How ATS Screening Actually Works: Keywords, Knockout Questions, and a Human at the End

What an ATS actually does

An applicant tracking system (ATS) is the software behind most online job applications. For screening, it does three things:

• Parses your resume into fields: name, contact details, education, employers, dates, skills. If the parser misreads the file, the rest of the process works from the misread version.
• Asks knockout questions. Work authorization, graduation year, location, willingness to relocate. An answer outside what the employer set can remove an application before anyone reads it.
• Searches and ranks by keywords. Recruiters filter or sort applicants by terms from the posting: skills, tools, degrees, titles.

A human then reviews a filtered subset. The machine doesn't pick the best candidate; it decides who gets looked at.

A worked example: our own scanner

This site has a free ATS scan on the resume page. Its first version made exactly the mistakes a literal screen makes, which is why it is a useful example.

That first version counted the most frequent words in the job description and checked which ones appeared in the resume. It scored a semiconductor resume 45 out of 100 for “missing” SAMSUNG, YOUR and SEMICONDUCTOR: the company's own name, a stopword, and a word the resume plainly contained. Its score was also clamped: it could never go below 45 or above 92, whatever the resume said.

We rebuilt it. The scan now:

• matches only a fixed list of named skills, never arbitrary words from the posting;
• matches each skill as a whole word or phrase, so “Java” is not found inside “JavaScript”, and the word “rest” on its own never counts as REST API;
• never counts the employer's own name as a skill;
• scores by coverage: each skill the posting names is weighted by how often it names it (up to three times), and the score is the share of that weight your resume covers.

The lesson generalizes. Screening is literal, bound to a vocabulary, and wrong in specific, predictable ways. It can miss a skill you have because you called it something else, and it can reward a word that means nothing.

Practical rules

• Use the posting's own words for skills you genuinely have. If it says “financial modeling” and you wrote “built valuation spreadsheets”, a literal match misses it.
• Never stuff keywords or hide white text. The person who does read your resume sees it as an attempt to game the screen, and many systems flag it.
• Tailor one version per role family (for example data analyst, software, product). It beats one generic resume sent everywhere.
• Keep the layout plain: standard section headings (Education, Experience, Projects, Skills) and a single column, so the parser doesn't mangle the file.
• Answer knockout questions accurately. A wrong answer can remove you, and a misleading one can cost you an offer later.

What a keyword score can and cannot tell you

A keyword score tells you how many of the posting's stated skills your resume shows, in the posting's words. It says nothing about the quality of your experience, how well you fit the team, or whether anyone will call you. Use it to find gaps in wording, not to judge yourself.

Try it

The free ATS scan needs no account: paste your resume and a job description to see which of the posting's skills you cover and which are missing. On the same page, Resume Tailor rewrites your bullets for one posting; with a free account you get 2 tailorings in total.

Last updated: November 7, 2026