What Is ATS? How Applicant Tracking Systems Work (2026 Guide)

Whose tool it is, and what it holds

The employer buys the system to run a hiring process: post the vacancy, collect applications, keep them in one place, move people through stages, and keep a record of it all. You are a record in that database with a file attached.

Oracle's Taleo Enterprise documentation is the most explicit about what gets stored. Its resume parsing extracts first, middle and last name, address, city, postcode, home phone number, email address, place of residence and the pasted CV text; under education it takes education level, graduation date, start date, institution and programme; under work experience it takes start date, end date, employer, job function and responsibility. That list is the shape of the thing. Your CV becomes those fields.

Around the fields sits a search tool. Workday's administrator guide documents search prefixes for recruiters, including cand: for a candidate and jr: for a job requisition. Greenhouse's talent filtering searches "job titles, skills, locations, and other keywords in the full text of resumes and internal notes". iCIMS exposes a search API that queries candidate records with filters and operators. None of this is a machine forming an opinion about you. It is a recruiter looking things up.

What happens to your application, in order

1. The file is read into fields

The parser tries to turn your document into the fields above. Taleo Enterprise requires the file to be Word, WordPerfect, text, RTF, HTML, PDF, Excel or OpenOffice Writer, and says the CV "cannot exceed 100 kilobytes or the size defined by the system administrator". iCIMS extracts text by default from a long list of types including pdf, doc, docx, rtf and txt, "though client configurations may vary". Workday's guidance is blunter: parsing results "can vary based on resume format and order of words", and it does not auto-fill Languages or Skills at all.

When parsing fails, Greenhouse tells the recruiter that they "need to manually input the candidate's details into the fields". That is the whole consequence in the vendor's own words: somebody types it in, or the fields stay empty. The ATS-friendly format guide covers what makes a file fail this step.

2. The questions on the form are checked against rules

This is the step that gets missed, and it is the one that can end an application in seconds. Greenhouse describes application rules as "automatic actions applied to job post questions", triggered "when an applicant's response to a question fulfills the rule's conditions". One of those actions is auto-reject, which "uses custom application questions to screen and reject candidates based on their responses" and can assign a rejection reason and send the rejection email without a person reading anything. Greenhouse adds that anyone who normally gets notified of new applicants "will not be notified of auto-rejected applicants".

Oracle Taleo calls the same idea a disqualification question: "a single-answer question that contains the minimum requirements for a candidate to be eligible for a job", where "a candidate not meeting the required response can be instantly exited from the application process". Workday uses condition rules that can read questionnaires "completed when a candidate first applies", with documented examples of declining candidates "ineligible to work in the country where the position is located" or who "failed the background check".

Notice what all three have in common. They act on a structured answer you gave — a yes, a no, a licence, a location, a salary figure — not on the prose of your CV. The tick-box section of the application form deserves the care most people spend on their font choice.

3. A recruiter searches what survived

Greenhouse supports boolean queries on stored candidates: AND, which "limits your search because you're requiring that any search term following AND must appear in the results"; OR, which "provides options in a search so that the results will include either or both terms"; NOT, "the command of exclusion"; quotation marks for an exact phrase; parentheses to "group clauses to form sub queries"; and a wildcard to "represent any number of characters", with its own examples (MIT OR Harvard) AND "Computer Science" and Java*.

Its talent filtering makes the logic explicit for recruiters who do not write queries: preferred keywords mean "only one term must appear in a candidate's application to be included in the search results (OR statements)", while required keywords mean "all terms must appear (AND statements)". iCIMS builds the same shape into its search API, combining filter groups with AND and OR operators. This is how a shortlist is really produced on a high-volume role, and it is why the specific words on your CV matter more than any score.

4. The record stays

Nothing is deleted when you are not shortlisted. Your record remains in the database that Greenhouse's talent filtering searches, which is why an application that went nowhere in March can produce a call in September.

"Rejected by the ATS" means two different things

A failed parse still costs you something, which is why the distinction is not an excuse to ignore formatting. A field nobody filled in is a field nobody can search, and the recruiter searching by keyword will not find what the parser never stored. Assume anything the parser missed may as well not be on your CV.

But the honest summary is narrower than the folklore. These systems do not read your career and form a view of it. They store what they can extract, apply the rules the employer wrote, and wait for a person to search. The place where an application genuinely ends without human involvement is a question you answered, and you can usually see those questions on screen before you press submit.

What the numbers do and do not show

The figure everyone repeats comes from Jobscan's annual usage report, which says it "detected an ATS for 97.4% of Fortune 500 companies" in 2026 — 487 of 500 — by reverse-engineering each company's careers pages. Two caveats belong with it: those are large US corporations, not UK employers, and Jobscan sells CV-optimisation software. I could find no equivalent published figure for the UK at all. If an advert sends you to the employer's own branded application portal rather than to somebody's email address, a system is receiving your application; that is as far as the evidence goes, and it is far enough to be useful.

Two more Jobscan numbers get quoted without their method, so here it is. Its State of the Job Search report says candidates whose CV job titles matched the target title "had an interview rate 10.6 times higher than those who did not", measured across more than 2.5 million applications made by people using Jobscan's own tools — so it describes users of a CV-optimisation product, not a controlled study. In a separate survey of 384 recruiters in February and March 2025, 55.3% said they use job titles as a keyword to filter applicants. Both point the same way as the vendor search documentation above: job titles are search terms.

What I have deliberately not given you is a target match percentage. No vendor documentation publishes a threshold, no source I could open supports the "75%" figure that circulates, and a score from a third-party checker is that tool's arithmetic rather than a reading from the employer's system.

What to do with this

Three things follow, in the order they affect an application.

1. Make the file readable. One column, real text, contact details in the body of the page, the file type the advert asks for. The format checklist traces each rule to the vendor page that documents it.
2. Answer the form's questions carefully. This is the documented automatic exclusion. Read every question about right to work, location, notice, salary and qualifications, and answer it accurately — an untrue answer here is a problem you carry into the interview.
3. Use the employer's words. Recruiters search stored text with AND, OR and exact phrases, so the terms on your CV are the terms that find you. The keywords guide covers choosing them, and tailoring your CV to a job description shows a full pass over one advert.

None of this substitutes for being a plausible candidate. It removes the failure modes that have nothing to do with whether you can do the job.

Sources

• Greenhouse Support: Unsuccessful resume parse
• Greenhouse Support: Application rules overview
• Greenhouse Support: Auto-reject
• Greenhouse Support: Search candidates using Boolean queries
• Greenhouse Support: Talent Filtering
• Oracle Taleo Enterprise: Candidate Management (resume parsing)
• Oracle Taleo Enterprise: Candidate Prescreening (disqualification questions, ACE criteria)
• Workday Administrator Guide: Concept: Resume Parsing
• Workday Administrator Guide: Steps: Automatically Advance or Decline Candidates
• Workday Administrator Guide: Reference: Search Prefixes for Recruiting
• iCIMS Developer Resources: Binary Files (text extraction)
• iCIMS Developer Resources: Search API
• Jobscan: Applicant Tracking System Usage Report (Fortune 500)
• Jobscan: State of the Job Search