The short answer
Most of what circulates about applicant tracking systems is unverified, and several of the most repeated claims lead people to make their resumes worse. Here is what holds up to checking, what does not, and where the honest answer is that behaviour varies by employer.
What matters most
- The widely quoted rejection percentages have no traceable source and should not drive your decisions.
- A ranking score, where an employer uses one, orders a list rather than emptying it.
- Text-based PDFs are read correctly by mainstream systems.
- Where behaviour genuinely varies by product and employer configuration, no universal rule can be true.
Why this field breeds myths
Three conditions make a subject a myth factory: the truth is hidden, the stakes feel high, and the audience is anxious. Job hunting supplies all three. Add an industry of tools selling optimisation against a black box, and confident numbers appear from nowhere and are then cited by people who assume somebody, somewhere, measured them.
A useful habit when you meet any claim in this area is to ask what the evidence would have to look like. A statistic about how many resumes are automatically rejected would require access to the internal data of many employers running different configurations. When you cannot picture who could have measured something, treat it as a story.
The claims, one at a time
The commonly repeated percentages have no traceable origin. What actually removes people automatically is employer-configured screening questions on the application form β work authorisation, required licence, minimum experience. The document itself is rarely the trigger. Where ranking features are enabled, they order a list a recruiter still opens.
Mainstream systems read text-based PDFs without difficulty. The genuine failure is a PDF containing an image of a document rather than selectable text. Send a PDF unless the posting asks for Word, in which case send Word.
No published threshold exists, and the numbers quoted by optimisation tools are that tool's own estimate against one posting. Cover the terms genuinely central to the role, evidence them, and stop.
Colour, bold, rules, white space and standard fonts are drawing instructions and do not touch the extracted characters. The genuine risks are columns, text boxes, headers, footers and text saved as images.
Cover letters are usually stored as an attachment and read, if at all, by a person later in the process. Write it for the human, because that is who it is for.
They appear in the extracted text exactly like any other characters, recruiters have seen the trick repeatedly, and it reads as an attempt to deceive. It is detectable and it is treated as a credibility problem rather than a clever hack.
The numbers you should not repeat
Two figures dominate this topic and neither should be treated as fact. The first is a rejection rate β a specific percentage of resumes supposedly discarded by software before any human involvement. It circulates without a source, and the mechanism it describes does not match how ranking features are documented to work.
The second is a recruiter attention span, expressed as a precise number of seconds spent on each resume. Versions of this trace back to small commercial studies with methodology that does not support a universal claim, and the figure quoted drifts depending on who is repeating it. The underlying point survives without a number: a first pass is fast, and your strongest material should be at the top of page one. That is worth acting on. The statistic is not worth citing.
Myths that make your resume worse
Some folklore is inert. These four actively cost people interviews, which is why they are worth separating out.
- "Strip everything to plain text." This produces a document that parses perfectly and reads terribly, and the person deciding whether to interview you is reading it. You lose nothing mechanically by keeping clear headings, sensible spacing and a legible hierarchy.
- "Add a keyword block at the bottom." A block of comma-separated terms with no evidence behind it is visible to the human reader, dilutes the document, and invites questions you cannot answer. Terms belong inside achievements.
- "Remove dates to avoid age filtering." A resume without employment dates makes a recruiter suspicious immediately, and it damages parsing badly. Age discrimination against workers 40 and over is unlawful in the United States under the ADEA; managing which optional signals you volunteer, such as a graduation year from decades ago, is reasonable, but deleting employment dates is self-harm.
- "Apply to everything, the system decides." Volume applying produces untailored documents and worse outcomes, and it is the main reason people conclude the process is rigged. Ten considered applications beat fifty generic ones by a wide margin.
Things that are true and get dismissed as myths
Scepticism can overshoot. A few commonly repeated pieces of advice are sound, and they get thrown out along with the nonsense.
- Search really is how recruiters find people in large pools. If your document does not contain the words someone would search for, you are effectively absent. This is the honest core of all keyword advice.
- Single-column layouts really are safer. Not because anyone rejects columns, but because extraction order is unpredictable and you cannot test the employer's parser in advance.
- Application forms really do filter. This is the part of the process that genuinely automates decisions, and it is where careless answers do measurable damage.
- Tailoring really does change outcomes. Not because of a score, but because a document that speaks the language of the role reads as a better fit to both search and human.
- Applying early really does help. Not because of anything the software does, but because recruiters review in batches and shortlists fill before postings close.
How to evaluate the next claim you read
Three questions handle most of it. Who could have measured this, and did they publish how? Does it describe a behaviour that would be identical across every employer, when nearly everything in these systems is configured per employer? And does following it make the document worse for a human reader, which is a strong signal that the advice is optimising for a mechanism that does not exist.
For a non-commercial second opinion on resume conventions generally, the Purdue OWL resume writing section has no product to sell you, and the UK government's National Careers Service guidance is a useful check on how much of the received wisdom is actually about a specific national market rather than a universal rule.
And if you would rather test than argue, run your own document and a live posting through our ATS checker. Seeing the extracted text is a better cure for this whole genre of anxiety than any list of claims, including this one.
Questions people actually ask
Do applicant tracking systems reject 75% of resumes?
That figure and its variants have no traceable source, and the mechanism they describe does not match how these products are documented to work. Automatic removal generally comes from employer-configured screening questions on the application form rather than from the resume. Where employers enable ranking, it sorts a list rather than deleting anyone from it.
Is the six-second recruiter scan real?
Versions of that figure come from small commercial studies whose methods do not support it as a universal fact, and the number quoted varies by source. The useful part is uncontroversial without a statistic: a first pass is quick, so your strongest evidence belongs at the top of page one.
Should I submit a Word document instead of a PDF to be safe?
Not unless the posting asks for one. Text-based PDFs are read correctly by mainstream systems, and PDF protects your layout in transit. Follow the posting where it specifies a format, because that instruction beats any general rule.
Does a fancy template ruin my chances?
Only if it uses columns, text boxes, headers or images to hold text. Colour, typography and spacing are harmless to extraction. The test is whether you can select and copy the text out in a sensible order, not how plain it looks.
Will a higher match score get me an interview?
A match score is one tool's estimate against one posting, not a threshold anyone hires against. Use the missing-terms list to find genuine gaps you can honestly close, and ignore the number itself. Adding terms you cannot defend improves the score and damages the interview.
