What an ATS actually does (and why most advice is wrong)
The internet is full of myths about ATS software auto-rejecting your CV. Here's what Greenhouse, Lever, and Workday actually do — and where your real risk sits.
A friend told me last month that her CV got "rejected by the robot" at a Berlin startup. She'd found a LinkedIn post claiming that if your keyword match score is below 80%, you're filtered out before a human sees you.
That startup uses Greenhouse. Greenhouse does not have a keyword match score. It does not auto-reject anyone.
The advice industry around ATS software is mostly nonsense, and it's making people anxious about the wrong things. Let's sort out what's real.
What an ATS actually is
An Applicant Tracking System is a database. That's the honest one-line description.
It stores applications, tracks candidates through stages (applied, screening, interview, offer), sends automated emails, and gives recruiters a pipeline view. The big players in Europe are Greenhouse, Lever, Workday, SmartRecruiters, and Personio. Each works a little differently, but the core is the same: it's CRM software for hiring.
The myth is that the ATS reads your CV, scores it against the job description, and tosses you below some threshold. That's Taleo-era folklore applied to modern tools that don't work that way.
What Greenhouse, Lever, and Workday actually do
Here's what happens when you hit submit on most modern European ATSes:
- Your CV file gets stored as-is.
- The system tries to parse text from it — extracting name, email, work history, education — to auto-fill fields.
- That parsed data goes into a searchable database.
- A recruiter sees your application in a pipeline view, usually sorted by date.
No score. No auto-reject. No algorithm deciding your fate.
What does happen: a human recruiter (or sometimes a hiring manager) opens the pipeline, filters or searches, and decides who to look at. That's the gate you need to get through.
The part that is real: parsing
Here's what the keyword-anxiety crowd gets half-right. Parsing matters.
If your CV is a fancy two-column design with icons and graphics, the parser might extract your job title as "Berlin" and your company as "Senior Product Manager." When the recruiter searches their database for "Product Manager at SAP," you won't come up, because the parser never correctly captured that.
That's not the ATS rejecting you. That's the ATS failing to index you correctly, so searches miss you.
This is the actual risk, and it's preventable.
A concrete example
I ran a test last year with a CV a job seeker sent me. It was a beautifully designed two-column PDF from a popular template site. Her job at Zalando from 2021-2024 was the most recent.
I uploaded it to Greenhouse's parser (the one that auto-fills new candidate fields). The result:
- Name: correctly extracted.
- Email: correctly extracted.
- Current company: parsed as "Berlin, Germany."
- Current role: parsed as blank.
She'd been applying for six months with zero callbacks. We rebuilt the CV as a single-column document with clean headings. Same content. Within three weeks she had three first-round interviews.
The ATS didn't reject her. It indexed her as a job-title-less candidate, so when recruiters searched "Product Manager" in their database, she wasn't in the results.
What actually filters you
Three things, in order of frequency:
- A human recruiter skimming your CV for 6-8 seconds and moving on. This is where most "rejections" happen, and it's not automated.
- Knockout questions on the application form. "Do you have a work permit for Germany?" "Do you have 5+ years of experience?" These yes/no questions are real filters, often set by the employer.
- Bad parsing causing you to be invisible in searches when recruiters go looking for candidates with specific experience.
None of these are solved by stuffing keywords at the bottom of your CV in white text. (Yes, people still do this. No, it doesn't work — and some ATSes flag it.)
The takeaway
Stop worrying about fooling a scoring algorithm that mostly doesn't exist. Start worrying about two things: whether your CV parses cleanly into a database, and whether a human reading it for 8 seconds will want to keep going.
Those are the actual gates. Everything else is noise.