Myths and evidence

Resume screening myths, with verdicts

Twelve things job seekers believe about how resumes get screened. Some are folklore repeated until it sounded like data. Some are half-right in a way that matters. Three of them are simply true, and those are the ones worth your attention.

The distrust behind these beliefs is not irrational. In a Pew Research Center survey of 11,004 US adults (2023), 71% opposed AI making final hiring decisions and 66% said they would not apply to an employer that used it that way. A process that stopped explaining itself gets explained by rumor instead — so here is each belief with what the evidence actually says, and where the evidence comes from.

Free · No sign-up · No credit card

How to read the verdicts

False
No evidence behind it, or evidence that says the opposite. Acting on it wastes effort, and in one case actively hurts you.
Partly true
A real mechanism, described wrongly. The correction usually changes what you would do about it.
True
Documented, and we are not going to talk you out of it. Two of the three have something you can do; the third does not, and pretending otherwise would be worse.

The twelve beliefs

False

“Three out of four resumes get rejected by the software before a human sees them.”

The figure traces to a 2012 sales pitch from a resume-optimization vendor that was gone by 2013, and no methodology behind it was ever published. When Enhancv interviewed 25 recruiters working across more than ten applicant tracking systems in 2025, nearly all of them said their system does not automatically reject anyone on the content of a resume — and most said they hear this claim from candidates constantly. Tracking systems rank, search and store. They do not form an opinion of your bullet points.

What to do instead

Spend the worry somewhere it pays: on whether your file extracts into readable text at all, which is a real failure and a visible one.

what a parser actually does with your file
False

“Paste the keywords in white text and you slip straight through.”

Parsing keeps the characters and throws away the display layer — color, size, position, all of it. Your hidden block arrives in the recruiter's view as plainly visible pasted text sitting in the middle of nothing, which reads exactly as it is. Vendors now ship detection built specifically for this trick. It is the one piece of advice in this category that can convert a maybe into a permanent no.

What to do instead

If a skill is worth smuggling into the margin, it is worth a line in the role where you actually used it.

False

“A PDF gets thrown out — you have to send a Word file.”

Text-based PDFs are read fine by current parsers, and plenty of employers ask for PDF precisely because the layout survives. The real failure is narrower and much easier to miss: a PDF that contains a picture of a resume rather than a resume. Design-tool exports, scans, and text baked into a graphic have no characters in them to extract, so what reaches the recruiter is a blank or a fragment.

What to do instead

Send the format the posting asks for, then confirm your file is text rather than an image of text — reading the extracted version takes ten seconds and settles it.

Partly true

“A robot reads my resume and decides.”

The tracking system itself is a database with a search box, and searches are run by people. What is genuinely automated sits on top of it: in a survey of 948 hiring decision-makers (ResumeBuilder, October 2024), roughly half said AI handles rejections at the initial screen, and 16–21% said it does so at every stage with no human review. That is an employer switching something on, not a property the software has by default — which is why the same resume meets a wildly different process at two companies.

What to do instead

Write for a skeptical human and a literal machine at once. They want the same thing: your claims in plain words, next to evidence.

Partly true

“I have to hit a keyword match percentage.”

The words are real. Recruiters search their database by skill, title and certification, so a resume that never uses the posting's vocabulary is genuinely harder to surface. The percentage is not real. No employer publishes a threshold, nobody outside the company can see what their system scored you, and any match rate you are shown was calculated by the tool showing it, against a formula it invented. Writing to raise that number is how resumes end up stuffed — and stuffing repels the person who does the actual rejecting.

What to do instead

Use the posting's own wording for skills you can already evidence somewhere in your experience, and ignore every percentage, including one you paid for.

Partly true

“Columns, tables and graphics break everything.”

They break reading order, which is a different problem from rejection. A two-column layout is often read cell by cell, so a sidebar of skills can arrive detached from the jobs that prove them, and dates can land next to the wrong employer. Text inside an image does not arrive at all. The resume is not discarded; it competes as a scrambled version of itself, which is worse in a quieter way, because nothing tells you it happened.

What to do instead

Keep anything load-bearing in one linear column, and check the extracted text before trusting a template you liked the look of.

Partly true

“Recruiters only look at a resume for six seconds.”

The famous number comes from eye-tracking sessions with about 30 participants, run in 2012 and again in 2018 — we quote the later study on our own home page, and it measures a first triage pass, not the whole read. The shape of the finding is solid and it is the reason this product exists: the opening pass is keep-or-kill and it lasts seconds. What is wrong is treating the figure as an attention budget for your entire resume, when the resumes that survive triage go on to get minutes of real reading.

What to do instead

Make the top third do the surviving — target role, current evidence, the numbers you would want read first.

test your own top third against the skim
Partly true

“Everyone uses one of these systems, so this applies everywhere I apply.”

At the top of the market, assume it: about 98% of Fortune 500 companies run an applicant tracking system, counted by checking 492 of the 500 (Jobscan, 2023). Below that it thins out fast — a great many small and mid-size employers have nothing between your attachment and a hiring manager's inbox, and AI scoring layered on top of a tracking system is still a minority practice rather than the default everyone assumes.

