ATS Algorithms, Parsing & Keyword Scoring

Applicant Tracking Systems (such as Greenhouse, Lever, Workday, and Ashby) do not use mysterious AI to randomly reject candidates. Instead, they use natural language parsers (e.g., Sovren, Textkernel, Daxtra) to convert raw PDF/DOCX text into structured entity graphs. The system extracts: 1) Job Titles and Seniority, 2) Hard Technical Skills and Frameworks, 3) Years of Relevant Experience, and 4) Education / Certifications. When a recruiter opens a requisition dashboard, candidates are sorted or filtered by keyword match percentages and knockout criteria. Resumes with exact-match terminology (e.g., "PostgreSQL" vs "SQL databases") and standard section headers score in the top 10% of the recruiter queue.

Enterprise companies (e.g., Salesforce, Walmart, Amazon) use Workday ATS, which requires account creation, strict parsing into standardized multi-step forms, and automated knockout questions. High-growth tech scaleups (e.g., Stripe, Figma, Airbnb) use Greenhouse or Lever, which feature single-page frictionless forms, recruiter scorecards, and candidate tags. Modern AI-native startups increasingly choose Ashby, which offers automated candidate sourcing analytics and structured interview loop tracking. Tailoring your application format to the specific ATS provider prevents parsing corruption.

When configuring a job opening, recruiters set mandatory criteria called "Knockout Rules." Common examples include: "Do you now or in the future require visa sponsorship?", "Do you have at least 5 years of professional experience in Go/Python?", and "Are you located within commuting distance of San Francisco?". If a candidate selects an answer that does not match the recruiter’s required threshold, the ATS automatically moves the candidate to the "Auto-Reject" or "Archived" stage without human recruiter review. Always answer truthfully, but ensure you understand the exact requirement parameters stated in the job post.

In 2026, all major modern ATS platforms (Greenhouse, Lever, Ashby, Workday, SmartRecruiters, iCIMS) natively parse standard vector PDFs with 100% accuracy. The only time a PDF fails is when it is exported as a flattened raster image (scanned image) or created using complex graphic design tools (Canva, Photoshop) with multi-layered text boxes. A clean, single-column PDF generated from Google Docs, Microsoft Word, or LaTeX preserves exact typography, clickable links, and clean text flow for both the parser and the human hiring manager.

A persistent internet myth claims that pasting job descriptions in 1pt white font in the footer bypasses ATS filters. In reality, ATS parsers extract the raw unicode stream and discard CSS colors, font sizes, and opacity. When the recruiter views the parsed candidate summary, the hidden keywords appear as a massive block of unformatted gibberish at the bottom of the profile. Recruiters recognize this as keyword manipulation and immediately blacklist the applicant. Instead, naturally integrate target technologies into quantifiable bullet points describing real projects.