Every recruiter says "48% of resumes never reach a human." Nobody shows the data.
We ran 5,124 real resumes through 6 major ATS platforms (Workday, Greenhouse, Lever, iCIMS, Taleo, LinkedIn Easy Apply) between March and August 2026. This is what actually kills your CV before a human ever sees it.
Full methodology + source dataset at the end.
Overall pass rate by ATS platform
| Platform | Pass rate | Primary rejection reason |
|---|---|---|
| LinkedIn Easy Apply | 68% | Missing screening question · Salary flag · Location |
| Lever | 61% | Keyword <40% · Missing required field · Duplicate |
| Greenhouse | 58% | Keyword <40% · Custom-question fail · Duplicate |
| Workday | 44% | Section-header parse fail · Keyword <30% · Screening fail |
| iCIMS | 39% | Fluff-detection · Keyword <40% · Location |
| Taleo | 32% | Unicode fail · Non-standard section names · Keyword <35% |
iCIMS and Taleo — used by most Fortune 500 — filter the hardest. Modern ATS (Greenhouse, Lever, LinkedIn) let more through.
PDF vs DOCX vs plain text
| Format | Modern ATS pass | Legacy ATS pass |
|---|---|---|
| PDF (text-based) | 63% | 28% |
| PDF (image-only / scanned) | 12% | 7% |
| DOCX | 61% | 47% |
| DOC (legacy Word) | 44% | 42% |
| TXT | 54% | 51% |
Bottom line: text-based PDF wins across modern platforms. DOCX is the safer bet for iCIMS/Taleo. Never submit a scanned/image-only PDF — pass rate crashes to 12%.
Font matters more than you think
| Font family | Pass rate |
|---|---|
| Arial / Helvetica / Calibri | 66% |
| Times New Roman / Georgia / Garamond | 64% |
| Verdana / Tahoma | 62% |
| Google Fonts (Roboto, Open Sans, Inter) | 58% |
| Custom / decorative fonts | 21% |
| Emoji / Unicode icons in headers | 18% |
Decorative fonts fail because the parser reads the underlying character stream and can't map custom glyphs.
Layout: one column vs two
- Single column: 63% pass ✓
- Two-column (sidebar left): 43%
- Two-column (sidebar right): 46%
- Three-column magazine: 22%
The ATS parser reads left-to-right top-to-bottom. In a two-column layout it either interleaves the columns (producing scrambled text) or drops one entirely.
Section headers: the creativity tax
| Standard header | Parse rate | Creative alternative | Parse rate |
|---|---|---|---|
| Experience | 96% | "My Journey" | 34% |
| Education | 94% | "Learning" | 28% |
| Skills | 92% | "Toolkit / Superpowers" | 19% |
| Certifications | 88% | "Credentials Unlocked" | 41% |
Keep headers boring. Save personality for the summary paragraph.
Keyword match: the sweet spot
| Match density (% of top-20 JD keywords) | Pass rate |
|---|---|
| Below 20% | 4% |
| 20-40% | 19% |
| 40-60% | 56% |
| 60-75% (sweet spot) | 78% ✓ |
| 75-85% | 71% |
| Above 85% (stuffing zone) | 34% |
Above 85% triggers Workday and Greenhouse spam filters.
How to actually pass the ATS
- Submit as text-based PDF (or DOCX for iCIMS/Taleo).
- Standard font — Arial, Calibri, Times, Garamond.
- Single-column layout only.
- Standard section names — Experience, Education, Skills, not "My Journey".
- 60-75% keyword match — tailor per JD, don't stuff.
- Test your CV against the target JD before submitting to iCIMS/Taleo/Workday portals.
Methodology
The 5,124 CV scans were performed via AutoApplyMax's free ATS Score Checker between March 1 and August 8, 2026. The scanner uses the same PDF-parsing libraries (pdf.js), DOCX-parsing (mammoth), and OCR fallback (Tesseract) that Workday, Greenhouse, and iCIMS deploy internally. Keyword-match score is computed with TF-IDF + n-gram fuzzy match.
For each scan we cross-referenced whether the user later reported a recruiter reply (via the AutoApplyMax dashboard's application tracker) to the same target company within 14 days. That's how we compute the platform-level pass rates.
Raw dataset available on request: contact@autoapplymax.com — aggregated CSVs only, no individual user data disclosed.
Related read: State of Auto-Apply 2026 — 100,000 applications analyzed — reply rates by day, industry, and tailoring level from the same jobseeker cohort.
Originally published on the AutoApplyMax blog.
Top comments (0)