Apply for a job with AI—and still sound like yourself.

Start with the vacancy and your own evidence—not ‘write me a perfect CV’

You want an employer to understand what you can contribute, not admire an application that could belong to anyone. Start with the actual vacancy and rough notes about work you really did. Ask the assistant to find relevant examples and missing information before it writes a sentence.

ChatGPT, Claude and Gemini can support this conversation when their available tools can read your material. Use redacted text, remove addresses and reference contact details, and do not upload confidential employer documents. Check that the assistant actually read the file. A polished answer does not establish that it did.

Check what this employer permits before using AI

Application editing, rehearsal and an assessed task are different activities. Read the rules for the actual stage. If the instructions are unclear, ask; do not interpret silence as permission or assume an accessibility arrangement needs no discussion.

The UK Civil Service offers one concrete example, not a universal policy: its published guidance permits clarification and preparation but prohibits misleading experience and AI-generated responses during interviews. Other employers may set different restrictions. Keep any required disclosure and remain responsible for everything you submit.

• Which tools are permitted for this stage?
• Must AI assistance be disclosed, and how?
• What data can you share?
• Are live interviews or assessment answers explicitly restricted?

Let AI interview you before it writes your cover letter

Answer in rough language first: what work appeals to you, which real task demonstrates a relevant skill, what you personally changed and what you still want to learn. A useful assistant asks for those details instead of inventing a lifelong passion for the employer.

Then ask for a short draft with one reason specific to this role and one supported example. Supply a short sample of your own ordinary writing if tone matters. Edit it aloud. If you would never say a phrase, replace it; if you cannot explain a claim, remove or clarify it.

Replace application clichés with something you can explain

These are fictional editing examples, not recommended claims to copy. Only use a concrete version if it describes what actually happened to you. The point is evidence and your own voice, not adding typos, avoiding a particular punctuation mark or fooling a detector.

Check the final CV, letter and form against the same fact sheet

Tailoring a CV means making relevant experience easier to find. It does not mean upgrading your role, claiming a required qualification you lack or inserting keywords for work you have never done. Keep team achievements separate from your contribution.

Check final files and pasted form answers together. Dates, qualification status, job titles and numerical claims can drift between drafts. Leave contradictions visible until you resolve them from your records. There is no guaranteed ATS keyword formula or AI-generated pass score in this workflow.

• Could I explain every claim with a real example?
• Are unfinished courses clearly unfinished?
• Are team results and my actions distinguished?
• Do dates and responsibility descriptions agree across documents?
• Have I followed the actual word limit, format and disclosure instructions?

Practise the conversation, not a memorised perfect answer

Choose real examples for the requirements that matter. STAR—situation, task, action and result—can organise an answer, but use it as a scaffold rather than a script. The National Careers Service recommends concise, conversational examples and preparation for follow-up questions.

Ask one question at a time, answer without a generated script, then let the assistant probe the detail: ‘What did you decide yourself?’, ‘What did not work?’ or ‘How do you know the outcome?’ Feedback should identify a missing detail, not fabricate the answer. Practise an honest answer for a skill you are still learning and prepare your own questions about the work. Do not use rehearsal prompts to generate answers secretly during a restricted live interview.

Improve one weak answer, then test it with a different follow-up

Make the loop specific. If your answer spends a minute describing the company but never says what you did, shorten the setup and add your decision. If it claims success without evidence, state what is known and what was not measured. Then rehearse again with a new follow-up so you are practising understanding, not recital.

A useful test is whether the revision is relevant, factual, clear about ownership and natural when spoken. AI feedback is not an employer's scoring rubric unless the employer supplied that rubric. Finish with the facts and questions you want to remember, rather than memorising a wall of text.

Can an employer tell whether a CV was written with AI?

Not reliably from polished wording, generic phrases, punctuation or a detector score alone. These features do not establish authorship or prove that the experience is false. Asking another LLM for an authorship percentage is not independent verification.

Liang and colleagues' 2023 study found detector misclassification in its samples of non-native English writing. It studied English essays and detectors available at that time—not today's CVs, every language or every current product. Its relevance here is the demonstrated risk of false accusation, not a universal numerical accuracy claim. Do not use writing style or language background as a shortcut for honesty, competence or an adverse hiring decision.

• Understand generic AI phrasing and the limits of authorship clues

For employers: check the claimed experience with fair follow-ups

The more useful questions are whether a claim is supported, what the candidate personally contributed and whether they can discuss the work. A vague paragraph is a reason to clarify, not proof of AI use. Use published role criteria and your approved recruitment process, with consistent core questions and appropriate access arrangements.

Ask about decisions, constraints, checking and learning. Where your process permits it, a relevant work discussion or proportionate work sample can gather evidence. Clearly state permitted tools and respect confidentiality. Do not demand private chat histories, confidential client figures or an authorship confession. Humans remain responsible for assessment; these prompts do not score, rank, reject or select candidates.

Use the prompts for the stage you are actually in

Start with an evidence map if you have not drafted yet. Use the generic-phrase edit when the letter sounds unlike you. Reconcile your final documents before submission, and practise weak points before an interview. Recruiters should choose evidence questions, not an AI-detection exercise.

All examples need your actual facts and the employer's current instructions. No prompt guarantees an interview, a detector result or fair hiring by itself.

• Copy the job-application and interview prompts
• How to check an AI result