Tailor your CV to a job description — without inventing anything
Upload a CV once, paste a job description, and get a version aimed at that specific role — assembled only from facts your existing CV supports, with every claim on the output checked back against it.
Most CVs are written once, for a career, and then read many times, for specific roles. That mismatch is the whole problem. Eight years of real work supports several different honest emphases — and the version that answers the posting in front of the reader lands very differently from the general-purpose one, without a single new claim being made.
Tailoring by hand is the correct solution and almost nobody sustains it past the fifth application. Tailoring with an unconstrained generator is fast and quietly introduces claims you will have to defend out loud.
What actually happens to your CV
Your CV is parsed once, into facts rather than text. On upload, the document becomes a fact graph: atomic statements about what you did, where, when, with what, alongside an index of the skills those statements support. This happens once per CV, not once per application.
The job description is parsed into requirements. Not keywords — requirements, each with a kind and a seniority signal, so “five years of Python” and “familiarity with Python” are not treated as the same demand.
A match plan is built before anything is written. It pairs requirements against the facts that speak to them, and — this is the part that matters — it records the requirements nothing in your history covers. Those stay visible as gaps. Nothing is written to fill them.
Selection and rephrasing, not authoring. The rewrite step chooses which facts to lead with, how to phrase them for this reader, and what to cut. It can promote a project you had buried in the third bullet of your second role. It cannot give you a project you never had, because it is drawing from the graph rather than from the job description.
Every claim on the draft is then extracted and checked. A separate pass reads the finished document as if it had never seen your CV, pulls out each claim, and labels it against the fact graph: supported, unsupported, or contradicted. Independence is the point — the step that checks is not the step that wrote.
One correction pass, at most. If the check finds a contradiction or a serious unsupported claim, the draft goes back once with the offending text named and only two permitted moves: soften it to what the evidence supports, or remove it. Never “find support for it”. The corrected draft is then re-extracted and re-verified from scratch, because old verdicts describe a document that no longer exists — and the correction is kept only if the weighted risk actually dropped.
The scores are code, not a model’s opinion
Four dimensions — Job Match, ATS Readability, Evidence Strength and Truthfulness — plus an overall figure. The model contributes structured facts; the arithmetic is TypeScript. That has two consequences worth stating plainly.
The same CV and job always score the same. And a bad truthfulness result cannot be outweighed by a good match: below 85, truthfulness caps the overall score at 59. A contradicted claim marks the whole result blocked.
Be precise about what “blocked” means, because it would be easy to oversell: it is a verdict, not a lock. Nothing stops you exporting the PDF anyway. What it stops is you doing it without knowing.
What you get, and what you can change
A tailored CV in one of four templates — Default, Modern, Compact and Elegant — switchable per application, with the PDF rendered from the same template you previewed, so the export cannot drift from the preview.
Every line is editable inline. Edits go back through the audit rather than around it. Alongside the document sits the report: the verdict, the score breakdown, the risks worst-first, and the match plan with its uncovered requirements still listed.
Where this is the wrong tool
It tailors a CV you already have. If you are writing your first one from a blank page, it has nothing to parse and a builder will serve you better — we say the same thing on every comparison page, including the ones where it costs us the sale.
It also cannot audit your source material. If your uploaded CV overstates something, that overstatement is a fact as far as the graph is concerned, and it will be faithfully carried forward. The guarantee is that nothing new is invented, not that what you wrote was true.
Questions
- How do you tailor a CV to a job description without lying?
- By separating selection from invention. Your CV is parsed once into a fact graph of atomic statements — what you did, where, when, with what. Tailoring then selects from that graph and rephrases, so a line can change its emphasis and its wording but cannot acquire a fact the graph does not hold. A separate pass afterwards extracts every claim on the draft and labels it supported, unsupported or contradicted against the same graph.
- Does tailoring a CV for each job actually help?
- It helps for a reason people often misstate. It is not that recruiters reward effort; it is that most CVs are written for a career and read for a role. The same eight years of work supports several different honest emphases, and the one that answers the posting in front of you reads as a much better fit than the general-purpose version — without a single new claim being made.
- What happens if the job asks for something my CV doesn't have?
- The gap stays a gap. It appears in the match plan as a requirement your history does not cover, and no line is written to paper over it. This is the case worth testing on any AI CV tool, because it is the one where an unconstrained generator invents — and it never comes up while you are trying a tool against a job you already match.
- Are the scores generated by AI?
- No. The model produces structured facts — the claims, and whether the fact graph supports each one. The scores are then computed in TypeScript from those facts, so the same CV and job always produce the same numbers and each one can be traced to the inputs that moved it. A truthfulness score below 85 caps the overall score at 59 no matter how well the CV matches.
- Can I edit the tailored CV afterwards?
- Yes, inline, line by line — and an edit re-runs the audit rather than leaving stale verdicts on screen. If you rewrite a bullet, its claims are re-extracted and re-verified; claims whose text you did not touch keep the verdict they already had, which is what keeps a one-line edit fast.