves, in order of how much they matter:
Answer the form honestly and carefully. This is the only automatic rejection in the system and it eliminates more people than anything else.
Make it parse. Single column, standard headings, consistent dates, .docx unless the posting says otherwise. Confirm every field extracts correctly rather than assuming.
Name things precisely. Exact tool names, exact certification names, exact job titles, for work you actually did. This is what makes you findable in a search.
Then write for the person. Once you are findable, the resume has to survive a human scan — how long recruiters look at a resume covers what happens in those seconds, and what each reader checks covers where the two audiences want different things.
Invisible: "Managed container orchestration and cloud infrastructure for a distributed platform."
Findable: "Ran Kubernetes on AWS EKS for a 40-service platform, cutting deploy time from 25 minutes to 4."
Same work. The first one never appears in a search for "Kubernetes" or "AWS", so no human ever forms an opinion about it.
The reframe worth keeping, once you know how an ATS works: it is not judging you. It is a search engine, and your job is to be findable in it and then convincing to the person who finds you. Neither of those requires beating an algorithm — and the anxiety about whether employers can detect an AI resume is a distraction from both, since no system in this pipeline looks for that at all.
Frequently asked questions
How does an ATS work?
It stores your resume, parses it into structured database fields, records your answers to application questions, and places you in a queue recruiters search and filter. A human makes every advance or reject decision.
Does an ATS automatically reject my resume?
Not based on resume content. The only automatic rejections come from knockout questions on the application form — work authorisation, required certifications, minimum experience — which trigger on your answers, not your resume text.
What gets most candidates eliminated?
Knockout answers first, invisibility second. An analysis of 50,000 Workday applications found knockout questions eliminate more candidates than formatting does, and boolean search means a resume missing a searched term never surfaces at all.
Does the recruiter see a match score?
Usually not. Workday, Greenhouse, Lever and Ashby do not show a universal score out of 100 by default; the list opens sorted by application date and pipeline stage. Scores on consumer checker sites are calculated by those sites.
Which ATS do most companies use?
Workday leads at roughly 39% of large employers, followed by Greenhouse at about 13%, then SAP SuccessFactors, Oracle Cloud, iCIMS and Taleo.
Do ATS systems understand synonyms?
Increasingly, in the ranking layer — modern systems use natural language processing trained on large volumes of resumes and postings. But recruiter boolean search is often literal, so exact nouns for tools and certifications still matter.
Should I send a PDF or a Word document?
A text-based .docx extracts most reliably across the widest range of parsers. A normal text-based PDF works in most modern systems, but a PDF exported from a design tool often does not.
Can I trick an ATS with hidden keywords?
No, and it backfires. Hidden text and keyword stuffing are actively detected and can flag an application as manipulative, and the human reading the shortlist sees the result either way.
Comments