AI Skills for Your Resume: What to List and How to List Them (2026)
Which AI skills belong on a CV, what the UK vacancy data and employer surveys actually show, and eight AI-skill bullets each paired with the question a sceptical interviewer would ask about it.
You use an AI tool most days and have no idea whether it belongs on your CV, or what to call it if it does. Put it in a skills list and it looks like padding; leave it out and you wonder whether the employer wanted it. This page covers what UK employers are advertising for, and gives eight AI bullets paired with the question a sceptical interviewer would ask about each one. (UK employers usually say CV where this page says resume.)
What the evidence says, and what it does not
The figure most often quoted comes from Microsoft and LinkedIn's 2024 Work Trend Index, published on 8 May 2024 from a survey of 31,000 knowledge workers across 31 markets: 66% of leaders said they "wouldn't hire someone without AI skills", and 71% said they would "rather hire a less experienced candidate with AI skills than a more experienced candidate without them". Worth knowing before you build a CV around it: that is what leaders told a survey, run by two companies that sell AI products, not a record of who got hired.
The UK advertising data is more sober. The Department for Science, Innovation and Technology's AI Skills for Life and Work: job vacancy analysis, published 28 January 2026, looked at UK job postings from January 2021 to December 2023 and found 448,484 of them AI-related — 1.7% of all UK postings. Demand fell in 2023 rather than rising: postings for the most technical group, which the study calls AI Experts, were down 47% year on year. In those Expert roles the skills employers asked for were Python (68%), data science (64%), machine learning (63%), SQL (29%), AWS (18%) and Azure (11%). Median advertised pay for them was £62,700, a 42% premium over wider IT roles, and 99% required at least a bachelor's degree.
The companion AI Labour Market Survey 2025, carried out for DSIT by Gardiner & Theobald and published the same day, asked employers where the gap is. The biggest one is understanding: the share naming "understanding AI concepts and algorithms" as a skills gap rose from 55% to 60% over five years. Where candidates fell short, employers named missing work experience (31%) and insufficient technical skills (30%). Apprenticeships went from 3% of AI hires in 2020 to 19% in 2025, and 88% of organisations train on the job.
Two conclusions I would draw. If you are going for an actual AI role, the thing being bought is engineering and statistics, and your CV should read like an engineer's. If you are going for anything else — which is almost everyone — AI use is a supporting claim that makes an existing strength more credible, not the strength itself. For what employers report they are short of more broadly, the skills and action verbs guide covers the Employer Skills Survey and the Future of Jobs Report figures.
On claiming tools you have barely used
My position is flat: if you have not used a tool on real work with consequences, it does not go on the CV. Not in a smaller font, not under "familiar with", not in a list of twelve where you hope nobody asks about the fourth one. Prospects puts the practical version plainly: "Don't exaggerate your abilities, as you'll need to back up your claims at interview." It also notes that lying on a CV is a criminal offence, which is a different and larger problem than being caught out.
AI claims are unusually easy to test, for two reasons. The interviewer often uses the same tools you do, so there is no specialist vocabulary hiding the gap. And the interesting questions are not about features. Nobody asks which button you pressed; they ask what the tool got wrong and how you noticed. Somebody who has genuinely used a model for six months has three answers to that ready. Somebody who has read about it has none.
The test before a tool goes on the page: can you name one thing it consistently gets wrong in your work, one change you made to your process because of that, and one person other than you who used what you built? Three yeses, it goes on. Two, it stays in the experience bullet without a tool name. Fewer, leave it off.
Eight AI bullets and the question that follows each
All of these are invented examples, with invented numbers, written to show the shape. The point is the pairing: write the bullet, then write down the question a sceptical interviewer would ask about it. If you cannot answer the question, the bullet is not ready.
1. Customer support
Bullet: "Wrote and maintained a library of 14 reply templates for the team's AI assistant, covering refunds, delivery delays and account access, cutting average first-reply drafting time from 12 minutes to 4 across a team of 15."
