How an ATS reads your CV — and what it doesn't

You press apply, the page thanks you, and the application disappears. If the employer runs an applicant tracking system (ATS) — the software a company uses to receive applications and work through them — your CV (resume) has just landed in one. What happens next is less mysterious than it feels, because the companies that build these systems publish how they work: what is read automatically, what a person does, and which rules can end an application before anyone reads a word of it.

Every sentence below about how these systems behave links the vendor page it came from, in the same paragraph, so you can check it yourself. Where the documentation says something different from what the internet says, this guide follows the documentation. Where no vendor documents a thing at all, this guide says nothing about it — which is why you will not find a single percentage on this page.

This guide is about what happens to your CV after you press apply: how the software files it, what a recruiter can search for, and which rules can end an application. How to rewrite your CV for one posting is how to tailor your CV to a job description; which words to take from the posting is pulling the right keywords from a job posting.

What an ATS is, in one paragraph

It is the employer's filing system for applications, with a workflow bolted on. When you apply, the system files your application against that job; a recruiter then reviews it and either moves it forward or declines it, and records why. Workday's own course manual for recruiting administrators describes exactly that shape — an application creates a review step, a recruiting user reads the CV, and chooses to move the candidate forward or decline them, with a disposition recorded (Workday, Recruiting Subprocesses). That is the spine of the whole thing. Almost everything else in this guide is detail hanging off it.

What it does with your file: parsing

The first thing that happens is mechanical. The system reads your file and copies what it finds into fields on your candidate profile — Greenhouse's help centre calls this being parsed, and says the point of it is that a recruiter does not have to type your information out (Greenhouse, Manually add a candidate or prospect).

Which fields depends on the system. In Personio, for example, parsing turns the file into plain text and fills fields such as name, contact details, current employer and current title — and a recruiter can then filter the candidate list by those values (Personio, CV parsing for candidate profiles). That is worth sitting with for a second. Your current job title is not decoration on your CV; in that system it is a value somebody can filter on.

What breaks it

Parsing works on text, and it is layout that breaks it. Greenhouse documents the causes of a failed or partial parse: tables, headers and footers, columns, graphics, letter-spaced words, image files, and contact details placed inside text boxes (Greenhouse, Unsuccessful resume parse). Personio's help page lists much the same set from its own side — details inside images, ready-made templates, scanned files, headers or footers, tables or multi-column layouts, very small fonts and unusual font styles (Personio). Two vendors, independently, naming the same handful of design choices.

What happens when it breaks is duller than the version you have probably read. Both pages say the details then have to be entered by hand (Greenhouse; Personio). Neither says the application is rejected, and neither says it vanishes. What they describe is a profile whose fields are empty or wrong until someone fixes them — and those fields are what a recruiter's filter runs against. Note also what the documentation actually names: a scanned CV, or an image of one (Personio; Greenhouse). Not a file format. The common advice to avoid one particular format appears on neither page.

What actually filters

Three different things narrow the pile, and only one of them is automatic — because somebody configured it to be.

Knockout questions — a rule a person set

A knockout is an application-form question whose answer can end the application on its own. It is not a property of the software; it is a setting somebody switched on. In Greenhouse, an administrator sets a rule so that one specific answer to a job-post question rejects the applicant automatically, attaches a rejection reason, and can send an email (Greenhouse, Auto-reject). In Recruitee, whoever sets the job up toggles knockout on a yes/no or multiple-choice question, chooses which answer disqualifies, and picks the reason that gets recorded (Recruitee, Set up knockout questions to automatically disqualify candidates).

Where the application then goes differs by system, and it is worth knowing. In Workable a disqualifying answer moves the application to a Disqualified list: it is not deleted, the candidate is not told by the rule itself, and a recruiter can revert them to qualified (Workable, Auto-disqualify candidates using application form questions). In SmartRecruiters, an administrator marks which answers knock a candidate out; the rejection is immediate, and the email — if the employer sends one at all — can be delayed (SmartRecruiters for SAP SuccessFactors Academy, Creating Rejection Screening Questions). That last detail quietly kills a popular inference: since the timing of that email is a setting, a rejection arriving twenty minutes after you applied tells you nothing about how it was decided (SmartRecruiters).

