How AI resume screening really works: myths vs facts

Few parts of job hunting attract more rumours than resume screening. You'll hear that robots reject most resumes before a human sees them, that white-text keywords trick the system, and that one wrong font sinks you. Some of this has a grain of truth. Most of it doesn't. Here's what is actually going on, and where to spend your effort. For formatting specifics, see our guide to an ATS-friendly resume.

What an ATS is

An applicant tracking system is the software a company uses to receive, store and manage applications. It turns your resume into structured fields (name, job titles, dates, skills), lets recruiters search and filter, and moves candidates through stages. Many now add AI features: ranking candidates against the job, summarising resumes, or suggesting matches.

Myth: 'Most resumes are rejected by a robot'

Fact: in most systems, a recruiter sets up the process and makes the decisions. Automatic rejections usually come from knockout questions you answered on the form, such as 'Do you have the right to work in this country?' or 'Do you hold this licence?'. The more common problem is not rejection but invisibility. When a recruiter searches or sorts a long list, resumes that are missing the key terms, or that parsed badly, simply sit lower down or don't appear.

Myth: 'Hidden white-text keywords beat the system'

Fact: parsed text is plain text. Hidden keywords show up in the recruiter's view of your resume, and some systems flag them. Even when they don't, a resume that ranks well and then reads badly gets dropped at the human step. It's a trick that costs more than it gains.

Myth: 'You need a 100% match score'

Fact: match scores in checkers and in some ATS products are estimates, built differently by each vendor. A strong application often lands well below 100%, because job descriptions mix essentials with wish lists. Cover the essential skills you genuinely have, using the posting's wording, and stop there.

Myth: 'PDFs can't be read'

Fact: modern systems read text-based PDFs well. What they can't read is a scanned image of a document, text inside pictures, or content in headers and footers in some cases. If you can select and copy the text in your PDF, a parser can usually read it. Send Word only when the application asks for it.

Fact: layout still matters

• Tables and two-column templates often parse out of order, so your job titles end up attached to the wrong dates.
• Unusual section headings ('Where I've been') may not be recognised as Experience.
• Icons for phone and email can replace the text the system is looking for.
• Dates in a consistent format (for example 'Mar 2023 – Present') parse more reliably.

Fact: AI ranking reads meaning, not just words

Newer screening tools compare the meaning of your experience to the job, not just exact keyword matches. That's good news if you've been forced to keyword-stuff: clear, specific bullets about what you did and the result tend to rank well. 'Reduced invoice errors by fixing the supplier data import' is easier for both a model and a human to understand than 'Responsible for finance processes'.

Fact: humans still make the call

In many places, rules on automated decision-making and AI in hiring are tightening, and employers are expected to keep humans responsible for hiring decisions. Practically, that means your resume has two readers. Write for the person and format for the software, and you've covered both.

What to do before each application

1. Read the posting and note the three to five skills that are clearly essential.
2. Make sure each one you genuinely have appears in your summary or a bullet, in the posting's own words.
3. Run the ATS resume checker with the job description to catch missing terms and parsing problems.
4. Apply early, while recruiters are still actively reviewing.