What Should You Never Let AI Write on Your CV?

AI is good at structure and phrasing. Hand it a messy list of responsibilities and it will hand back clean, readable bullet points in seconds. That part is a genuine time saver, and you should use it.

There are five specific things it should never write for you. Get these wrong and the damage is not cosmetic; it can cost you the interview, the offer, or in a small number of cases the job itself after you have started.

The Dividing Line: Facts Versus Phrasing

The simplest rule to carry around is this. AI can decide how something is said. Only you can decide what is true.

Every failure mode below is a version of the same mistake: letting a language model supply a fact it has no way of knowing. Large language models are built to produce plausible text, not verified text. When a detail is missing, they fill the gap with something that reads well.

1. Your Contact Details and Personal Data

This sounds obvious, and yet it is the most common practical failure. People paste an AI-generated CV into a document and never check the header, because the header is boring and the bullet points are exciting.

AI tools sometimes reformat phone numbers into US conventions, drop the leading zero from a UK mobile, or invent a placeholder email such as name@email.com. Check every character of your name, phone number, email address, and location by hand.

Your CV also does not need your National Insurance number, date of birth, marital status, nationality, or a photograph. UK employers do not expect them, and including them invites unconscious bias into a process where you want none. The Information Commissioner’s Office is clear that employers should only collect personal data they actually need, and the same discipline applies to what you volunteer.

There is a practical reason too. Your CV does not stay in one inbox.

An incorrect phone number in a database is a missed call you never know about. A National Insurance number in that same database is a risk with no upside.

2. Specific Quantified Achievements

Numbers are the single most persuasive thing on a CV, which is exactly why AI is so willing to supply them. Ask a general chatbot to strengthen a bullet point and it will cheerfully offer “increased team efficiency by 30%” without having the faintest idea whether that happened.

Do not let a number reach your CV unless it came out of your head or your records. If a hiring manager asks how you arrived at 30%, “the AI suggested it” is not a recoverable answer.

The right use of AI here is as an interviewer, not an author. Ask it what kinds of metrics your role typically produces, then go and find your own figures in old appraisals, dashboards, handover notes, or team reports. Once you have the real number, AI is genuinely useful for phrasing it well.

If you want the fuller method for this, our guide on the best prompts for AI CV writing covers how to ask for quantification without inviting invention.

3. Named Clients, Projects, and Confidential Work

AI does not know what you signed. It has no view on your NDA, your former employer’s client confidentiality policy, or the fact that the project you are proud of was never announced publicly.

Three rules cover most situations:

• If the client relationship is publicly known, naming it is usually fine.
• If it is not public, describe the client by sector and size instead: “a FTSE 250 retail group” rather than the name.
• If you are in doubt, leave it out and save the detail for the interview, where you can judge the room.

This applies with particular force in the public sector, in legal and financial services, and in defence, where the wrong project name on a document that gets emailed around is a disciplinary matter rather than an awkward moment.

4. Salary Expectations and Notice Periods

AI has no idea what you are worth in your city, in your sector, in this month’s market. It also does not know your notice period, your visa status, or whether you would relocate.

Most UK CVs should not carry salary expectations at all. Naming a figure before you have had a conversation either prices you out or anchors you low, and there is rarely a version where it helps you. If an application form demands one, answer it there, based on your own research from sources such as Prospects or sector salary surveys, not on an AI guess.

Notice periods are a related trap. An invented “available immediately” on a CV you have not read closely can start a relationship with an inaccuracy, which is a poor way to begin.

5. Anything Sensitive

Health conditions, disability, religion, sexuality, political affiliation, and family circumstances are yours to disclose, on your terms, at a time of your choosing. AI has no judgement about timing or context, and it will sometimes weave a detail you mentioned in passing into a personal statement.

Employment gaps sit slightly differently. You should account for a gap, because unexplained time reads worse than explained time. What you should not do is let AI write the explanation for you, because the reason belongs to you and the wording needs to sound like a person, not a press release.

The same caution applies to references. Never let a tool generate a referee’s name, job title, or contact details, and never list anyone who has not agreed to it.

What AI Should Be Doing Instead

None of this is an argument against using AI on your CV. It is an argument for giving it the right job.

AI is strong at structure, ordering, and consistency. It is very good at turning “responsible for the rota” into an active, specific bullet point once you have supplied the facts. It is good at spotting where your CV and a job advert use different words for the same skill, and at cutting a rambling paragraph to three tight lines.

It is also good at a first-pass check on whether your CV will get past both AI and human screening, because format problems are pattern problems, and patterns are what these tools see best.

Where it needs supervision is voice. An AI draft that is factually correct can still read like it was assembled by committee, and fixing that is a human job. Our guide on making an AI-written CV sound like you covers the specific edits that put your own register back in.

A Two-Minute Check Before You Send

Run this every time, regardless of which tool produced the draft:

• Read the contact block character by character. Phone, email, name, location.
• Highlight every number on the page. For each one, say out loud where it came from.
• Search for client and project names. Ask whether each is public.
• Confirm there is no salary figure, NI number, date of birth, or photograph.
• Read the personal statement aloud. If a sentence is not something you would say, rewrite it.

Five minutes of checking protects an hour of drafting, and it is the difference between AI as a useful assistant and AI as a liability you have signed your name to.

A Faster Way

Running a general chatbot with a careful checklist works. Plenty of people get good CVs that way, and if you enjoy the control, carry on.

Zappli was built to remove the checklist. It never asks a model to invent your figures, it keeps personal data out of the document by default, and it prompts you for the facts rather than filling them in for you. Start with a free ATS diagnosis and one rewrite, no card needed. If you want the full set, a 7-day Pro-Pass is £7.99, Pro-Monthly is £11.99, and the Agent at £24.99 a month scans the UK market and drafts applications overnight.

Zappli is paid by candidates, never by employers, which is the only arrangement where the tool is working for you. If a free diagnosis sounds more useful than policing an AI’s guesses line by line, the link below is the place to start.