What does the number actually mean?

It is a weighted blend of four measurements: keyword overlap with a pasted posting, how many recognised section categories were found, a format assessment built from specific deductions, and whether bullets begin with accomplishment verbs. It is a readability measurement, not a verdict.

Can the same resume produce two different scores?

Yes, and both can be correct. Pasting a job description changes the weighting, so the same unchanged document scores differently with and without one. They answer different questions, so compare versions under the same conditions.

Does the score predict whether I will be hired?

No. It has no information about the employer, the role's seniority bar, the other applicants or the hiring process. It measures whether your document processes cleanly, which is a precondition rather than a prediction.

Should I aim for a perfect score?

No. The last few points usually cost more in honesty than they return, because closing a keyword gap you cannot defend produces a document that scores well and interviews badly.

Updated August 13, 2026

Reading an ATS readability score honestly

A score on a resume checker invites a story the number cannot support: that you are 72% likely to get the job, or that another tool’s 88 means you are a better candidate. Most of those stories are marketing. The useful reading is narrower.

This guide is how to interpret ResumeTune’s free readability score — what it is built from, what changes the mix, and what it is not. It is not a diagnostic walkthrough of every deduction; that belongs in the companion article on why scores come back low. Here the job is expectation-setting.

The free path is deterministic. Fixed rules run over the extracted text. No language model is involved. The same file, scored twice under the same conditions, returns the same number. That reproducibility is why the number cannot pretend to be a recruiter’s judgement.

The number is a weighted blend, not a verdict

ResumeTune reports a single integer from 0 to 100. That integer is four component scores, each already on a 0–100 scale, combined with fixed weights, then rounded and clamped so it cannot fall outside that range.

The four components are keyword overlap with a job posting, recognised section headings, format heuristics on the extracted text, and action-verb use in bullet lines. Blending them into one number is a convenience for the result screen. It is not evidence that those four questions are equally important to an employer — they are not, and employers do not share this formula.

Read the number as a summary of those checks on this extract, under the conditions you chose (job text pasted, or not). Do not read it as a percentile of applicants, a probability of interview, or a grade on your career.

The four components, in plain language

Keyword overlap exists only when you paste a job description. The scorer pulls up to twenty prominent terms from that posting and asks which appear in your resume text. The component is the share that appear. No posting means this component is skipped — its weight becomes zero, not a hidden 50.

Sections are a count of recognised heading categories in the extracted text, out of four, turned into a percentage. The check looks at lines, not at the visual template.

Format starts at 100 and subtracts for specific text-level problems listed below. Action verbs look at bullet lines (or, if there are no bullet markers, every line) and score how many start with a recognised accomplishment verb, plus a small bonus for using more than one of those verbs. None of the four is an AI review of writing quality.

The weights change when you paste a job

With a job description, the mix is 50% keyword, 20% sections, 15% format, and 15% action verbs. Overlap with the posting dominates because that is the only role-specific evidence the free check has.

Without a job description, keyword weight is zero. The remaining mix is 40% sections, 30% format, and 30% action verbs. The same file is being asked a different question: how cleanly does this document process, with no posting to compare against.

Those two numbers are not two attempts at the same truth. If you compare versions of a CV, compare them under the same condition — both with the same posting, or both with none — or the difference is mostly the formula, not the edit.

Format points: the exact deductions behind that slice

The format component is a budget of 100, reduced by the findings below. A 15-point contact finding on a 15% slice of the overall score is not the same event as a collapsed keyword component on a 50% slice.

Extracted text under 300 characters: −25. Usually an extraction failure, not a short career.
Extracted text over 8,000 characters: −10.
No email-shaped token and no phone-shaped digit run in the text: −15. Either one is enough to avoid this.
Fewer than two recognised section headings: −20.
Lines longer than 120 characters: −5 each, capped at −15.
Tab characters, or three or more spaces in a row: −10. A common fingerprint of columns or table layout in the extract.
More than three pages: −15, and only when page count is marked reliable. On DOCX, ResumeTune does not treat page count as reliable, so this deduction does not run. On a successfully parsed PDF it can. The scorer does not invent a page count when the flag is off.