What to do instead

Keep the formatting discipline everywhere, because it costs nothing and helps a human too. Drop the assumption that an algorithm is always in the room.

Partly true

“A gap in my dates gets me filtered out automatically.”

The Hidden Workers study from Harvard Business School and Accenture (2021) found gap filters of six months or more configured at more than half of the companies surveyed, and 88% of employers agreeing that their own screening rejects qualified people. Two things temper it: that research is from 2021 and no one has repeated it since, so treat it as evidence the practice exists rather than a current rate — and the filter is a setting a person chose, which means it only ever sees an unexplained hole in a timeline.

What to do instead

Put one factual line where the gap sits, or a dated entry for what you were doing. Explained time is not a hole.

the one-line versions that work
True

“One wrong answer on the application form can disqualify me instantly.”

This is where automatic rejection actually lives. Knockout questions — work authorization, licenses, shift availability, minimum salary, years in a field — are configured to route a disqualifying answer straight to a rejection, no human involved and no appeal. It is the one mechanism recruiters across systems consistently confirmed when Enhancv interviewed them in 2025. Your resume is not consulted, because the form already answered the question.

What to do instead

Read the form questions before you spend an hour on the document, and answer them accurately. A convenient answer is caught at offer stage, where it costs far more than the rejection would have.

True

“I could be interviewed and scored by an AI without being told.”

In Greenhouse's candidate survey (Greenhouse candidate survey, May 2026, n=2,950), 63% of job seekers said they had been through an AI interview, 70% of them were not told in advance, and only 26% believe AI evaluates candidates fairly. Those are candidate reports rather than a count of employers, but the direction is not in dispute: this is now a normal part of the process and it is frequently undisclosed. Being uneasy about it is an accurate reading of the situation, not paranoia.

What to do instead

Ask what the interview format is before you accept the slot. Wanting to know who or what is judging you is a reasonable question, and the answer tells you something about the employer either way.

True

“Screening tools can be biased against people like me.”

A study (University of Washington, peer-reviewed, 2024) had language models rank more than 550 resumes against real job postings and found white-associated names preferred 85% of the time, against 9% for Black-associated names. The EEOC settled a case against iTutorGroup for $365,000 in 2023 over software configured to reject older applicants — a settlement, so that one is established. Mobley v. Workday, an age-discrimination collective action preliminarily certified in May 2025, is an allegation currently being litigated, not a finding of fact.

What to do instead

Nobody can tell you whether any of this touched your application, and a tool that claims to detect it is guessing. This one is a policy problem rather than a personal verdict — worth knowing, not worth reading your rejections through.

What you can actually control

Most of the list above is either someone else’s configuration or somebody else’s rumor. Three things are yours, and two of them you can verify in the next few minutes.

The text that actually comes out of your file

Not the layout you designed — the characters a parser extracts from it. This is the one failure on the whole list that is invisible to you and completely fixable, and it takes one look to rule out.

how the parse pass works

What survives the first few seconds

Triage happens whether or not the number attached to it is precise. You can find out what a stranger retains from your top third before an employer does it for you.

run the skim test

The answers you give on the form

The genuinely automatic rejection is the one people skim past to get to the resume upload. Slow down on those five questions; they carry more weight than the next hour of editing.

Check yours against the first two

The check reads your file the way a parser does, shows you the text it got, and names every skim risk in what a first pass would reach. No prediction about whether you will get an interview, because that is not knowable — just the specific things in the document that are working against you.

Check my resume — free

Screening myths — common questions

Does an applicant tracking system reject resumes automatically?

On the content of your resume, almost never — recruiters interviewed across more than ten platforms in 2025 said their systems do not do this (Enhancv). On application-form answers, yes: knockout questions about authorization, licensing or availability route a disqualifying answer to rejection with nobody reading it. And some employers switch on AI screening above the tracking system, which about half of 948 hiring decision-makers reported at the initial screen (ResumeBuilder, October 2024). The distinction matters because it changes what you would fix.

Is AI hiring biased?

There is measured evidence that it can be. A peer-reviewed 2024 study from the University of Washington found language models ranking resumes preferred white-associated names 85% of the time against 9% for Black-associated names, and the EEOC settled a $365,000 case in 2023 against an employer whose software was configured to reject older applicants. A separate age-discrimination case against Workday was preliminarily certified in May 2025 and remains an allegation. What no tool can tell you is whether bias affected a particular application of yours.

Do recruiters really only look at a resume for six seconds?

The first pass is that quick, and the precise number is shakier than it sounds — it comes from eye-tracking with roughly 30 participants. Read it as a description of triage rather than a stopwatch: seconds to decide keep or kill, then minutes of real reading for whatever survives. The useful version of the finding is what it implies about your top third, not the digit.

Can recruiters tell if I used AI to write my resume?

They can tell when a resume reads like nobody in particular. In a May 2025 survey of 600 hiring managers (TopResume), 33.5% said they can spot an AI-written resume in under 20 seconds — self-reported confidence that has never been tested blind, and two-thirds made no such claim. Independent audits through 2025 and 2026 found no major applicant tracking system running AI detection at all. What gets penalized is generic writing, which is a property of unedited output rather than of the tool that produced it.

What happened when we actually pasted resumes into a chatbot
Menu