AI applications all sound the same.
Because they don’t know you.
The objection is fair: a language model that knows nothing about you writes the sentence everyone writes. So JACVault personalizes with you — we ask you how you want to sound and what you stand for. Your answers become the source every one of your applications is built from.
- 4 questions · 2 minutes · optional
- Your answers, not invented ones
- Every change with a reason
You answer a few questionsFree text, in your own words
That becomes your voiceTone, banned phrases, signature language
It feeds every applicationWithout you entering it again
of applicants in Germany use AI to write their cover letter — in 2023 it was 13%.
softgarden · “KI trifft Recruiting 2026”
n ≈ 7,000 applicants
of recruiters say applications are less tailored to the specific role than they used to be.
The Stepstone Group · 2025
704 recruiters, 3,495 candidates
of applicants believe AI has to be used in a way that keeps it undetectable.
Statista · January 2026
n = 500
The third figure is the wrong reflex to the first two. In German recruiting the objection isn’t “no AI”, it is: nothing generic. People are still screened out for missing skills and missing experience — not for the tool the text was written with.
How to spot a generic AI application
Not by its style, not by individual words — by the missing evidence. Six markers recruiters see within seconds.
You can spot generic AI applications by the fact that they claim qualities instead of naming roles: the text says what someone is like, but not what they have done.
Interchangeable adjectives instead of evidence
“Motivated”, “committed”, “solution-oriented” — qualities anyone can claim about themselves. Nobody can back them up: they stand for nothing verifiable.
No numbers, timeframes or results
Where a text has no “three years”, no “twelve projects”, no outcome, the model had nothing to build on — and filled the gap with language.
No reference to the actual role
The text fits the ad — and thirty others. If you can swap the company name without a single sentence becoming false, it wasn’t written for this role.
Always the same sentence melody
Paragraphs of equal length, lists of three, the closing outlook sentence. Without a source of their own, language models fall into the same pattern — which is why the texts resemble each other across industries.
Stock phrases that overwrite the person
“Hands-on mentality”, “proactive”, “a passion for innovative solutions”. These phrases aren’t wrong — they have just been used so often that they no longer say anything about the person writing.
One tone for every application
The same person sounds equally friendly and non-committal in every application, whether it goes to a corporation or a family business. If you read the text aloud and don’t recognize yourself, nobody else will either.
None of these markers proves AI was involved — people write like this too. The reverse does hold, though: a text none of the six applies to won’t stand out as generic, no matter what it was written with.
A model without a source writes the average
If the job ad and a prompt are the only basis, a language model has nothing about you — so it takes whatever occurred most often across millions of application texts. The result is linguistically flawless and interchangeable in substance.
A better prompt only moves the problem: it changes the tone, not the source. What’s missing is material about you — your roles, your numbers, your rules for phrasing. That is exactly what personalization is: not a style dial, but an additional source.
The most probable sentence is the one everyone writes.
Your voice is a source, not a setting
You answer a few short questions about your writing style. Your answers are stored and passed into every generation as a hard source — just like your profile and the job ad. The AI doesn’t invent your style, it applies it. AI-assisted, not AI-decided.
In your own words
Free text, no picklist. Whatever you never want to read, you write down — it lands on a banned list, not in a recommendation.
Into a style profile
Your answers become a fixed profile: tone, banned words, signature language, anti-goals. You see it on your dashboard and can restart it at any time.
Without you doing it again
CV, short profile, cover letter — in every future application your style profile sits on the table next to your profile and the job ad.
You don’t have to take our word for it. You see it line by line.
The personalization changelog puts the unpersonalized version and your version side by side — with a reason for what was phrased differently, and why. No leap of faith, no “the AI will handle it”.
It is also the honest answer to the question of whether personalization changes anything at all: you can look it up instead of hoping.
The whole engine in detailFour questions for everyone. Five more if you want to go deeper.
Both stages are optional, and they build on each other. We tell you here what sits in which plan — so you don’t have to look for it in the small print.
Basic personalization
Four short questions, about two minutes. They set how your documents sound — and what must never appear in them. In the free plan too.
- How should the text feel when you read it? Name three adjectives.
- Which words or stock phrases must never appear in your documents?
- What do you want to stand for in your work — in one sentence?
- What must never happen in your documents?
Result: tone, banned list, signature language, anti-goals — part of every generation from then on.
Premium personalization
Five more questions that go deeper: no longer just how you sound, but how you work and how you come across. It adds to the basic stage, it doesn’t replace it.
- What makes your contribution to projects genuinely valuable?
- How do you work when things get unclear or chaotic?
- How do you want to communicate with stakeholders?
- How should people perceive you after ten seconds?
- Which three words or terms best match your signature style?
Result: impact, working style, communication — on top of everything from stage 1.
If you answer nothing at all, everything still works: your application is then built from your profile and the job ad. Personalization is added the moment you set it up — and you can restart it whenever you like.
Applied to every application — without you doing anything again
Your style profile isn’t a template you pick. It automatically sits next to your profile and the job ad on every run.
Short and verifiable
What recruiters actually notice, what personalization changes and what it costs.
Why do AI applications all sound the same?
Because a language model with no information about the applicant picks the statistically most common phrasing — and that is the same for everybody. If the job ad and a prompt are the only basis, the result is a linguistically clean but interchangeable text. Texts only start to differ once the model gets different material: the person’s real roles, numbers and rules for phrasing.
Can recruiters tell whether an application was written with AI?
Not reliably. What recruiters do notice is something else: that a text is interchangeable. In a survey of 704 recruiters (The Stepstone Group, 2025), 69% said applications were less tailored to the role than they used to be and 73% found them less authentic — while 74% credited them with a more professional appearance. The more important figure from the same survey for applicants: people are screened out for missing skills (60%) and missing professional experience (42%); AI use does not appear as a reason at all. So what stands out isn’t the AI, it is the absence of evidence, the missing link to the specific role and the ever-same sentence melody.
How do you stop an AI application from sounding generic?
By giving the model a source rather than a better prompt. Three things work: first, complete data of your own — roles, timeframes, results, numbers; second, an explicit banned list for phrases you would never write yourself; third, a fixed tone that applies to every text. Write those three down once and you don’t have to retype them into a prompt for every application.
What is a personalization changelog?
A side-by-side view: on the left the unpersonalized version of a sentence, on the right the version that came out of your own style answers, and underneath the reason for the change. In JACVault it is created during the explicit personalization run and makes it verifiable what personalization actually changed — instead of merely claiming it.
Do I have to answer the questions — and does it cost extra?
No, both stages are optional. Without answers every generation runs as normal — it then draws on your profile and the job ad. Basic personalization with its four questions is included in every plan, the free one too. The five deeper questions of premium personalization are unlocked from Professional up and add to the basic stage rather than replacing it.
What is the difference to ChatGPT with a good prompt?
A prompt is transient: it applies to one text and has to be rewritten for every application. Your style answers in JACVault are stored sources — on every run they automatically sit next to your complete profile and the job ad, together with fixed checks against invented content. The difference isn’t the model, it is the path to the result.
How JACVault works — the whole journeyMake it sound like you
Sign up for free, upload your CV and answer the four questions — if you feel like it. From then on every application carries your voice.
Sign up for free