Let's kill the biggest myth first. An ATS does not read your CV, decide you are unworthy, and trash you before a human sees the file. That is not how the software works, and it has never worked like that at Workday, Greenhouse, Lever, Taleo, SAP SuccessFactors, or iCIMS.
What actually happens is more boring and more useful to know. The ATS parses your PDF or Word file into structured fields, stores those fields in a database, and lets a recruiter search and filter that database. The recruiter is the one who rejects you. Or, more often, who never sees you because your CV parsed badly and your name does not show up when they search for "senior data analyst, SQL, Madrid".
Once you accept that distinction, everything about CV writing for ATS changes. You stop trying to trick a robot. You start helping a parser do its job.
What an ATS actually is
Think of an ATS as three pieces glued together. First, a parser that converts your file into text and tries to assign each chunk to a field: name, email, company, job title, dates, skills. Second, a database that stores those parsed fields for every applicant. Third, a search and filter layer the recruiter uses to find candidates.
There is no scoring committee inside the machine. Some platforms add a relevance score, but that score ranks candidates against the job description for the recruiter to review. It does not delete anyone. The decision to discard a CV is human, even when it feels automated because the recruiter has 400 applications and 90 minutes.
The interesting consequence: a clean parse beats a clever design every single time. If your job title gets stored as "PROFESSIONAL EXPERIENCE" because the parser confused your section header with a role, you do not exist for any keyword search. That is the real failure mode, and it has nothing to do with AI gatekeeping.
The most-used ATS in 2026
The market is concentrated. Workday dominates large enterprises and is the system you meet most often when applying to Fortune 500 companies. Greenhouse and Lever own the tech and scale-up segment. Taleo and SAP SuccessFactors still run a huge chunk of legacy enterprise hiring, especially in banking, pharma, and energy. iCIMS sits in the middle, big in retail and healthcare in the US.
They are not identical. Workday is famously strict about parsing tables and unusual layouts. Greenhouse is more forgiving. Taleo has the oldest engine of the bunch and punishes anything fancy. But the differences matter less than the shared logic: all of them index your CV into searchable fields, and all of them reward files that look like a document, not a poster. For a deeper breakdown of which company uses what, see the most used ATS in 2026 and what that means for your CV.
How parsing actually works
The parser walks through your file top to bottom and tries to recognize patterns. Dates with a hyphen between them look like an employment range. A line in larger or bolder text near a date looks like a job title. An indented block under that line looks like responsibilities. An email address pattern looks like contact info.
This is why layout choices have outsized consequences. Two-column CVs scramble the reading order, so the parser sometimes glues the right column onto the left in odd ways. Tables get flattened into walls of text where job title, company, and date no longer line up. Text inside images is invisible. Fancy fonts can break character encoding so "é" becomes a square. Section headers buried inside graphics never register.
If you want the deep version with examples, ATS parsing rules: how fonts, columns, and headers decide whether your CV gets read walks through each rule. And if you are still using a sidebar layout, why that two-column CV is costing you interviews is the post to read next.
Keyword matching, not keyword stuffing
Here is where most ATS advice goes wrong. You will read that you need to "match keywords" with the job ad. True. You will also read that you should stuff every variant of every term into your CV. False.
Recruiters typing into the search bar use natural phrases. "Python", "AWS", "team lead", "healthcare". They use boolean filters in maybe 5% of cases. So your CV needs to contain the real terms the job uses, in the places a parser will find them: inside your job titles, inside the bullet points under each role, in a skills block. Keyword vomit at the bottom of the page in white-on-white text was a trick that died around 2015. Modern ATS strip formatting and recruiters laugh at the result.
The right move is targeted alignment. If the job ad says "SQL" three times and you have written "databases", change one mention to SQL. If they say "product manager" and your title was "product owner", note both. Read ATS keyword optimization without keyword stuffing for the full method.
PDF or Word: the honest answer
For 80% of platforms in 2026, a PDF exported from a normal word processor parses fine. The remaining 20% includes some older Taleo and SuccessFactors setups that prefer .docx. The actual problem is rarely the format itself, it is what is inside the file. A Word document with text boxes parses worse than a clean PDF. A PDF made of scanned images parses to nothing.
When the job application form gives you a choice, .docx is the safer default for legacy systems. When you cannot tell, ship a PDF generated from real text, not a screenshot.
What the recruiter actually does with the output
The recruiter opens the ATS, types a query, and reviews a list. They look at names, current job titles, current employers, and a few lines of summary. They open maybe 15 to 30 CVs out of the search results in detail. They decide who goes to a phone screen.
Your CV does not need to be perfect for an AI. It needs to be readable to a human in 20 seconds and findable in the database when the recruiter searches. Those two goals overlap more than people think.
If you want the full reverse-engineering of why CVs get filtered out, the 9 reasons that actually matter in 2026 lays them out, ranked by how often each one matters.
How to test your own CV
Do not guess. Run your CV through a parser and read what it sees. Free tools like Resume Worded, Jobscan, and the parser preview built into many job boards will show you the extracted fields. If your name lands under "experience" and your most recent job title is missing, you have a parsing problem before you have a content problem.
How to test your CV with an ATS parser walks through which tool to trust and how to read the output without panicking. The first time you do this, expect surprises.
Putting it together
An ATS is a filing cabinet with a search bar. Treat your CV as a file the parser will index for a human reader. Match the language of the job ad without stuffing. Use a single column, plain text, standard section names, real fonts. Test before you submit.
The deeper this gets, the more it touches the rest of your career stack. How to write a CV covers the content side. LinkedIn profile optimization covers the channel recruiters actually find you on before they ever open the ATS.
If you want to skip the manual rewrite, Postulit pulls your LinkedIn profile and generates an ATS-friendly CV in one step, with the parsing rules above already baked in. Either way, the action is the same: stop trying to outsmart the system. Help it read you.