The advice that PDFs break applicant tracking systems was true in about 2012 and has been repeated ever since.
Where the advice came from
Early applicant tracking systems genuinely struggled with PDF. Text extraction was less reliable, multi-column layouts scrambled, and some systems simply rejected the format.
That was fixed years ago. The major platforms now use mature extraction and handle a normal text-based PDF as well as they handle a Word file — often better, because a Word file can reflow differently depending on the fonts installed on the parsing machine.
The advice persists because it is repeated by career sites copying each other, which is the same dynamic that keeps a lot of outdated document advice circulating.
The exception that is real
One kind of PDF genuinely does fail, and completely.
A CV exported as images — from a design tool set to rasterise, from a phone scan, or from a "print to image" workflow — has no text layer. An applicant tracking system extracts nothing and files an empty record. You have applied with a blank document and received no error.
- Test it in ten seconds: open your PDF and try to select a line of text. If a rectangle is drawn instead of characters highlighting, it is an image.
- Also test: Ctrl+F for your own surname. No result means no text layer.
- Canva, Figma and similar can export either way — check the setting rather than assuming.
This is the actual risk behind the folklore, and it applies equally to a Word file with the content in a text box or a picture. The format is not the problem; a missing text layer is.
What actually parses badly
| Element | Parses? | Note |
|---|---|---|
| Single-column layout | Reliably | The safest structure |
| Two columns | Often scrambled | Reading order is inferred |
| Text in headers and footers | Often ignored | Never put contact details there |
| Tables for layout | Unpredictable | Cells may be read out of order |
| Icons instead of labels | Lost entirely | An envelope icon is not "email" |
| Graphics and charts | Ignored | Any skill rating bar conveys nothing |
| Standard section headings | Reliably | "Experience", "Education", "Skills" |
The headers and footers row is the one that costs people interviews. Contact details placed in the page header are frequently not extracted, so the system has a parsed CV with no way to contact the candidate.
A CV that survives both readers
It has to parse for a machine and read well for a person, and those wants are compatible if you keep it simple.
- One column. Two columns look designed and interleave when parsed.
- Contact details in the body, at the top of the page, not in the header area.
- Conventional section headings. "Work Experience", not "My Journey".
- Real text for everything, including dates and contact details. No icons carrying meaning.
- A common font, embedded. The embedding setting is in PDF font problems.
- Name the file properly —
Surname-Firstname-CV.pdf. It is what a recruiter sees in a folder of two hundred.
File size
Some portals cap uploads at 1 or 2 MB, which a CV should never approach. A two-page text CV is well under 200 KB.
If yours is several megabytes, something is wrong: an uncompressed photograph, an unsubsetted font family, or a design tool embedding assets at print resolution.
Compressing at 150 DPI fixes it, and if the file is mostly text the saving will be small because it was never the problem — in which case check the fonts. Zipping is not a fix: a PDF is already compressed internally, so ZIP saves nothing, and a zipped CV is one extra step between a recruiter and reading it.
Follow the form
All of the above is subordinate to one rule: if the application specifies a format, send that format.
A system asking for .docx may be feeding a parser configured for it. Sending a PDF because it is generally better is an unforced error at the one stage where you cannot afford one.
Where both are accepted, PDF is the better choice — it renders identically for the human reader, cannot reflow, and cannot be accidentally edited. Where nothing is specified and you are emailing a person directly, PDF, every time.
Frequently asked questions
Can applicant tracking systems read PDFs?
Yes, reliably. The advice against PDF dates from around 2012 and has been repeated since. Modern systems parse text-based PDFs as well as Word files, often better — a Word file can reflow differently depending on the fonts on the parsing machine.
When does a PDF CV actually fail?
When it has no text layer — exported as images from a design tool, scanned, or printed to image. The system extracts nothing and files an empty record, with no error shown. Test by selecting a line of text; if a rectangle appears instead, it is an image.
Should my CV be one column or two?
One. Two-column layouts look designed and interleave when parsed, because extraction infers reading order from position. The same applies to using tables for layout, where cells can be read out of order.
Why should contact details not go in the header?
Because page headers and footers are frequently ignored during extraction. The result is a parsed CV with no contact details attached, which is the most expensive formatting mistake available. Put them in the body, at the top.
How small should my CV file be?
Under 200 KB for a two-page text CV. Some portals cap at 1 MB. If yours is several megabytes, the cause is an uncompressed photo, an unsubsetted font, or print-resolution assets from a design tool — not something zipping will fix.
PDF or Word — which should I send?
Whatever the application asks for; that overrides everything. Where both are accepted, PDF, because it renders identically for the human reader and cannot be accidentally edited. Where nothing is specified, PDF.