The same capability exists in the larger suites, wired slightly differently. Workday lets an employer attach a questionnaire to a job requisition and configure answers that disqualify (Workday, Job Requisition Components), and separately lets an administrator write condition rules — over the application, the requisition or the questionnaire answers — that automatically advance or decline a candidate (Workday, Steps: Automatically Advance or Decline Candidates).

So yes: an application can end without a person reading it. In every case the documentation describes, it ends because an employer chose the question, nominated the answer, and turned the rule on. That is the whole of the automatic part, and it is why the screening questions deserve more of your attention than they usually get.

Recruiter search — where the posting's words actually matter

The second filter is a person typing. Greenhouse's candidate search is a full-text Boolean search over every CV in the system and the notes attached to them — this and that, this but not that, quoted phrases (Greenhouse, Search candidates using Boolean queries). Its talent filtering goes a step further — a recruiter searches CV and note text for titles, skills and locations, and can mark a term Required, in which case every required term must appear, while one Preferred term is enough (Greenhouse, Talent Filtering).

That is the moment the posting's own vocabulary matters, and it has nothing to do with a score. When a recruiter marks a term required and your CV does not contain it, your CV is not in the list they are looking at (Greenhouse). Nothing in that documentation says how the results they do see are ordered, so this guide says nothing about that either.

Match and ranking views — optional, configured, and read by a person

Some systems offer a view that compares your application against the job's requirements. These are features an employer turns on, not properties of every ATS. In Greenhouse, talent matching is an add-on: the recruiter chooses the skills, and applications are grouped by how well they match — nobody is advanced or rejected by it (Greenhouse, Talent Matching). Its FAQ explains what the comparison does, and it is not arithmetic on repetition: it relates terms on your CV to the calibrated skills, and because several terms can map to a single skill, more matching terms do not always mean a higher match (Greenhouse, Talent Matching FAQ). And where there is a rating, the decision is still not the software's — Greenhouse's data-processing FAQ states that humans are required to make the decision, and that nothing is advanced or rejected automatically on the strength of an AI rating (Greenhouse, Talent Matching – Data Processing FAQ).

Other vendors do rank. On jobs that use the Workable Agent, candidates are ranked by a match score against criteria the employer wrote, weighting must-haves more heavily; an optional setting will auto-disqualify candidates missing a criterion the employer marked that way, and only if the employer switches it on (Workable, Using the Workable Agent). Workday's HiredScore grades applications by comparing the job's requirements with the CV, as a way of prioritising whom a recruiter looks at first (Workday, Reference: HiredScore AI for Recruiting).

Read those pages together and one pattern holds across all of them: the comparison runs against requirements a person wrote, and what comes out of it is an order to read in — unless an employer has deliberately enabled something stronger.

The myth, quoted and corrected

You may have read that "the ATS rejected you". Here is what is knowable about that.

No vendor documentation read for this guide states how many CVs are ended by software, or what share reach a person. The figures that circulate about it appear on none of these pages, which is why there is no figure on this one. What the documentation does describe is narrower, and more useful to you. An application can end automatically — Greenhouse's auto-reject (Greenhouse), Workable's disqualification (Workable), Workday's condition rules (Workday) and the SmartRecruiters knockout (SmartRecruiters) — and in each documented case it ends because an employer configured a rule, on a question they wrote, with an answer they nominated.

The opposite over-correction is wrong too, and this guide will not make it. It is not true that every application is seen by a person; the pages above describe exactly how that can fail to happen. What is documented is that where a system rates the match, a person still makes the call — Greenhouse's documentation says so in those words (Greenhouse) — and that the one assistant here which can end an application by itself does so only where the employer switched that setting on (Workable). Which turns the useful question around. It is not what the software thinks of you. It is which of these specific rules your application is actually meeting — and that one you can answer.

