How it works
A score you can argue with
Atsy's rubric is published in full. You can read every check, see what it is worth, and predict your result before you upload anything.
Five pillars, 100 points
- Parse and structure — 35. Can the text be extracted at all, in the right order: columns, reading order, tables, header and footer bands, fonts, bullets.
- Contact and identity — 10. Can a recruiter reach you, and is the top of the page clean and free of details you should not be giving away.
- Experience and dates — 20. Titles, employers and dates a machine can actually read, in an order it can follow.
- Content and impact — 25. Action verbs, quantified results, bullet length, clichés, consistency.
- Skills presentation — 10. Whether your skills are findable text rather than pictures.
Two things Atsy will not do
It will not invent a number. Scoring is deterministic and runs the same published checks every time — no language model is involved, so the same CV always gets the same score. Any figure you see is one Atsy worked out from your own file.
It will not pretend to be an ATS. Every applicant tracking system ships a different parser, most are configured per employer, and none publishes a candidate score. Nobody can tell you "your Workday score" — what Atsy can tell you is how readable your file is, exactly where it breaks, and which engines are most likely to trip on it.
Why Atsy is free
Because the people who need it most are between jobs. There is no paid tier, no credit pack and no upsell — the whole thing runs on infrastructure that costs a few dollars a month, and it is built to stay that way: the scoring is plain arithmetic on your file, not an expensive model call.
What that buys you: no reason for Atsy to hold your data, show you ads, or tell you your CV is worse than it is.
The full rubric
Every check, what it is worth, and what it means. Nothing here is hidden, weighted secretly, or decided by a model.
Parse & structure — 35 pointsCan the text be extracted at all, in the order you wrote it.
- P01 caps your score at 25
- The PDF has no text layer — it is a picture of a CV, and a parser reads nothing from it.
- P02 6
- The page is in two columns, so extraction runs across the gutter and interleaves them.
- P03 3
- The order the text is stored in does not match the order you read it.
- P04 3
- Content sits in a running header or footer, the band parsers most often skip.
- P05 3
- Content is in a table: cell order is not reading order.
- P06 2
- A font is not embedded, so the text can extract as the wrong characters elsewhere.
- P07 2
- The extracted text is corrupted — ligatures, missing character maps, or spaced-out glyphs.
- P08 2
- Bullet characters come from an icon font and extract as junk.
- P09 1
- Emoji in the text.
- P10 2
- Skill ratings are graphics: a parser records nothing from them.
- P11 3
- Section headings are not ones a parser recognises.
- P12 2
- More than three pages, an unusual page size, or a very large file.
- P13 1
- A filename that upload forms mangle, or a generic one like "resume.pdf".
- P14 caps your score at 40
- Invisible or white-on-white text. Parsers see it and reviewers treat it as keyword stuffing.
- P15 1
- Links exist only as annotations, with no visible URL.
- P16 2
- Type set too small or too tight to survive a print or preview.
- P17 1
- Mixed scripts, or no document language set.
Contact & identity — 10 pointsWhether a recruiter can reach you, and whether the top of the page gives away more than it should.
- B01 2
- No name found on the first lines.
- B02 2
- No email address.
- B03 2
- No phone number, or one without a country code.
- B04 1
- No city and country — or a full street address, which you do not need to give.
- B05 1
- No LinkedIn or portfolio link in plain text.
- B06 1
- Contact details only in the page header, where parsers often miss them.
- B07 0.5
- A photo, in markets that do not expect one.
- B08 0.5
- Date of birth, marital status or an ID number.
Experience & dates — 20 pointsWhether your work history can be read as a history.
- C01 5
- A role without all three of title, employer and dates.
- C02 4
- Dates in a format a parser cannot read, or two formats in one CV.
- C03 3
- Roles not in reverse-chronological order.
- C04 2
- An unexplained gap of more than six months.
- C05 1
- The current role not marked "Present".
- C06 1
- Dates that cannot be right — an end before a start, or a future start.
- C07 2
- Job titles nobody searches for ("Ninja", "Guru", "Rockstar").
- C08 2
- No education section, or one missing qualification, institution or year.
Content & impact — 25 pointsWhether a person reading it would be persuaded.
- D01 3
- No summary, one that runs long, or one that says nothing specific.
- D02 4
- Fewer than 80% of bullets start with an action verb.
- D03 5
- Fewer than 40% of bullets contain a number: scale, money, time, or a change.
- D04 3
- Bullets far outside 8–30 words, or paragraphs inside a role.
- D05 2
- First-person pronouns.
- D06 2
- Filler phrases: "responsible for", "duties included", "team player".
- D07 1
- The same verb used more than three times.
- D08 1
- Tense that switches between and within roles.
- D09 2
- Typos, two spellings of the same word, or misspelt product names.
- D10 1
- Bullets punctuated inconsistently.
- D11 1
- Unusual acronyms never expanded.
Skills presentation — 10 pointsWhether your skills are findable text rather than decoration.
- E01 4
- No skills section.
- E02 2
- Skills only inside a table, columns or graphics.
- E03 2
- Fewer than six concrete skills, or nothing but soft skills.
- E04 1
- A very long list, or skills that never appear in your experience.
- E05 1
- Self-rated proficiency levels nobody can verify.
Score bands
- 90–100 Excellent — parses cleanly everywhere we test. Work on content.
- 75–89 Strong — safe on modern parsers, a few fixes left.
- 60–74 Needs work — real risk of being mis-read. Fix the criticals.
- 0–59 At risk — likely to be scrambled or dropped. Start at the top.