How we use AI

Reading the ad

Claude Haiku reads the ad, or photos of it, and finds the role, employer, country and interview round. You confirm or change them.

Rules Check

Claude Haiku lists the ad's rules. Our code then checks that each rule's quote is in the ad word for word, and drops any that aren't.

Free preview

Claude Haiku picks the question you're most likely to be asked first. Claude Opus writes one full answer from your resume.

Your stories

Claude Sonnet finds up to 12 stories in your resume. Each must quote the resume line it came from.

Interview Kit

Claude Opus plans the questions and writes each answer card, built on your stories.

Honesty check

Claude Haiku checks every claim about you in each card against your resume and your stories.

Company brief

Claude Haiku searches the web for the employer's news and values. We show only links to pages that search found.

On the day

Claude Haiku writes your day-of checklist, tips for extra stages and a thank-you note to edit.

Application documents

Coming in 2027. Claude Opus writes a resume for the job, and a cover letter or selection-criteria answers if the ad asks for them. Our code counts the words and checks the pages.

Some parts use no AI at all, such as practising with a friend and ticking off your checklist. Today, Claude is the only AI we use.

• Every claim about you is checked. Each answer is checked against your resume, your real stories and the answers you gave us. If a claim isn't backed, the card is rewritten once. If it still isn't, we flag it: “Check these facts before you use this answer.”
• Gaps stay gaps. Where an answer needs a detail your resume doesn't give, such as a number, we leave a [bracket]. You fill it in. We don't guess.
• Examples are labelled. If your resume has no story for a common question, we write an example, marked “Example · made up for you”. It shows the shape of a good answer. It's never treated as yours, and never counts as proof for a real answer. Swap it for a true story. Once you edit it in your own words, it's yours.
• Rules quote the ad. Every rule shows the ad's own words, so you can check it.
• Chances are estimates. The chance shown for each question is our estimate, not a fact.
• Limits are counted, not guessed. Our code counts the words and measures the pages of each application document.
• Links are real. Company brief links come only from the search results. Still, check anything important at its source.

• Help you during a live interview, test, assessment or recorded video interview.
• Do a take-home task for you. If you paste a real one, we switch to coach mode: we explain, plan, research and proofread, and you do the work.
• Ignore an employer's rule on AI. If the ad or an email says not to use AI, we show you that line. Then we pause AI help for that stage.
• Invent employers, job titles, dates, degrees, certificates, security clearances, citizenship, numbers or results.
• Apply for jobs or post anything for you.
• Analyse your face, your emotions or your voice.
• Sell your data, or use it to train AI.

• Read every answer and document before you use it.
• Fill in each [bracket] with what really happened, or take it out.
• Swap every example for a true story of your own.
• Say your answers in your own words. A kit is for practice, not for reading out.
• Follow each employer's rules on AI.

For each job, we show what you wrote and what Seal the Offer changed. So if an employer asks how you used AI, you can answer honestly.

• We send Anthropic only what each step needs: usually the job ad, your resume and your stories.
• Anthropic doesn't use data sent through its API to train its models, and keeps it only for a limited time.
• We log each AI step's cost and timing, never its content.
• Photos of an ad are read and then thrown away. We keep only the text.
• Our Privacy policy has the full details.

Email [email protected]. We're happy to explain how any part of your kit was made.