Study · 2026-10-09
How we test our resume templates
An applicant tracking system can only read what it can extract from your PDF. So we export real PDFs from every Resumity template and read them back with open-source text extractors, then check whether the name, contact details, section headings and every bullet come back intact and in order.
Result
21 PDFs (3 sample resumes × 7 templates), each read by 3 extractors: 63 extractions.
- Name, email and phone recovered
- 100%
- Section headings found
- 100%
- Bullets intact
- 100%
- Jobs in the right order
- 100%
- Bullets in the right order
- 100%
- Garbled characters
- 0
What this does and does not show
- It shows that the text layer of our PDFs is complete and in reading order for three widely used open-source extractors.
- It does not test commercial applicant tracking systems such as Workday or Taleo. Each employer configures its own system, so no template can promise how a particular employer will read it. Our ATS guides summarise what those vendors document.
- The sample resumes are synthetic and in English. All seven templates are single-column layouts.
- It checks that text is extracted, not whether a recruiter likes the resume.
Results by template
Method
- Three synthetic resumes: a one-page mid-career designer, a two-page senior engineering manager, and a one-page graduate with an internship.
- Each one exported to PDF in all seven templates by the same exporter users download from, running in headless Chrome 154.0.8037.98.
- Text extracted from every PDF with pdf.js 4.10.38, pdftotext (Xpdf) 4.00, pypdf 6.13.0.
- Scored against the source data: name, email and phone (phone digits in order, any separators); the section headings the template drew; a bullet counts as intact when at least 95% of its four-word sequences appear; order compares consecutive jobs and consecutive bullets.
Data and how to reproduce it
Every extraction is in the raw results (JSON), and the 21 PDFs are published next to it (for example senior-two-page__daniel.pdf). The study script is scripts/research/template-parse-study.mjs (run at commit 1a14a0c). Two runs give identical results. Our automated test suite also extracts the text of every template's PDF.