How to Use AI to Write a CV Without Inventing Anything

Use AI for wording, not facts: a step-by-step way to write your CV with AI, check every detail, protect your privacy and stay honest with employers.

AI chatbots are genuinely useful for CV writing. They can turn a clumsy sentence into a clear bullet, trim a long paragraph and catch typos you have read past a dozen times. They can also add a result you never achieved, a tool you never used or a title you never held, and phrase it so smoothly that it is easy to miss. On a document you may be asked to confirm as true, that is a serious problem.

This guide explains what AI does well on a CV, where it goes wrong and why, and a step-by-step workflow that keeps every fact yours: you supply the facts, the AI helps with the wording, and you check everything before you send it. It also covers privacy, tailoring your CV to a job, and how to answer if an employer asks how you used AI.

What AI does well on a CV

Treat a general chatbot such as ChatGPT, Gemini, Copilot or Claude as an editor, not an author. Harvard's Mignone Center for Career Success describes that role well: AI can help you brainstorm better versions of your bullet points, work in relevant keywords from a job description and improve what you already have, but it should not be the main author, partly because its output tends to be generic.

The tasks where it helps most:

• Rewording. Turning a passive line such as "was responsible for the preparation of monthly reports" into a direct one such as "Prepared monthly sales reports for the regional manager".
• Cutting. Reducing a four-line description to two lines without losing the point.
• Structure. Splitting a block of duties into separate bullets, or suggesting a clearer order for a section.
• Matching language to a job. Comparing your CV with a job description and pointing out which of your real skills and experiences answer what the employer asks for.
• Proofreading. Catching spelling and grammar errors, mixed tenses and inconsistent date formats. The UK Civil Service's guide for candidates counts checking your spelling, grammar and clarity among the acceptable uses of AI.

All of these tasks have one thing in common: the AI works on material you have already given it. That is the safe zone.

Where AI goes wrong: details that sound right but are not true

A language model does not know your work history. It generates likely-sounding text based on patterns in the data it was trained on. The US National Institute of Standards and Technology calls the resulting risk confabulation in its Generative AI Profile (NIST AI 600-1), defining it as "confidently stated but erroneous or false content", and explains that it follows naturally from the way these models are built rather than being a rare glitch. The Saudi Data and Artificial Intelligence Authority (SDAIA) makes the same point in its generative AI guidelines for the public: text models can get information wrong, and it is up to users to verify what they produce against reliable sources.

On a CV, confabulation usually looks like this:

• a number nobody measured, such as "increased sales by 30%";
• a tool, method or certification you have never used or earned;
• a quiet promotion, with "Senior", "Lead" or "Head of" added to your title;
• a bigger role than you had, with "supported" becoming "led", or "a team" becoming "a team of 12";
• a date, employer name or degree detail changed while the text was being tidied.

These errors are hard to spot precisely because they read well. The polish that makes a real achievement sound strong also makes an invented one sound true.

Example (hypothetical). Original bullet: "Handled customer complaints by phone and email." AI suggestion: "Resolved 95% of customer complaints within 24 hours, raising satisfaction scores by 20%." Both figures are invented; the original contains no numbers at all. A safe rewrite stays with what happened, for example: "Handled customer complaints by phone and email, and wrote reply templates for the most common issues", but only if you really wrote those templates.

This matters because a CV is a factual record. The UK Civil Service tells candidates that AI tools must never be used to give misleading or false information during an application, and asks them to confirm that what they submit is true and accurate. Harvard's advice is just as direct: know your CV well enough to discuss any line of it in an interview. A figure you cannot explain becomes a problem the moment someone asks about it.

A safe workflow: your facts first, AI for the wording

1. Write a fact sheet before you open a chatbot

For each role, note the employer, your exact job title, start and end dates, what you actually did, and the results you can back up, with real numbers if you have them. Harvard's advice is to begin with what you already have: your current CV or detailed notes on past roles, including the difference you made and any relevant numbers. This fact sheet becomes your reference for checking everything the AI returns.

2. Work on one section or bullet at a time

A focused request is far easier to check than a whole CV rewritten in one go. Paste one bullet and the facts behind it, not your entire career.

3. Set the rules in your prompt

Tell the tool what it may and may not do:

Example prompt. "Rewrite the CV bullet below in clear, concise English, starting with an action verb. Use only the facts I give you. Do not add numbers, tools, job titles, dates or results that are not in my text. If a detail would make the bullet stronger but is missing, ask me for it instead of guessing."