What helps, in order

1. Answer the screening questions truthfully and precisely. This is the one place where your own answer, alone, can end the application, because the employer nominated the disqualifying answer in advance (Greenhouse; Recruitee). Read each question for what it literally asks — about a licence, a location, a notice period, a right to work — and answer that. Rounding an answer up is how people remove themselves from jobs they could have done.
2. Give it a CV it can read. One column; real text rather than text inside an image; contact details in the body rather than in a header, a footer or a text box; nothing load-bearing trapped in a table. Those are the causes both vendors name for a failed or partial parse (Greenhouse; Personio). The cost of ignoring it is not a rejection; it is a profile with the wrong employer in it and no job title, until somebody notices and types the details in by hand.
3. Use the posting's own terms, where they are true, inside your own sentences. A required term that is missing keeps your CV out of the list the recruiter is reading (Greenhouse). Deciding which of the posting's words are load-bearing is the whole of pulling the right keywords from a job posting.
4. Make every must-have visible on the page. Where a match view exists, it compares your CV with requirements a person set (Greenhouse; Workable); where none exists, a person is making the same comparison by eye. Either way, a requirement your CV answers in the last line of your oldest role is a requirement your CV does not answer. Rearranging the document around one posting is how to tailor your CV to a job description.

What does not help

Repeating the posting's terms is the first mistake in common CV tailoring mistakes, and that guide covers the fix. Here is only the mechanism: a search finds a word, and then a person reads the sentence it sits in.

The matching side says the same thing in its own words. Greenhouse's FAQ notes that several terms can map to one calibrated skill, so a longer list of matching terms does not always produce a higher match (Greenhouse, Talent Matching FAQ). Repetition, in other words, is not a lever you are pulling. It is a lever attached to nothing, sitting next to a sentence somebody is about to read.

The tricks are in the same category, for a plainer reason: nothing sourced here covers them. Hidden text, white-on-white keywords, a job description pasted invisibly into a footer — this guide has no vendor documentation behind any of it, so it says nothing about how any system treats it. What the documentation does cover is the part that makes the question moot: when an application reaches the review step, a recruiting user reads the CV and decides (Workday).

Example: one requirement, and the story that answers it

Example — a composite posting and CV, assembled to show the shape. Not a real candidate.

The posting says: Required — experience running a supplier tender end to end.

The CV says: Procurement Analyst, 2022–2025. Supported the procurement team across sourcing, contracts and supplier reporting.

The term a recruiter would mark required: tender. That one word decides whether this CV is in the results at all, because every required term has to appear for the CV to come back (Greenhouse, Talent Filtering). On the CV as written, it does not appear — and not because the candidate has never run a tender. It is missing because "supported the procurement team across sourcing" is the sentence people write when they are compressing three years into one line.

What actually happened, in the candidate's own account:

• Situation — the packaging contract was three years past its review date and the incumbent supplier had raised prices twice in a year.
• Task — the head of procurement asked the candidate to take the category to market before the next price review, with the operations lead as the technical reviewer.
• Action — the candidate wrote the specification with the operations lead, ran a market scan, shortlisted six suppliers, built the scoring sheet, ran the bid clarifications, and recommended a supplier to the approval panel.
• Result — the category moved to a new supplier on a two-year contract, and the candidate wrote the review schedule that stopped it lapsing again.

The line that now goes on the CV: Ran the packaging supplier tender end to end — wrote the specification with the operations lead, scored six bids against it, and moved the category to a new two-year contract.

The required word is on the page now, and it is on the page because the work is. That is the only order in which this is safe to do — the story first, the term second, never the other way around.

The example stops at one story. If the requirement is one your CV does not cover at all, a STAR story for the requirement your CV lacks is the full method — including the case where the honest answer is that you have not done it.

Check yourself

• Does it parse? One column, real text, contact details in the body, nothing that matters trapped in a table, a header, a footer, a text box or an image (Greenhouse; Personio).
• Are the screening questions answered truthfully, and exactly as asked? An answer the employer nominated ends the application by itself (Greenhouse; Recruitee).
• Is every must-have there in the posting's own words, inside a sentence that is true? A required term that is missing keeps the CV out of the recruiter's results (Greenhouse).
• Is anything repeated for its own sake? More matching terms do not always mean a better match (Greenhouse).
• Would the document survive being typed out by a stranger? If parsing fails, that is close to what happens next (Greenhouse).

None of this is a trick, and none of it is a guarantee. It is a short list of places where an application can quietly stop — a question you answered carelessly, a layout that lost your job title, a required word your CV never says — and each of them is fixable before you press apply rather than guessed at afterwards. HireReady24 does the last two mechanically: it reads the posting, shows which requirements your CV visibly answers and which it does not, and builds the missing line out of what you actually did. The screening questions stay yours, because only you know the true answer.