The question: "Who checked those replies before a customer saw them, and what happened the first time one of them said something wrong?"
What you need: the review step, the specific failure (a refund window quoted wrongly, say), and what you changed in the template afterwards. Without that, the bullet reads as a time saving nobody checked.
2. Marketing and content
Bullet: "Drafted 40 product descriptions a month with an AI writing tool and rewrote each against the brand guide, freeing about a day a week for campaign work."
The question: "What does it reliably get wrong about your products, and how would I spot it?"
What you need: two concrete recurring errors — invented materials, a certification the product does not hold, sizes that do not exist — and the check that catches them. This answer is also the one that shows editorial judgement, which is the actual skill being bought.
3. Finance
Bullet: "Built an Excel and Copilot workflow that drafts monthly variance commentary from the management accounts, reviewed line by line before it enters the board pack."
The question: "What happens when the commentary is confidently wrong about a variance?"
What you need: the boundary you drew. The tool drafts sentences; the numbers come from the ledger and you tie them back. A finance interviewer is testing whether you know which half of that job cannot be delegated.
4. Data and insight
Bullet: "Coded 3,000 free-text survey responses into 11 themes using Python and an LLM, hand-checking a 10% sample against my own coding and reporting the disagreement rate alongside the findings."
The question: "What was the disagreement rate, and what did you do with the responses you disagreed on?"
What you need: the number, and the decision it led to. Publishing your own error rate is the part that makes a research team trust the rest.
5. Software engineering
Bullet: "Added retrieval-augmented search over eight years of policy documents using the OpenAI API and a vector store, with an evaluation set of 120 questions run against every release."
The question: "Talk me through the evaluation set. How did you decide an answer was wrong?"
What you need: how the questions were chosen, who wrote the expected answers, the threshold for shipping, and one failure mode you never fully fixed. Engineers listening for depth are listening for that last part.
6. HR and recruitment
Bullet: "Drafted job adverts with an AI tool and checked each against the inclusive language checklist and the pay wording agreed with legal, across 60 vacancies."
The question: "What did you stop it doing?"
What you need: examples you rejected — invented essential criteria, a salary range it guessed at, wording that narrowed the field for no reason. In HR the value is the veto, not the draft.
7. Operations and administration
Bullet: "Automated the weekly supplier report with Power Automate and Copilot, replacing two hours of manual copying, and documented it so two colleagues can run it without me."
The question: "What breaks it?"
What you need: the known failure — a supplier renaming a column, a file arriving late — and how anyone finds out it has failed. An automation nobody can fix when you are on holiday is a liability, and an operations manager knows it.
8. Changing career or self-taught
Bullet: "Built a question-answering tool over 200 pages of my own study notes in Python, and wrote up the cases it answered wrongly and why, in a public repository."
The question: "What did it get wrong, and what would you do differently if you built it again?"
What you need: nothing more than the honest answer, which is why this is the strongest bullet available to someone with no commercial AI experience. The write-up is the evidence, and it is the part almost nobody does. The career change CV guide covers placing work like this when your job titles do not match the target role.
Where it goes on the page
One line in the skills section for the tools, the real evidence in the experience bullets, and a mention in the profile only if the advert asks for it. Do not create a standalone "AI" section unless you are applying for an AI role; it separates the tools from the work they were used on, which is the opposite of what you want.
Skills Analysis: SQL, Python (pandas), Power BI Automation and AI: Power Automate, Microsoft Copilot, OpenAI API (internal tools) Project: Agile (Scrum), Jira
Use the advert's own words where they are true of you: if it says "Copilot", write Copilot rather than "AI productivity tools". The tailoring guide covers pulling those terms out of an advert, and using ChatGPT to write a CV covers the drafting side, which is a separate question from what you claim.
Leave these off
- "AI literacy" and "familiar with AI tools." Neither states anything a reader can check.
