How ATS Really Works: Resume Scanners Explained

 

By Sidra Salman · October 26, 2026

You send off a resume you spent hours on, and nothing comes back. No rejection email, no interview, just silence. It's easy to picture a robot binning your application — and plenty of resume sites tell exactly that story, usually with a scary statistic attached.

Here's the twist: the scariest statistic isn't true, and the real story is more useful. An applicant tracking system (ATS) is mostly a filing cabinet with a search bar, not a judge. What filters you out is usually plain formatting mistakes plus blunt rules set by humans. Once you see how the machine reads your resume, fixing it takes about ten minutes — here's how, with a worked example and a self-test.

The biggest ATS myth: "75% of resumes are rejected before a human sees them"

You've seen this number everywhere. It sounds precise, which is part of the problem: one resume-format guide traced it to a 2012 sales pitch, not a published study — yet it still sits on the first page of search results, repeated by sites selling templates.

What does real evidence say? In a survey of 25 US recruiters published by ResumeGeni and HR.com, 92% said their ATS does not automatically reject resumes — the systems sort, rank, and filter for review, not quietly delete applications. Greenhouse, one of the biggest ATS platforms, says the same in its documentation.

So where do rejections come from? Knockout questions and filters humans set up: work authorization, a required license, minimum experience, or — most controversially — an employment gap. A 2021 Harvard Business School and Accenture study, "Hidden Workers: Untapped Talent," surveyed 2,250+ executives in the US, UK, and Germany: 88% believed qualified candidates get filtered out for not matching a posting's exact criteria, and nearly half screened out anyone with a gap longer than six months. The study counted 27 million Americans in this "hidden worker" group — veterans, caregivers, career changers never assessed by a person.

The truth beats the myth because it gives you something to act on: you can't beat a robot judge that doesn't exist, but you can write a resume that survives a keyword search, a formatting parse, and a ten-second skim.

How an ATS actually reads your resume

When you hit submit, a parser pulls the text out of your file and sorts it into fields: name, contact details, work history, education, skills. Recruiters then search that database like a spreadsheet — by job title, skill, location, or years of experience.

So your resume has two audiences: the parser, which only understands plain, well-ordered text, and the human, who spends seconds deciding whether to open it. A beautiful resume the parser mangles fails at step one; a plain one with no proof fails at step two.

The one automatic rejection that does exist: knockout questions on the form itself — "Are you authorized to work in the US?" or "Do you hold an active CPA license?" Answer no to a mandatory one and the system can auto-reject you. No keyword trick gets around that, but none needs to: it's a stated requirement, not a hidden trap.

What breaks the parser (and what to do instead)

Most parsing failures come from design choices that look great to the eye and read as gibberish to software:

• Two-column layouts and tables. The parser reads straight across the page, so it can stitch your job title and a date from opposite columns into one nonsense line. Use a single column, no tables.
• Headers and footers. Some systems skip them, so contact details that live only in a header may never reach the recruiter's database. Put everything in the main body.
• Graphics, icons, and skill bars. A parser can't read a photo, a logo, or dots showing "Excel: 4 out of 5" — that information vanishes. Write skills as plain text.
• Creative section headings. "Where I've Been" instead of "Work Experience" can confuse the parser's field mapping. Stick to: Summary, Work Experience, Education, Skills.
• Text boxes and unusual fonts. Text in a floating box may be invisible to the parser, and decorative fonts can garble. Use Calibri, Arial, or Georgia, and normal paragraphs.
• Scanned or image-based PDFs. A photo of a document has no text to extract. When the posting doesn't specify a format, .docx is the safest default; if it asks for PDF, send a text-based one, not a scan.

Keep dates consistent throughout, like "Mar 2022 – Present" — inconsistent formats make the parser's job harder and your history harder to trust.

Keywords: match their words, prove them with results

Keywords matter — recruiters literally search for them — but stuffing repeated terms backfires. Modern systems use semantic matching, so a wall of keywords with no context reads as manipulation, and a human spots it in seconds.

Pull 8 to 15 must-have terms from the posting: tools, certifications, and responsibilities that repeat or sit in the requirements section. Ignore filler like "team player" — nobody searches for that. Then use the employer's exact wording wherever it's true for you. If the posting says "data visualization in Power BI" and you wrote "built dashboards," and you did use Power BI, say so. If your internal title was "Growth Ninja" but the market title is "Growth Marketing Manager," use the market title, with the internal one in parentheses.

