Start with the job ad, not with you
Paste the ad and ask the model to list the five things the employer clearly cares about most. Those five are what the application should answer, in that order. Most applications are about the applicant; the good ones are about the job.
Give the model facts, not adjectives
What you actually did, with numbers where they exist: “responsible for the web shop, revenue up 30% in one year”. The model can phrase, but it cannot know. Whatever you leave out, it invents, and that is where applications get embarrassing.
One job, one application
The same letter to ten jobs is what AI does most easily, and what recruiters notice fastest. Write one letter per job, with the five points from the ad as its spine. With help it takes twenty minutes, not two hours.
Proofread like a sceptic
- Cut every sentence that could be in anyone's application
- Cut “passionate”, “driven”, “results-oriented”
- Check every number and claim against reality
- Read it aloud: if you would not say it in an interview, it should not be there
The CV: short, chronological, concrete
Two pages, newest first, three bullets per role with what you did and what it led to. Ask the model to cut, not to decorate. No photo unless the industry expects one, and no skills you cannot defend.
Privacy
You are pasting your CV into a tool. Check what is stored and for how long, and avoid entering national ID numbers or other people's personal data. A tool that keeps little is the one to prefer.
In Marshi
Marshi's CV tool writes a CV and a cover letter for one job from a CV you upload or paste, plus the job ad. Three applications from that browser, then a free account. With an account, Marshi keeps a living profile — Experience and education on a personal profile — and for each job uses only what is relevant. LinkedIn import is for a signed-in account.
