INSIGHTS

Does the ATS Really Reject Your Resume? What Hiring Software Actually Does

6 MIN READ

SHARE THIS ARTICLE


Does the ATS Really Reject Your Resume? What Hiring Software Actually Does
SUMMARY

No major hiring platform documents rejecting a resume on a keyword score. Here is what Greenhouse, Ashby, Workday and Lever actually do with your application.

"75% of resumes are rejected by an ATS before a human ever sees them." You've probably read that line. It's on career blogs, in LinkedIn carousels, in the pitch of every resume optimizer on the internet.

It traces back to Preptel, a company selling resume optimization, around 2012. No published method. Forbes repeated it in 2014, CIO in 2018, CNBC in 2019. Preptel shut down in 2013. The number outlived the company that made it up.

So we went to the source. We read the documentation from Greenhouse, Ashby, Lever, Workday, Oracle and Teamtailor, the engineering notes of the companies that build resume parsers, recruiter surveys and the few real field studies. Then we checked it against more than 7,000 applications we filled in for job seekers over five weeks this fall.

Short version: the software rarely reads your resume and says no. The form does.

Four things get called "the ATS rejected me"

An ATS (applicant tracking system) is just the software a company uses to collect and sort applications. Four separate things happen inside it, and only one of them rejects anyone automatically.

1. Screening questions: the one that can say no on its own

"Are you authorized to work in the US?" "Do you need sponsorship?" "Can you work from Berlin three days a week?"

These are knockout questions, and they're the documented automatic rejection. Greenhouse's Auto-Reject fires when "an applicant's response to a question fulfills the rule's conditions" and can send the rejection email by itself. Oracle calls them disqualification questions. JazzHR and Workable have the same thing.

Notice what's being judged: your answer to a question, not your resume.

2. Parsing: where information gets lost

When you upload a file, the system turns it into fields: name, contact, job titles, employers, dates, skills. Workday's own docs say results "vary based on resume format and order of words" and tell you to avoid "image-based styles". Lever can't parse image files at all.

This is the real technical risk. Not a rejection, but a profile with a blank where your last job should be.

Parsers are getting better. Textkernel, which powers parsing for SAP SuccessFactors and integrates with Oracle, says its newer model reads multi-column resumes correctly 90% of the time, up from 62%. Better, still not perfect.

3. AI matching: ranks, doesn't reject

Every big platform now has an AI feature. Greenhouse has Talent Matching, Ashby has AI-Assisted Application Review, Workday has HiredScore, iCIMS has Role Fit.

They sort and summarize. Greenhouse states Talent Matching "does not automatically advance or reject candidates". Ashby's version marks each criterion the recruiter set as met or not met, with citations from your resume, and it's opt-in. HiredScore grades applicants A to D.

Ranking matters, because a recruiter with 400 applicants starts at the top. But a person still clicks the button.

4. Recruiter search: decides who gets found

This is where keywords actually matter, just not the way the myth says.

Recruiters search their candidate database. Greenhouse supports full-text Boolean search plus required and preferred keyword filters. In Jobscan's 2025 survey of 384 recruiters, 76% filter by skill keywords, 55% by job title and 51% by certifications.

So if a role says "PostgreSQL" and your resume says "relational databases", you're harder to find. Today, and three months from now when a better role opens at the same company.

Platform by platform

PlatformCan reject automaticallyAI featureWorth knowing
GreenhouseYes, on screening-question answers (Plus and Pro plans)Talent Matching tiers, never rejectsIdentity checks and fraud signals on applications
AshbyLimits on how many roles one person can apply toMet / not met per criterion, opt-inRe-reads your uploaded resume to prefill the form
LeverNo documented rule foundFit score, per recruitersCan't parse image files
WorkdayNo documented rule foundHiredScore A to D gradesSubject of an age-discrimination lawsuit, no liability ruling
Oracle / TaleoDisqualification questionsPrescreening scores rankTextkernel parsing available
TeamtailorNo documented rule foundCo-pilot summaryDetects and logs hidden white text

None of these companies says which parser they use, except where Textkernel is named. Anyone who tells you "Greenhouse uses parser X" is guessing.

The myths, checked

"ATS can't read PDFs." False. Lever and Workday list PDF as supported. What fails is a scanned or image-only PDF, where the "text" is a picture of text.

"You need an 80% keyword match." False. 75 to 80% is the target inside Jobscan's own tool, and Jobscan itself warns against aiming for 100%. No hiring platform documents a percentage cutoff.

"Hide keywords in white text." False, and it gets caught. Teamtailor flags hidden text. Duke and hireEZ found hidden prompts in at least 1% of about 200,000 resumes, up sevenfold in 16 months. Some recruiters reject on sight.

"Designed templates get auto-rejected." Mostly false. No platform rejects on format. The risks are a bad parse of columns, and taste: 20% of recruiters in an Enhancv sample disliked Canva-style resumes.

"Keep it to one page." Unsupported. ResumeGo had 482 hiring professionals compare resumes, and two pages were preferred 2.3 times as often overall. That study is run by a resume company, so hold it loosely.

"Tailoring doubles your interviews." Correlation, not proof. Huntr reports a 4.23% interview rate for resumes tailored in its tool versus 2.07% for those that weren't. Self-reported, not a controlled test. Tailoring is still worth doing, just don't expect a guaranteed 2x.

What we saw across 7,000+ applications

Clera fills in applications on job seekers' behalf. Over five weeks this fall that was more than 7,000 applications for 700+ people across Ashby, Greenhouse, Lever, Workday and others.

The resume was almost never the problem. What stopped an application was the form: required questions only the candidate could answer, account creation, email verification codes, bot checks. On BambooHR, nine in ten applications paused for a question only the person could answer. On Workday, more than half did, and account creation stopped nearly a third outright.

That lines up with the documentation. The form is where applications live or die.

What actually helps

In rough order of how much evidence is behind it:

  1. Answer screening questions truthfully and the same way every time. It's the only documented automatic rejection. Say you're authorized to work in Germany on one form and that you need sponsorship on another, and you'll trip a rule somewhere.
  2. Send a text PDF or a DOCX, one column, standard headings, contact details in the body. Not in the header, not in an image. This avoids the parse failures the vendors themselves describe.
  3. Use the posting's exact skill names and a standard job title. Write both the acronym and the full form once: "Amazon Web Services (AWS)". Only where it's true for you.
  4. Write clearly. A 2025 field experiment in Management Science with about 480,000 job seekers found that writing help raised hires by about 8% and wages by about 10%. Clear writing beats keyword stuffing, measurably.
  5. Apply to fewer roles, more carefully. Huntr's data shows interview rates fall from 9.25% for people sending 11 to 20 applications to 2.58% at 100+. Eight or more applications to one company convert at 1.91% versus 6.07%. Observational, but the direction is hard to ignore, and Ashby and Greenhouse are adding per-person application limits.
  6. Put numbers on your results. Everyone advises it. We couldn't find a rigorous study proving it moves outcomes, so treat it as good practice, not a hack.

The short version

No major hiring platform documents throwing out your resume because a keyword score came in low. The automatic "no" comes from the screening questions. The AI ranks, a person decides. Keywords matter because recruiters search, and a clean, one-column text resume makes sure there's something for them to find.

Skip the "beat the ATS" tricks. Answer the questions honestly, make your resume easy to read for people and parsers alike, and send fewer, better applications.

Sources

Hiring? Let Clera source and screen your next hire

See how Clera compares to recruiting agencies