Understanding Your ATS Score and Its Limits

An ATS score looks like a measurement and behaves like an opinion. Knowing which parts of it are real is what stops you rewriting a perfectly good resume to satisfy a number nobody at the company will ever see.

The score is not coming from the employer

This is the most consequential misunderstanding in the category, so it is worth being blunt about.

When you run a scoring tool, you are using a third-party product that models what an applicant tracking system might do. It has no connection to the employer you are applying to. It cannot read their configuration, their required fields, their knockout questions, or their shortlist. The number it returns is that vendor’s estimate against that vendor’s own rules.

Employers do run their own systems, and those systems do produce internal signals. You will never see them, and they are set up differently at every company. So a score is best understood as a rehearsal, not a result.

What actually goes into the number

Strip the branding away and most scores combine four ingredients in some proprietary ratio.

Parseability. Did the tool extract clean text, in a sensible order, from your file? This is the objective part, and the part worth acting on.

Section detection. Did it find experience, education and skills under recognisable headings? Also objective, also worth fixing.

Term overlap. How many words from the pasted job description appear in your resume? Partly objective, but the handling of synonyms and near-matches varies enormously between tools.

Style rules. Bullet counts, summary presence, date formats, page length. This is vendor opinion dressed as measurement, and it is where scores diverge most.

The first two are diagnostics you should take seriously. The last one is a house style you are free to disagree with.

Why the same resume scores differently everywhere

Because there is no standard. No governing body defines an ATS score, no published methodology is shared between vendors, and no employer validates any of the scales.

That means the numbers are not on a common axis. A 71 and an 84 from two products are not two readings of one quantity, in the way that two thermometers would be. They are two different tools applying two different rulebooks and both calling the output a percentage.

The practical consequence: never average scores, never chase the highest one, and never treat a drop between tools as a regression in your resume.

The measurable half, and how to check it yourself

You can verify the reliable part of any score without a tool.

1. Export the resume as PDF and open it.
2. Select all the body text, copy it, and paste it into a plain text editor.
3. Read what comes out.

Complete, ordered, readable text means parsing is not your problem. Nothing selectable means the file is an image, and every downstream finding is a symptom of that. The same principle underpins accessibility standards: the W3C’s WCAG guidelines require that text be available as text rather than embedded in an image, which is precisely what any parser needs.

A resume that passes this test has cleared the part of the score that is genuinely mechanical.

What the number cannot see

A score has no visibility into referrals, internal candidates, whether the role is already filled in practice, hiring freezes, or how many other applicants have the exact certification the team wanted. It cannot assess whether your experience is convincing, only whether certain strings appear.

It also has no view of how employers are obliged to run selection. Screening practices sit within anti-discrimination law, an area overseen by the Equal Employment Opportunity Commission, and no consumer scoring tool models that.

Most rejections are decided by factors the score never observed. Reading a low number as the cause is usually a misattribution, and an expensive one if it sends you rewriting from scratch.

The point at which you should stop

There is a clear finish line, and passing it is more useful than maximising anything.

• The file parses to clean, ordered text.
• Experience, education and skills are found under standard headings.
• Terms from the posting that are genuinely true of you appear naturally in your bullets.
• Nothing important is trapped in a header, footer, text box or image.

Once those four hold, further score-chasing changes the resume for the tool’s benefit rather than the reader’s. Stuffing additional keywords past this point tends to make bullets read worse to the human who eventually opens the file, which is a real cost against an imaginary gain.

Before: Skills: leadership, communication, teamwork, problem solving, detail oriented, results driven.

After: Led a four-person support rota through a platform migration, cutting escalations to the engineering team.

The second version demonstrates the same qualities and survives a conversation about it. The first is a keyword list that a human reader skims past.

For the surrounding detail, see how the different checker tools compare, how to source keywords from the posting itself, and where the free tiers stop. You can run our ATS checker against your own file, or sidestep most layout problems by starting from an ATS-friendly template.

Does formatting matter more than content?

Formatting matters until it stops breaking things, then it stops mattering. Content matters the whole way through.

Should I use a plain-text resume to be safe?

Not necessary. A clean single-column PDF with real text parses reliably and reads far better to a person.

The score is a rehearsal, not the performance. Once your file parses cleanly, put the remaining effort into what the bullets actually say. You can scan your resume free to see which lines are carrying weight and which are filler.