How the score works, and what the AI does

Four dimensions, one hundred points

Each dimension is a real measurement you can check:

• Quantification density (30 pts): the share of bullets that carry a number a recruiter can judge. Full marks at 60%. In many roles, half the work is design review and stakeholder alignment. Asking for more just makes people invent numbers. [N] placeholders do not count, and neither do bare years.
• Verb strength (25 pts): each bullet is graded on its opening verb. Top verbs claim a result. Middle verbs name real work. Weak verbs hand the credit to someone else (supported, assisted). A bullet that does not start with a verb scores nothing.
• Filler rate (20 pts): we match your text against a dictionary of padding words (various, successfully, familiar with). Each hit costs points. Cutting them barely shortens the page. It just lets the reader see what you did.
• Structure (25 pts): whether the expected sections are there. Education, experience, skills, a contact email, and enough bullets with real content.

The [N] rule: missing numbers are never invented

If a number is not in your material, the AI does not supply one. Where the rewrite needs a figure, the draft shows a yellow [N] for you to fill in.

While any [N] is still unfilled, the total score is capped at 84. An unfinished resume should not score like a finished one.

The reason shows up in the interview. Every number on the page is one you will be asked about. A made-up number that falls apart costs you more than no number at all.

What the AI does, and does not do

The AI does two things. It rewrites your Chinese or English material into the English that international employers expect. The style follows the industry you picked. In review mode, it explains in words why a sentence lost points.

It does not produce the score. That is the rules engine’s job. It does not invent figures, and it does not invent experience you never gave us. Rewrites use your material only.

How the example pages are written

Every English example on the role pages is written to teach a point. None of them come from a real user’s resume. Where a metric belongs, the example carries [N]. Examples name no real employers or people, and cite no made-up statistics.

Each page shows its published and updated dates. Dates move only when the content actually changed.

What is not promised

The score measures how readable your resume is and how much evidence it carries. It does not measure your ability, and it does not predict your odds. No resume tool can guarantee interviews or offers, and this one does not pretend to.

Where each number on this page comes from

The "7.4 seconds" on the home page is not our measurement. It comes from The Ladders' 2018 resume eye-tracking study (The Ladders, "Eye-Tracking Study", 2018). It is the premise this product was built on, not a result we produced, so it is attributed.

Every other number on this site is computed rather than cited: the four dimension weights, the 84-point cap, the quantification threshold — all of them read from lib/rating/score.ts, and all of them identical if you score the same resume a hundred times. The [N], [X%] and [$Y] in the examples are blanks, not data. They are exactly the figures we refuse to fill in for you.

We do not cite studies we have not read, we do not write "surveys show" without a survey, and we do not put star ratings on a site that has no real ratings. The day we publish statistics of our own, the sample size and the method will be published with them.