Instructions like these reduce invented details but do not guarantee their absence, so step 5 still applies.

4. Ask for options, not a finished version

Request two or three versions and choose the one closest to how you would describe the work yourself. Harvard recommends treating AI text as a proposed edit, not a finished product, and reworking it until it describes your experience accurately and authentically.

5. Check every fact against your fact sheet

Go through the AI's version line by line and confirm:

• Names: employers, schools, clients and products.
• Titles and scope: your job title, the size of any team, and whether you led the work or supported it.
• Dates: start and end months and years.
• Numbers: every percentage, amount, count and time frame.
• Tools and credentials: software, methods, licences and certifications.

Anything that is not on your fact sheet either comes out or is replaced with the truth.

6. Put it back in your own voice

Read the final version aloud; Harvard suggests this as a way to spot repetition and inaccuracies. Swap out words you would never use in conversation. If English is not your first language, take particular care: the University of Edinburgh's careers service cautions non-native speakers against giving a false impression of their English. Your CV should sound like the person who will walk into the interview. Our guide to writing your CV in English when Arabic is your first language goes further.

7. Make sure you can talk about every line

Before you send the CV, go through it one last time and ask of each bullet: could I explain this with an example if an interviewer asked? If the answer is no, rewrite it or remove it.

Protect your privacy, and other people's

Treat anything you paste into a chatbot as information you may not be able to take back. Edinburgh's careers service warns that personal data you type in could become public, and SDAIA warns that users can reveal sensitive information without realising it. NIST lists data privacy, including the leakage and unauthorised disclosure of personal information, among the risks that generative AI creates or makes worse.

Before you paste anything:

• Remove your own identifiers. Your phone number, email address, home address and ID or passport numbers are not needed to improve a sentence.
• Leave other people out. Do not paste referees' names and contact details, colleagues' or managers' names, or details of clients, patients or students.
• Keep your employer's information confidential. The UK Civil Service tells applicants not to share sensitive details about their organisation, such as figures or personal information, with AI tools, and Harvard gives a similar warning about personal or proprietary data.
• Generalise when context helps. Write "a regional bank" rather than the bank's name, or "a hospital client" rather than the hospital.
• Check the tool's settings. Look at how the service handles your conversations, including whether they may be used for training, and follow your employer's rules if you use AI at work.

Check the result against the job

Tailoring is where AI is most helpful and where invention creeps in most easily: a job description lists skills, and an eager model may simply write them into your CV.

1. Compare line by line. Put the job description next to your CV and mark each requirement you genuinely meet, with the evidence.
2. Claim only what you can support. If you have not used a required tool, do not list it. Highlight your closest real experience instead, or address the gap honestly in your cover letter.
3. Use the employer's words where they are true. If you did the work but described it differently, matching the advert's wording is fine, and Harvard recommends using the job description's keywords where they fit naturally in context. Adding a skill you do not have is not tailoring.
4. Read it as a recruiter would. Every claim should have evidence behind it, and you should be comfortable being asked about each one.

Our guide to how applicant tracking systems read a CV explains how to use a job's keywords without stuffing them in.

Be open about how you used AI

Rules differ between employers. Harvard advises reading every application instruction, since some organisations set their own rules on using AI to prepare applications, and Edinburgh notes that some employers say openly that they will not accept AI-generated applications. Where a policy exists, follow it.

Interviews are different again. The UK Civil Service accepts AI as a way to prepare for interviews, but not to generate answers during one. If someone asks how you wrote your CV, a plain answer works: you used an AI tool to help with the wording, every fact in it is yours, and you are happy to talk about any line. That answer is only honest if you have followed the steps above.

How Tabhar helps

Tabhar's AI writing help is available on paid plans; the Free plan has no AI (compare plans). You can improve, shorten, expand or rewrite a section or a bullet. Each suggestion appears as a before and after, and nothing changes until you accept it. The AI's edit tools cannot change your contact details, dates, or employer and school names, and if a suggestion contains a number that is not in your CV, that number is flagged so you can check it.

If you are starting from a blank page, you can generate a starting draft from a job title or job description in which every personal fact is a bracketed placeholder for you to fill in, so it never invents your history. You can also tailor a CV to a saved job and accept or reject each suggestion, translate a whole CV between English and Arabic into a new copy while the original is kept, and run the free ATS checker, which uses rule-based checks (no AI) and shows how your CV matches a job title and description.

Related reading

• How to write a CV summary, with examples
• Common CV mistakes and how to fix each one
• Common interview questions and how to answer them

Sources