Improving your CV score: have it sharpened or rewrite it yourself?

Seven checks, one CV, two adverts — 4 September 2026

One anonymous CV check per run on swissjobs.app, German interface, 4 September 2026. The CV is fictional; both adverts are real and open in our own index. Two runs were repeated to measure the tool’s own noise: 48 against 48, and 53 against 51.

The routes compared — what each one actually changes

Every cell read on the provider’s own pages on 4 September 2026. Where a page does not state something, the cell says “not stated” rather than guessing.

What sharpening costs

Prices read on the providers’ own pricing pages on 4 September 2026. Foreign currencies converted at the 3 September 2026 rate and shown as approximations.

What we did, so the numbers mean something

We invented a CV of a kind Switzerland produces in quantity: a Konstrukteur EFZ (a federally certified design draughtsman — EFZ is the Swiss vocational qualification), five years in design, then technical administration in a metalwork firm since 2021, a project-management certificate, German as first language, French B1, English B2. About 210 words, one page, nothing remarkable.

Then two real adverts from our own index, both titled Projektleiter/in, both open that day: one from a firm near Bern asking for customer projects, quotations, costings and schedules, and one from a lift company in Wettswil asking for order management of lift projects, experience with replacement installations and modernisations, and CAD and SAP knowledge. Same title, two entirely different texts.

Three versions of the CV. The original. A sharpened version aimed at the first advert — swapping words, reordering sentences, adopting the advert’s vocabulary, inventing nothing. We followed exactly the instruction our own “improve with AI” button sends to the model: replace sentences rather than append them, weave in missing keywords only where the person genuinely has the experience, never invent employers, dates, numbers or skills. And a third version that breaks that rule on purpose.

Each run was its own anonymous check in a fresh browser profile, so no run could see the previous one. Two runs were repeated to measure the tool’s own variation: the original against advert A scored 48 both times; the sharpened version against advert B scored 53 and 51. The differences discussed below are larger than that noise — but this is one CV, two adverts, one day. It is an experiment, not a statistic.

What the sharpening bought

Against the advert it was made for, it worked immediately: 48 to 57, with the word-match dimension jumping from 62 to 82. Hard requirements rose too (68 to 80) and soft skills (72 to 85), because the advert names things like organisational talent, a structured way of working and commercial understanding, and the sharpened version now contains those phrases.

That is not a trick, it is the mechanism. A checking tool compares two texts. Use the other text’s words in yours and the two become more similar, and similarity is what is being measured. The candidate’s actual career is identical in both versions.

In our tool’s own bands, 48 to 57 crosses one boundary — from “partial fit” to “moderate fit” — and the displayed estimate for reaching an HR conversation moves from about 4 per cent to about 8 per cent. Our own interface prints, under that estimate, that it is derived from the score and is not a measurement of real applications. That caveat is the right one, and it applies to every such percentage on the market.

What it did not buy — and this is the answer to the question

Against the second advert, same job title and a different employer, the keyword score stayed at exactly 62. Not one point of the 20-point gain carried across. The overall score still rose a little (47 to 53), but through general dimensions — self-presentation, soft skills — rather than through fit to the text.

Two scores barely moved in any of the seven runs: motivation stayed between 12 and 18, achievements between 18 and 22. No amount of rephrasing shifted them, including the dishonest kind. That is logical once you look at it: a motivation is a reason for this one employer, and an achievement is a number or a result. Neither is in the document, because nobody wrote them down — and a tool forbidden to invent cannot add them.

So if your score is low and you want to know whether sharpening will fix it, read the breakdown rather than the headline. Missing words are a minutes-long problem. Missing numbers, results and reasons are an hour with yourself.

Why English-language Swiss adverts make imported advice look right

Most advice about optimising a CV for a score is written in English, for a US market, and it is not wrong there. Our index shows why it half-works here. Taking every open advert whose job title is also used by at least three other employers, removing apprenticeships and republished texts, we measured how much of one advert’s distinctive vocabulary appears in the other: 10.1 per cent in English-language Swiss adverts, against 7.3 per cent in German-language ones. Random pairs with different titles score 2.5 per cent.