Keyword matching is string overlap, including a compact form

When a posting is present, a term counts as found if it appears in the resume text after light normalisation (lowercase, collapsed spaces), or if a compact form matches. Compact matching strips dots, spaces, slashes, and hyphens, then looks for the compressed needle when it is at least three characters long. That is why “Node.js” can match “nodejs.”

It is still not meaning. Synonyms you did not write do not match. “JS” is not automatically “JavaScript.” The missing-term list is strings from the posting that were absent from your extract, not skills you must acquire before applying.

Chasing 100% overlap by pasting every missing token is a misreading of the score. An honest mid-range overlap on work you can discuss is more useful than a perfect number on a stuffed page.

What the score is not

It is not a hiring prediction. ResumeTune has no model of the employer, the other applicants, the hiring manager’s taste, or the seniority bar. A high score means this extract processed cleanly under these rules. A low score means it did not. Neither fact is an offer or a rank.

It is not a simulation of Workday, Greenhouse, Lever, or any other named ATS. Those products are not public scoring engines you can clone from a blog post. This tool reports its own checks.

It is not an AI assessment of your writing, and the free path does not call a language model. It is not a font or layout audit. Page-count findings run only when the parser marks page count reliable. Font and layout metadata flags stay off; the UI does not invent those findings.

Two scores on the same CV can both be correct

People treat a changed number as a changed quality of the person. Here the more common explanation is a changed input. Paste a posting and you add a 50% keyword slice. Skip the posting and you are looking at structure and format only. Export a new PDF and you may change what the extractor recovers.

Language detection is another input, not a grade. The free check scores English and Spanish resumes. An unsupported language returns an error with no score. If a posting and a CV are in different supported languages, the result can note that keyword matching may be less reliable. That note is a warning on the issue list; it is not one of the numbered format deductions above.

When you re-run, keep the file and the job text aligned with the previous run if you want to learn whether an edit moved the number. Otherwise you are comparing two experiments.

How to use the number without overclaiming

Use it as a consistency check on the file you will upload. Did text extract? Are there headings the rules recognise? Did email or phone appear? If you pasted a posting, which of its prominent terms are already in your wording?

Use the components, not only the headline. A total in the 60s with a strong format slice and a weak keyword slice is a wording question. A total in the 20s across every slice is usually extraction.

Do not use it as a reason to invent experience, to rewrite a truthful resume into keyword soup, or to assume a recruiter will see this same integer. They will not.

Where ResumeTune fits

The homepage upload is the free check: PDF or DOCX, optional job description, deterministic rules, no account. The score is a readability and overlap summary of the extract. Optional paid work, when offered, is a separate path and is not how this number is produced.

If you want a repair list — which deduction fired, in what order to fix extraction versus wording — read the guide on why a resume scores low. If you want to interpret the integer you already have, this page is the honest ceiling: a weighted blend of four public components, not a forecast of who gets hired.

Frequently asked questions

What does the number actually mean?: It is a weighted blend of four measurements: keyword overlap with a pasted posting, how many recognised section categories were found, a format assessment built from specific deductions, and whether bullets begin with accomplishment verbs. It is a readability measurement, not a verdict.
Can the same resume produce two different scores?: Yes, and both can be correct. Pasting a job description changes the weighting, so the same unchanged document scores differently with and without one. They answer different questions, so compare versions under the same conditions.
Does the score predict whether I will be hired?: No. It has no information about the employer, the role's seniority bar, the other applicants or the hiring process. It measures whether your document processes cleanly, which is a precondition rather than a prediction.
Should I aim for a perfect score?: No. The last few points usually cost more in honesty than they return, because closing a keyword gap you cannot defend produces a document that scores well and interviews badly.