Spell out acronyms the first time — "Search Engine Optimization (SEO)" — to match both versions of a search. Attach every major keyword to proof: a number, a scope, or an outcome. "SQL" alone is a claim; "Built SQL dashboards that cut weekly reporting time by 30%" is evidence.

Never paste the job description in white text, or claim a skill you can't discuss in an interview. Hidden text is detectable and gets applications discarded, and a keyword that wins you a call you then fail costs more than the missed match did.

Worked example: one bullet, before and after

Say you're applying for a customer support role, and the posting repeats "Salesforce," "ticket resolution," and "customer satisfaction":

Before: "Responsible for helping customers and handling support tickets and stuff, working with the team to keep customers happy."

Why it fails: no exact keywords, no numbers, and "stuff" tells the parser nothing at all.

After: "Resolved 40+ Salesforce support tickets per week, lifting team customer satisfaction scores from 88% to 94% over six months."

What changed: the posting's exact terms now sit inside a real achievement with scale and a result. The parser finds its keywords, and the human skimming sees someone who did something specific. Same experience, different outcome.

Run this pass over your five strongest bullets and you've done most of the tailoring for one application. The same rules apply to teaching roles — schools run the same systems, which is why I test tools by profession, including my roundup of the best AI tools for teachers.

Test your resume before you apply: the 10-minute self-check

Don't guess whether your resume survives the machine — run this check before every application:

1. The plain-text test (2 minutes). Paste your whole resume into a plain text editor. If sections come out jumbled or missing, the parser sees the same mess. Fix the layout until it reads cleanly top to bottom.
2. The keyword gap check (3 minutes). Highlight the 8 to 15 must-have terms from the posting and search your resume for each. Any term that's missing but true for you gets added inside a real bullet, not a dumped list.
3. The honest-claims check (1 minute). Read every keyword you added and ask: could I talk about this for two minutes in an interview? Delete anything that fails.
4. The format check (1 minute). Single column? Standard headings? Contact info in the body, not the header? .docx, or PDF only because the posting asked for it? Fix any no.
5. The scanner check (3 minutes). Run your resume and the posting through a free checker like Jobscan, which returns a match score plus missing terms. Its free tier allows five scans a month and recommends 75% or higher before applying. Treat the score as a guide: one benchmark found gaps of 15 to 20 points between tools on the same resume, so pick one tool and stick with it.

A match score measures keyword overlap with one posting, not your worth, and it differs for every job. Aim for the high-70s on your tool, fix the genuine gaps, and send it.

What to do when the silence continues

Sometimes you do everything right and still hear nothing, because the filter was a human rule — a degree requirement, a location filter, a gap rule. Don't give up on tailoring; just spend it where it counts. Tailor for roles where you genuinely match the core requirements, and move on fast from the rest.

Keeping your job hunt organized helps too. If your system is scattered across sticky notes and half-remembered tabs, my tested list of the best AI note-taking apps might save you more time than another resume template.

The ATS isn't your enemy or your judge. It's a database with a search box, run by people with blunt rules. Write for the parser first — plain text, standard headings, honest keywords — then for the human who reads what survives. That's the whole trick, and it takes ten minutes.

Browse the Latest Articles hub for more across this blog.

Frequently Asked Questions

Does an ATS really reject most resumes automatically?

No. In a survey of 25 US recruiters, 92% said their ATS does not auto-reject resumes — systems sort and rank, humans set the filters and reject. The real risks are parsing failures and human rules like employment-gap filters.

Should I submit my resume as a PDF or DOCX?

Follow the posting's instructions first. With no format specified, .docx is the safest default — every major ATS parses it cleanly. If you send a PDF, make sure it's text-based, not a scanned image.

What is a good ATS match score?

Jobscan, the best-known checker, recommends 75% or higher against a specific posting. But the score measures keyword overlap with one posting, not your quality, and tools score the same resume very differently — use one tool consistently.

Can I use a creative or designed resume if I also upload a plain version?

Most forms parse only the uploaded file, even with a text box offered. Use one clean, single-column version for the ATS, and save the designed version for emailing a hiring manager or bringing to an interview.

About the Author

Sidra Salman is the founder of AI Tools Hub, where she writes practical, research-backed guides on AI tools. Every guide is tested and written in plain English for a US audience.