English-language adverts in Switzerland are written closer to an international template — more standardised phrasing, more transferable vocabulary — so keyword work does carry a little further. It still only carries a tenth of the way. And the employer behind that English advert is frequently Swiss, with Swiss expectations about the application file, whatever language the posting is in.

The second figure matters more for anyone applying from abroad: only 9.5 per cent of open English-language adverts share their job title with another employer at all. Roughly nine in ten are, right now, the only advert with that title. There is rarely such a thing as “the CV for this role”. There is the CV for this advert.

The test that shows the limit: the version that lies

The third version claims what the second advert asks for: order management of lift projects, experience with replacement installations and modernisations, preparatory building and disposition plans, CAD and SAP. None of it is in the candidate’s career. The check returned 62 — the highest number in the experiment — with a keyword score of 88 and a hard-skills score of 93.

That is not a particular weakness of our check; it is the boundary of every checking tool. It reads two texts and compares them. Whether one of them is true is outside what it can see. Any number a piece of software gives your CV rewards the claim exactly as much as the fact.

In Switzerland there is a second reason not to test that boundary. A Swiss application file carries Arbeitszeugnisse — written references from every previous employer, describing what you actually did. A responsibility that shines in the CV and is missing from the reference is not a nuance; it is a contradiction inside the same envelope. That is the practical reason the rule “rephrase, never assert” bites harder here than in a market that only knows the resume.

Whoever writes the fix also awards the mark

Jobscan describes AI Optimize accurately: suggestions tied to the job’s required skills, accepted or rejected one at a time, plus the line “Watch your resume’s Match Rate increase in real-time as you accept suggestions.” Honestly described — and that is the problem. The party awarding the mark supplies the changes that raise it.

Ours would behave identically. Our “improve my CV with AI” button takes the check report and turns it into an instruction for the model, including the missing keywords from precisely the dimension that gets re-scored afterwards. That is why we did the sharpening by hand for this experiment and left the tool only the measuring. The yardstick is still ours; anyone who wants an independent read should send the same three versions through cvswiss.ch or Jobscan.

What the instruction contains is checkable, and it is a fair standard for any serious sharpening tool: weave in missing keywords only where the person genuinely has the experience; never invent employers, dates, numbers or skills; if a metric is missing, phrase the existing achievement more sharply rather than fabricating a figure; replace rather than lengthen, so the document stays the same size; and do not change the language of the document.

What rewriting it yourself gives you that no button does

First: you can repeat it in the interview. A sentence you wrote survives “tell me more about that”. A sentence you accepted because it lifted a score does not always.

Second: rewriting is where the missing material appears. How many projects at once? How big was the largest? What is measurably better since you arrived? Those answers exist in no document that could be rephrased — they surface while you think. Afterwards a tool can pour them into the right vocabulary in seconds.

Third: the Swiss lines an international tool does not know. Workload written as a Pensum in per cent, not as “full-time”; language ability written as A1 to C2, not as “fluent”; the permit line for anyone without a Swiss passport; Arbeitszeugnisse as part of the file; and the photograph, which is conventional here and discouraged in Germany and the UK. If you are relocating, that list is shorter than the advice industry suggests — and it is about content, not about English.

The honest counterweight: rewriting costs one to three hours the first time, and from the fifth application onwards the repetition becomes the real burden. That is exactly where a sharpening pass earns its francs or its minute.

The order we would recommend

Once, for yourself: write the facts down. Two or three lines per position covering what you were responsible for, with a number wherever one exists. That is the part you never have to do again and the part no tool can do for you.

Per application: take the advert, have the CV scored against it, and sharpen the wording — by hand or with a button, which changes little in the result. Expect a gain of the order our experiment produced, a band rather than a transformation. And do not throw the sharpened version at the next advert; it was not built for that one.

Never: leave a line standing that your Arbeitszeugnis will not support. The score rewards it, the interview does not.

If you are applying once and want no tool at all, lebenslauf.ch is free, needs no login and lets you edit directly in the document. If you want to check against a specific Swiss advert and then rewrite, our check is built for it: one free check per browser, a report across ten categories including permit, language levels and photo, and the advert comes from the same index you found the job in.