How our scores are calculated

Your resume is checked against a fixed list of rules covering contact details, section structure, writing content, and formatting. Every rule is deterministic: the same resume produces the same score every time, and each deduction traces to a specific finding.

Four category scores are produced, plus a fifth only when you have given us a job description to compare against:

• Parseability — whether the document can be read at all: contact details present, a template whose layout survives extraction, no photo dependence.
• Completeness — whether the sections an employer expects are actually filled in.
• Section quality — whether the writing does its job: a summary of usable length, bullets under your roles, evidence rather than duties alone.
• Formatting safety — font sizing, margins, and template choices known to cause extraction problems.
• Keyword relevance — your overall job-match score, described below. This one is shown only when you have pasted a job description; with no posting to compare against there is nothing to measure, so it is left out rather than guessed at.

Each category starts at 100 and loses points for what it finds. Structural gaps carry the heaviest penalties — a missing name or email costs 28 points of parseability each — and issues raised by the rule checks deduct 8 for a critical, 4 for a warning, and 2 for a tip. Scores are floored at 0.

The overall score is the plain unweighted mean of whichever categories applied, rounded. That means four categories without a job description and five with one — chosen because you can verify it by hand.

All of the scoring above runs in your browser and sends nothing anywhere. Every number you see is produced without your resume leaving your device.

A second, more detailed analysis engine also exists — six explainable categories including a Grammar score. It is not what scores your resume here; it serves our public API and the AI tailoring feature. One part of it IS available to you: an optional AI spelling and grammar check in the ATS step, which you trigger yourself. It sends the written parts of your resume — never your contact details — to an AI model and returns suggestions. It deliberately does not change the score above, because blending a second engine’s findings into this one’s number would produce a figure neither model explains.

This is a different number from the one above, and the difference matters. The ATS readiness score is about your resume. The parseability score is about the template — how well its layout survives being read by software, regardless of what you put in it. All 50 templates are scored the same way.

Scores map to four bands:

• 90–100 — Excellent ATS structure
• 80–89 — Strong
• 70–79 — Moderate
• Below 70 — Creative; better suited to direct submission than to an automated pipeline

A template scoring below 70 is not broken, and we do not hide these. A designer applying through a portfolio review is served well by a layout that would be a poor choice for a high-volume corporate application. The score exists so you can make that trade knowingly, which is why every low-scoring template carries a visible warning rather than a quiet deduction.

When you paste a job description, it is broken into skills, tools, technologies, responsibilities, and qualifications, and each of those is compared against your resume’s actual text. Five scores are produced, and the overall figure is their unweighted mean:

• Skills match — matched over total, across hard skills, soft skills, tools, and technologies combined.
• Experience match — your years, calculated from the date ranges on your own entries, against the years the posting asks for. This is a real ratio, not a keyword check. If the posting names no requirement, this scores 100.
• Responsibilities match — how many of the posting’s responsibilities appear in your experience and project bullets.
• Qualifications match — the same comparison against your education and certifications.
• Keyword coverage — matched over total across every term extracted.

A category the posting says nothing about scores 100, not 0. “Nothing to measure” and “measured and failed” are different facts and should not produce the same number.

One distinction runs through this whole feature: “missing from your resume” is a statement about your document, never about you. If a required skill is not found, the app says it is not in your resume and asks — it never records that you lack it. If you confirm you have it, it is tracked separately as something to evidence rather than folded into your match score, because an unevidenced claim should not raise a number.

Every AI rewrite is compared against what you originally wrote before you are shown it. The check asks one question per category of fact: does this appear in the rewrite, absent from the original, and not among the facts already known from your resume? If so, the AI introduced it, and it is flagged rather than presented as your text.

The locked categories are employer, job title, employment dates, degree, institution, certification, skill, technology, salary, numeric achievement, and award. Numbers are treated most strictly of all: any figure in a rewrite that was not in your original is treated as invented, without exception. That rule is the one that catches the classic failure — turning “Managed social media” into “Increased engagement by 40%”. Instead of that sentence, you get a question asking whether engagement, followers, traffic, or leads actually changed.

Suggestions are labelled by where their content came from — from your resume, provided by you, AI wording, or needs verification — and nothing is applied until you accept it. You can edit, reject, or undo any of it.

Being specific about the limits is what makes the rest of this page worth reading.

• We have no access to any commercial ATS. These systems are proprietary, numerous, and change without notice. Our checks model documented and widely observed parsing behaviour; they are not a test against Workday, Taleo, Greenhouse, or any other named product, and a high score is not a prediction that a specific system will accept your file. This is precisely why we never claim a resume is guaranteed ATS compatible.
• A score is not a hiring outcome. Nothing here evaluates whether you are a good candidate. These numbers describe a document.
• Text matching misses meaning. Job-match comparison works on words and known synonyms. A posting asking for one thing in language your resume expresses differently can read as a miss. Treat a gap as a prompt to check, not a verdict.
• Truth Guard’s detection is pattern-based, and patterns have edges. It reliably catches invented numbers, dates, and recognisable degrees, credentials, and well-known employers. A completely fabricated company name with no corporate suffix, or an obscure tool outside its vocabulary, can slip past. It is a strong safety net, not a guarantee, and you remain responsible for what your resume says.
• The optional AI grammar check is probabilistic. It can miss real errors and occasionally flag correct writing. It is the only part of this page that is not deterministic, which is why it is opt-in and why it does not affect your score.

When a scoring model changes in a way that would move your score, it is recorded on the changelog. The standards governing our written guidance are on the editorial standards page.