How ATS Systems Actually Work

I spent 12 years at ZipRecruiter, the last several as VP of Product Design. In that time I watched the hiring technology stack evolve from simple database searches to the complex ATS ecosystem that exists today. The single most important thing I learned: the gap between what a candidate submits and what a recruiter sees is larger than almost anyone realizes.

When you upload a resume to an ATS, the system doesn't store your PDF and let recruiters read it. It parses your document — converting formatted text into structured data fields: name, email, phone, work history entries, education entries, skills. This parsed data is what recruiters actually search, filter, and evaluate. The original document becomes a fallback, something a recruiter might open later if your parsed profile looks interesting.

The problem is that parsing is an imperfect process. Every creative formatting choice — columns, tables, text boxes, headers, footers, infographic elements — introduces ambiguity. The parser has to guess where one field ends and another begins, which text belongs to which job entry, whether a line is a skill or a job title. When it guesses wrong, the recruiter's view of your resume is a scrambled version of what you intended.

The most common parsing failure isn't rejection — it's corruption. Your resume passes through, but the structured data is wrong. Your job title from Company A gets attached to Company B. Your dates are off by a year. Your skills section gets merged with your summary. The recruiter sees a profile that doesn't make sense, and moves on.

At ZipRecruiter, we saw this at scale. Millions of resumes processed through various parsing engines, and the error patterns were consistent:

• Multi-column layouts cause the parser to read across columns instead of down them, interleaving unrelated content
• Tables lose their cell relationships — dates, titles, and companies get shuffled
• Text boxes and floating elements get extracted in unpredictable order
• Headers and footers get concatenated with body content
• Creative section headers ("Where I've Made an Impact" instead of "Work Experience") aren't recognized as standard sections

These aren't edge cases. At the volume of a major job marketplace, we saw them in a significant percentage of submissions.

Not all ATS platforms handle parsing and search the same way. After years of working adjacent to these systems, and later researching them for ResumeGeni's ATS comparison guide, the differences became clear:

Workday

The most widely deployed enterprise ATS. Workday's parsing is competent on standard formats but struggles with non-standard section headers and multi-column designs. Its internal candidate search is keyword-based, not semantic — if the job description says "Kubernetes" and your resume says "container orchestration," you won't surface in the recruiter's search. I wrote a detailed breakdown of how Workday processes resumes based on testing against the actual system.

Greenhouse

Greenhouse is fundamentally different from the others: it doesn't auto-reject resumes. Every application reaches a human reviewer. There's no match score. Instead, recruiters use structured scorecards and Boolean search across parsed fields. This means formatting still matters — not for passing an algorithm, but for being findable when a recruiter searches for specific skills. The full picture is in the Greenhouse resume guide.

Taleo

Now Oracle Recruiting Cloud, Taleo is the strictest parser I've encountered. It's aggressive about requiring specific file formats, breaks on tables and columns more than any other system, and has rigid expectations about section order and labeling. Many government and Fortune 500 employers still run Taleo. If you're applying to a large, established enterprise, read the Taleo-specific rules.

iCIMS

Widely used in healthcare, retail, and mid-market employers. iCIMS has its own parsing quirks, particularly around how it handles iframe-based job listings and candidate data extraction. It's more forgiving than Taleo but less sophisticated than Greenhouse in how it presents candidates to recruiters. Details in the iCIMS guide.

Lever

Popular with startups and tech companies. Lever uses an "opportunity-based" model where recruiters see candidates in context — attached to specific roles with notes, feedback, and history. The parsing needs to work for search, but the human-readable document carries more weight than in keyword-heavy systems like Workday. More on how Lever evaluates candidates.

The mental model most candidates have — a recruiter reading your resume from top to bottom — is wrong in most cases. What actually happens:

1. Search. The recruiter enters keywords into the ATS search bar: job titles, skills, tools, certifications. Your parsed profile either matches or it doesn't.
2. Filter. Results are narrowed by location, experience level, education, date applied. If your parsed data has errors in these fields, you're filtered out.
3. Scan. The recruiter scans the structured profile view — not your PDF. They see parsed fields: current title, company, skills list, education. This takes 6-10 seconds.
4. Open (maybe). If the parsed profile looks promising, the recruiter opens the original document. This is the first time anyone sees your actual resume formatting.

This workflow means two things for candidates: your resume needs to parse correctly (steps 1-3) before it has any chance of being read (step 4). And the keywords and structure that determine parsing success are different for each ATS platform.

After 12 years on the inside of hiring technology, three things stand out:

Your resume is a data input, not a design artifact. How it looks to you matters less than how it parses into structured data. A beautiful two-column resume that parses into garbage is worse than a plain single-column document that parses cleanly.

Know which ATS your target employer uses. Each system has different parsing behaviors. Workday wants exact keywords. Greenhouse cares about searchable fields. Taleo needs strict formatting. Optimizing for "ATS" in general is less effective than optimizing for the specific system. The ATS comparison guide breaks this down by platform.

Test before you submit. Upload your resume to a parsing test tool and look at the structured output. If your job titles, dates, skills, and company names aren't extracted correctly, fix the formatting before applying. This single step eliminates the most common reason qualified candidates don't get callbacks.