Applicant Tracking Systems do not “read” your CV the way a recruiter does. They ingest a file, turn it into text (and sometimes structured fields), then match keywords and rules. If that extraction step is messy, everything downstream is guessing.
This guide is about that first step — what the machine actually sees — not about gaming a score.
Parsing is lossy by design
A PDF is a drawing of glyphs on a page. A DOCX is a document object model. An ATS usually converts either into a stream of characters. Layout that looks fine to humans often collapses when that conversion runs:
- Text in sidebars, headers, or footers may move, duplicate, or vanish.
- Multi-column layouts can interleave lines (“Hospitality Manager London Led a team…”).
- Icons, logos, and text-as-images contribute nothing searchable.
- Tables used for layout (not data) scramble reading order.
None of that means “never design a CV”. It means know the trade-off, and prefer exports with a real selectable text layer over flattened images.
Round-trip is the honest test
The useful question is not “is this ATS-friendly?” as a vibe. It is:
- Take the exact file you will upload.
- Extract the text as a parser would (copy-paste from Preview, or a proper extractor).
- Compare that text to the CV you intended.
If names, dates, employers, or bullet points disappear or shuffle, the ATS is not “rejecting” you for personality — it never saw the facts.
Kandid calls this a round-trip check. We show fidelity as a score with a basis label because honesty beats a green tick. A designed template can look sharper and still score lower on round-trip; an ATS-safe single-column layout usually wins on extraction. That tension is real — we surface it on purpose.
What still helps (and what does not)
Helpful
- Single-column body copy for portal applications.
- Standard section headings: Summary, Experience, Education, Skills.
- Real job titles and employers you can defend in interview.
- Dates in a consistent format.
- Skills written as text, not icon grids.
Not helpful (or harmful)
- Hiding keywords in white text.
- Inventing tools, metrics, or employers to “match” the JD.
- Claiming AI-detector evasion as a feature.
- Stuffing the same synonym twenty times.
Keyword relevance still matters for many systems — but stuffing and fabrication create risk. Recruiters notice. Interviews expose it. Guardrails that strip unverifiable claims exist for a reason.
Headers, graphics, and “creative” PDFs
Creative portfolios are fine for design roles when humans open the file. For high-volume portal apply flows, treat the CV as a data packet first:
- Contact details in the body, not only in a logo header.
- No essential facts only inside images.
- Avoid text boxes that float outside the main flow.
If you need a designed version for networking, keep a clean ATS export for portals. Two artefacts, one career history — same truth both times.
How to check your file in five minutes
- Export the PDF or DOCX you plan to submit.
- Select all text and paste into a plain editor.
- Scan for missing phone/email, broken dates, and jumbled bullets.
- Fix layout (usually: simplify columns, remove decorative tables).
- Re-export and repeat until the paste looks like your CV.
If you want that check automated — with readability, ATS rules, and an explicit “what the ATS sees” panel — that is what Kandid’s free audit is for. It will not invent experience to raise a score.