What AI does, and what code does
- AI reads the posting. It lists the requirements and marks each one key, important or nice-to-have. Key requirements get a second check.
- AI looks for evidence in your resume. For each requirement it judges matched, partial or missing, and must quote the sentences it relied on.
- Code checks every quote. A quote that isn’t in your resume (ignoring case, punctuation and spacing) is thrown out. A matched or partial requirement left with no real quote becomes missing.
- Code does all the scoring. Every number below is calculated from your document and those checked results. No score is an AI opinion.
How a requirement counts
- Matched — a sentence in your resume shows it.
- Partial — your resume shows something related but narrower or weaker. A requirement that names two things (“team coordination and junior mentoring”) is partial when only one is shown. A hard skill shown only in coursework or certifications is partial until you’ve used it at work.
- Missing — nothing in your resume shows it yet.
- Your answers — confirming something is true for you doesn’t move the score on its own. It counts once a line in your resume shows it, judged like any other line.
ATS compatibility
How well screening software is likely to read your resume and match it to the posting. This is our model, not any vendor’s score: ATS products don’t publish how they rank, and many don’t rank at all — a person searches and filters what the software extracted.
Extraction integrity
30 pts
Whether a parser recovers your name, contact details, and the title, employer and dates of every role, plus education and skills, as separate fields. Employers and dates weigh more than a skills line.
Parse safety
20 pts
Seven checks: standard section headings, a single column, no tables, no graphics or text boxes, real selectable text, contact details on their own lines, and one date format throughout.
Requirement coverage
35 pts
Each requirement from the posting counts as matched (full), partial (half) or missing (none). Key requirements count more than important ones, and important ones more than nice-to-haves, and no single requirement can carry more than 15% of these points.
Title and vocabulary match
15 pts
How closely your current title (10) and previous title (5) match the posting's title.
Human recruiter read
How clear and convincing the same resume feels to a person skimming it for well under a minute.
First-pass landmarks
25 pts
What a six-second skim finds in the top third of page one: name, current title and employer, previous role, dates and education. How closely your current title names the role carries the most.
Top-third relevance
25 pts
How many key requirements are evidenced in the top third of page one or at the top of their role.
Bullet specificity
25 pts
Each experience bullet is graded: a figure (a count, share, money or scale) is full proof, a named outcome without a figure is most of it, a concrete action on a named system is a start, and a duty (“Responsible for…”, “Worked on…”) earns nothing. The first bullet of your most recent role counts double.
Scannability
10 pts
Bullets of 8 to 30 words (6), no dense paragraphs in experience (2.5), nothing running past two lines (1.5).
Consistency and trajectory
15 pts
Titles, dates and employers agree everywhere, with no unexplained gap over six months between roles. Time covered by dated education is not a gap.
What comes off
- Keyword repetition (ATS) — only words the posting asks for count, so describing your own work (“patients”, “students”) never costs points. A one-word ask (“Python”) counts as that word; a phrase (“Data analysis”) counts only when the whole phrase is repeated, so “data” on its own never does. Once your most-repeated posting keyword is used 7 or more times and passes 3% of all words (counted over at least 80 words), points come off, reaching the full 8 at 7%. Each use of a word from your own job titles counts half. A vendor name that starts a product name (“Oracle” in “Oracle Integration Cloud”) isn’t counted when the posting names that vendor too.
- Over-length (recruiter) — 4 points per page over the guide below, capped at 8.
Screening questions are not scored
These are listed separately: years of experience (“5+ years”, “minimum five years”, “several years of”), degrees and education, location, residence and relocation, work authorisation and visa sponsorship, availability (notice period, start date, work pattern, travel share), named certifications and licences (such as PMP, CISSP, CPA or RN), and spoken languages. An ATS doesn’t reject on them and a recruiter asks about them on the application form, so they never move either score, never become a question, and are never written into your resume.
Potential
Next to each score we show what it would reach if every requirement you haven’t shown yet were evidenced. It uses the same parts as above and leaves screening questions out.
Page-length guide
There is no fixed page rule by years of experience. Length is judged on density: how many pages your real evidence needs at a comfortable reading size. The page count we score is the one your export actually produces, so the number in the editor, on the tailoring card and in the downloaded file is always the same. Three pages apply only to an academic, research or clinical CV, which we detect from sections such as Publications, Grants, Teaching or Clinical Experience.
Everyone
The shortest readable length that still carries your strongest evidence
A second page
Fine when real evidence fills it, not when it is repetition or padding
Academic, research or clinical CV
3 pages
Same input, same numbers
The first time we read a resume against a posting, we keep that reading. The same resume and the same posting give the same result after that, instead of a new AI guess each time. When you edit, the scores recalculate in code straight away, with no new AI reading. When we improve our checks, a stored reading may be refreshed once.
What the scores can’t tell you
- They don’t predict an interview. We haven’t yet measured whether higher scores lead to more callbacks, so treat them as a checklist, not a forecast.
- The AI can misjudge a requirement — calling a related tool a match, or missing a synonym. Every judgement shows its quote so you can check it, and you can answer any requirement yourself.
- Requirements come from the posting as written. If a posting is vague, the list will be too.
What we never do
No score is improved by inventing anything. A matched requirement always points to wording already in your resume. Tailoring only rearranges and rewords what your resume says, and a line you add from your own answers says only what you told us. A missing requirement stays missing until you tell us it is genuinely true for you.
Score your resume against a posting — the first one is free.