How the Resume X-ray scores
No resume scanner publishes how it scores. We do, because a score you can't check is just an opinion with a number on it.
The score measures how clearly your resume shows what recruiters screen for. It doesn't predict whether you'll be hired: no public study links any resume score to callbacks, ours included. We'll publish callback data by score band only once we have enough of our own to say something honest.
Six categories
| Category | Weight | What it measures |
|---|---|---|
| Reads cleanly | 20% | Whether an applicant system gets all of your text, in order, with your name and contact details intact. |
| Proof and impact | 25% | Whether your lines show what changed because of your work: numbers, results, scale and how you did it. |
| Target fit | 20% | Whether the job title, location, years and skills a recruiter filters on for your target are easy to find. |
| Writing and voice | 15% | Typos, buzzwords, pronouns, tense, and lines that all sound the same. |
| Visual design | 10% | Readable type, clear headings, sensible margins and spacing, contrast and tidy alignment, measured from your file. |
| Structure and completeness | 10% | Contact details, standard sections, length, dates, and keeping the detail where it counts. |
Target fit is scored only when you give a target role or paste a job posting; without one you get a general resume score and the other weights are scaled up to fill 100%. Design is scored from a PDF (fully) or a .docx (declared styles only); pasted text has no layout, so it isn't scored.
Inside each category, every check is worth set points, and the category's checks add up to 100. A check loses points in proportion to how much of your resume it affects: one weak line out of twenty costs far less than twelve out of twenty. Each fix on your report shows how many points it's worth to the total.
Five bands
| Band | Score | Meaning |
|---|---|---|
| Excellent | 90–100 | Ready to send; only polish left. |
| Strong | 75–89 | Competitive. A few targeted fixes. |
| Solid foundation | 60–74 | Good bones. The top three fixes will move it most. |
| Needs work | 40–59 | Real problems a recruiter will notice. |
| Not ready to send | 0–39 | A file-level problem: unreadable, unreachable or hidden text. |
These band edges are our design choice, not a validated standard.
Caps
A few problems make a resume unusable whatever else is right, so they cap the total:
- At most 35: No text layer: an applicant system would see an empty resume.
- At most 39: Hidden or instruction-like text: recruiters look for it, and finding it ends the review.
- At most 39: No email and no phone number: nobody can reach you.
- At most 59: Over 2% of the characters don't convert to text.
- At most 74: Something marked Critical is still open, so the score stays below Strong until it's fixed.
- At most 89: Something marked Fix is still open, so the score can't reach Excellent (which means only polish is left).
What each check rests on
Every check on your report carries one of these labels, so you can decide how much weight to give it:
- Backed by a large study: a peer-reviewed study or a large field experiment.
- Recruiter survey: what recruiters or hiring managers report; useful, but surveys are often run by companies that sell resume services.
- Applicant-system documentation: what applicant-system vendors document about their own parsers.
- Published standard: a published standard, such as the WCAG contrast minimum.
- Seen in your file: something we can see directly in your file.
- Style convention: typographic or career convention, widely followed but not tested against outcomes.
- Our rule of thumb: our own heuristic, to be tuned as we gather data.
What we deliberately don't do
- We don't guess whether AI wrote your resume. Seven AI detectors wrongly flagged 61% of human essays by non-native English writers, on average, in a Stanford study, and OpenAI withdrew its own detector. We flag patterns recruiters say read generic (stock words, every line the same shape, the same opening verb) and cap that check at 3 points of the total.
- We don't penalise style. A study of 94,000 applications found no reliable interview difference for two-column layouts, colour or template choice. We score whether the page reads cleanly and comfortably, not whether we like it.
- Repeating keywords doesn't raise the score. Each skill counts once, and stuffing loses points.
- We never add skills or numbers for you. Suggestions use [placeholders] for anything you haven't told us.
Detail where it counts
Recruiters read the most recent years most closely. We suggest a simple budget: your current role 4–6 lines, roles in the last 7 years 3–5, roles 7–15 years back 1–2, and anything older on one “Earlier career” line. Keep titles and dates; cut the oldest, weakest lines first.
Age signals: in a field experiment with over 40,000 applications, older applicants got 18–35% fewer callbacks, with age signalled mainly by graduation year. We point these out as an option, never as a mistake.
Privacy
Your file is read in memory in an isolated process and not stored unless you choose to save it. Only extracted text is sent to our AI provider for the written review, once. Details in our privacy policy.