The short version
- A failed parse does not trigger an automated rejection. Greenhouse's own documentation says a recruiter has to key your details in by hand.
- The real damage is empty and misfiled fields, which drop you out of the recruiter's search results before anyone reads a word.
- Columns are a genuine risk but not a death sentence — Textkernel's parser renders 90% of column resumes cleanly, up from 62% before it added column detection.
- OCR is not involved unless you upload a scan or a photo. A PDF exported from Word already carries a text layer.
- The five-minute test: paste your resume into a plain text file. If the job titles separate from their dates, a parser will make the same mistake.
Corporate job postings received an average of 244 applications each in 2025, according to Greenhouse's hiring benchmarks, drawn from more than 640 million applications across 6,000 companies. That is more than double the 116 per job they measured in 2022. No talent team opens every attachment by hand at that volume, so the applicant tracking system ingests each file first, extracts the text, and maps it into database fields.
Which is where the folklore starts. Search for why resumes get rejected and you will find hundreds of pages claiming the ATS silently bins you the moment it fails to read your file. That is not what the vendors document, and believing it sends people chasing the wrong fix.
What actually happens when a parser fails
Greenhouse publishes a support article specifically about unsuccessful resume parses. Its instruction to the recruiter is to manually input the candidate's details into the fields. A partial parse, it says, "will need to be manually corrected and verified." Nowhere in that article is rejection, disqualification, or scoring mentioned at all.
Workday's admin guide describes the same shape: parsing populates fields from a resume, and the system then lets a human review that data. It is a data-entry convenience with a review step, not a gatekeeper.
So the mechanism that hurts you is quieter and harder to notice. Neither vendor documents what follows next, so treat this as reasoning rather than something either company states: when extraction goes wrong, your job titles, employers, and dates land in the wrong fields or in no field at all. Recruiters filling a role search that database — by title, by skill, by years in a function. A field that is empty cannot match a query. You are not rejected so much as never returned, and no rejection email ever explains that.
“A parse failure does not reject you. It makes you invisible to the search that decides who gets read.”
There is a second-order effect worth naming. Recruiters do configure knockout questions and filters, and a field that parsed as blank can fail one. That is a real path to rejection — but the cause is the empty field, not the parser deciding you are unqualified.
Which layout choices actually break extraction
Greenhouse names the causes directly in that same article: columned layouts, tables, headers and footers, graphics and word art, contact details inside text boxes, resumes uploaded as images, and letters spaced apart. This is a first-party list from an ATS vendor, not a guess from a resume blog.
| Layout element | What the parser does with it | What to do instead |
|---|---|---|
| Two-column grid | May read straight across both columns, interleaving unrelated lines | Single column, top to bottom |
| Contact details in a header or text box | Frequently dropped, which costs you the fields a recruiter searches on | Put name, email, phone and location in the document body |
| Tables for skills or dates | Cell boundaries are lost; values merge into neighbouring text | Plain lines, one item per line |
| Skill bars, rating dots, icons | Carry no text, so they extract as nothing | Write the proficiency out in words |
| Scanned or photographed resume | Requires OCR, which some parsers run only as a paid add-on | Export a text-layer PDF or DOCX from your editor |
Here is the part most articles leave out, because it undercuts a tidier story. Modern parsers have got considerably better at columns. The resume-parsing vendor Textkernel published its own results after adding machine-learning column detection: well-rendered CVs rose from 62% to 90%, judged by human annotators across roughly 700 documents, with contact-information fill rates improving by 4 to 10 percentage points across more than 12,000 CVs. These are the vendor's figures for its own product, so read them as a direction of travel rather than an industry benchmark.
Read that in both directions. Before the fix, nearly four in ten column resumes rendered badly — the folklore had a real basis. After it, nine in ten come through clean, so "a two-column resume is unreadable" is now overstated. But the residual risk is not zero, it is entirely avoidable, and the damage lands specifically on contact information. If you want to know how a specific system treats your file, we go deeper on that in how Greenhouse and Lever parse and rank resumes.
Where OCR does and does not come into it
A common claim is that OCR skips your headers and footers. That conflates two unrelated things. Optical character recognition reads characters out of images. A PDF exported from Word or Google Docs already carries a text layer, so nothing needs recognising — the parser reads that layer directly.
Textkernel sells OCR as a separate add-on that detects scanned or photographed documents and "automatically activates only when necessary." So OCR matters in exactly one case: you scanned or photographed your resume instead of exporting it. Header and footer content gets lost for a different reason entirely — reading order and page-artifact handling during text extraction. Same symptom, different cause, and the fix is the same either way: keep load-bearing details in the body. If you are weighing formats, PDF versus DOCX covers which to send where.
See how your resume parses through a real ATS
Optimize your application materials in under 2 minutes.
Formatting rules that parse cleanly
None of this requires an ugly resume. It requires a readable structure underneath the design. These rules maximise the chance your history is captured accurately across the systems you will actually meet:
- Keep a single-column layout from top to bottom.
- Use conventional section headers — Work Experience, Education, Skills — since unusual ones give the parser less to anchor on. Greenhouse lists inconsistent section formatting among its causes of a failed parse.
- Put phone, email, and location in the body of the document, never in a header or a text box.
- Write skill levels as words rather than bars, dots, or icons.
- Export from your editor as PDF or DOCX; never submit a scan or a screenshot.
- Use a common typeface such as Arial, Calibri, or Helvetica, and avoid letter-spacing effects.
- Start from a layout that is already single-column rather than retrofitting a design-led template.
Clean parsing gets your experience into the database intact. What you put in those fields still decides whether you match the role — that is a separate discipline, covered in strategic keyword matching and in the action-context-result formula for the bullets themselves.
Test your own resume in five minutes
Open your resume, select all, and paste it into a blank plain text file. You have just done crudely what a parser does deliberately: stripped every visual cue and kept only the character stream.
Now read it. Do job titles still sit next to their dates? Did your phone number survive? Does anything appear out of order, or arrive as a run-on line stitched from two columns? Every defect you can see, a parser sees too — and the ones involving your contact details are the expensive ones.
It is worth knowing what employers themselves say about the outcome. In a 2021 Harvard Business School and Accenture study, 88% of the 2,275 senior leaders surveyed agreed that qualified high-skills candidates are vetted out because they do not match the exact criteria in the job description. Two caveats matter: that is employers reporting their own impressions, not a count of rejected resumes, and the fieldwork was conducted in early 2020. Treat it as evidence about how executives see their own filters, not as a measured rejection rate.