- A list of twelve models. Six months on one tool beats a week on each of twelve, and the list invites a question about the one you have used least.
- "Prompt engineer" as a self-applied title, unless the advert uses it. Describe the work instead.
- A percentage you cannot source. "Improved productivity by 60%" with no method behind it is the first thing an interviewer picks at. Scale works instead: how many, how often, who else used it.
- The AI tool in place of the underlying skill. "AI-assisted data analysis" without "data analysis" reads as the tool doing the job.
If your current employer has not settled its position
Plenty of people are using these tools in workplaces that have not decided what they think. The same Work Trend Index found 52% of people using AI at work were "reluctant to admit to using it for their most important tasks", so this is common rather than unusual. It still does not license a vague CV. Write the bullet about the work and the result, name the tool only where you were allowed to use it, and keep the detail for the interview, where you can give the context a bullet cannot hold.
Sources
- Microsoft and LinkedIn: 2024 Work Trend Index Annual Report, "AI at Work Is Here. Now Comes the Hard Part" (8 May 2024)
- Department for Science, Innovation and Technology: AI Skills for Life and Work — job vacancy analysis (28 January 2026)
- DSIT and Gardiner & Theobald: AI Labour Market Survey 2025, executive summary (28 January 2026)
- Prospects: How to write a CV
Frequently Asked Questions
Should I put ChatGPT on my CV?
Only if you can describe a piece of work it changed and answer questions about what it got wrong. The tool name on its own says nothing: almost everyone applying has opened it. A bullet saying what you built with it, who else used it and what you checked before the output went anywhere is worth a line; "ChatGPT" in a skills list is not.
Is prompt engineering a real skill employers hire for?
Rarely under that name. The Department for Science, Innovation and Technology's analysis of UK job adverts found the skills most often asked for in the most technical AI roles were Python (68%), data science (64%) and machine learning (63%). Outside AI roles, I would describe the work you did rather than claim a job title that few UK adverts use.
Do employers actually care about AI skills?
The stated appetite is much larger than the advertised demand. In Microsoft and LinkedIn's 2024 Work Trend Index, 66% of leaders said they would not hire someone without AI skills and 71% said they would rather hire a less experienced candidate who had them. But DSIT's analysis of UK job adverts from 2021 to 2023 found only 1.7% of all UK postings were AI-related. Treat AI as something that strengthens your existing case, not as the case itself.
How do I list AI skills if I am self-taught?
The same way you list anything else self-taught: by what you have produced. A built thing you can show and talk about carries more weight than a course completion certificate, and it answers the question interviewers actually ask, which is what the thing got wrong and how you found out.
Should I say I used AI to help write my CV?
No. The CV is read as your account of your own work, and how you drafted it is not a claim about your abilities. What matters is that every line is true and that you can talk through any of it. If the drafting changed a fact, fix the fact.
What if my employer has not allowed AI tools?
Then do not invent AI experience to fill the gap. Write the bullet about the work: the process you improved, the analysis you ran, the thing you automated with the tools you were allowed. An interviewer can tell the difference between someone who has used a tool and someone who has read about one.
Keep reading
Skills and Action Verbs for a Resume: 200+ Examples by Category
200+ resume skills and action verbs by category, what UK employers report they are short of (DfE Employer Skills Survey 2024, WEF Future of Jobs Report 2025), how to list skills and languages on a CV, and one bullet rewritten from weak to strong.
Read articleBest AI Resume Builder in 2026: How the Top Tools Compare
Which AI resume builder actually gets you more interviews? We compare the top tools — Jobscan, Teal, Rezi, Kickresume and more — on features and price.
Read articleHow to Use ChatGPT to Write Your Resume: 18 Prompts and Where It Falls Short
How to use ChatGPT to write your resume (a CV in the UK): 18 copy-ready prompts, each paired with what to check in the output, a worked example of correcting a bullet, and the privacy settings and personal details to keep out of the chat.
Read article