70% of graduate employers develop AI literacy in their early career programmes.

only 37% actually recruit for it.

That's from the Institute of Student Employers' latest survey. So the "put AI skills on your CV" advice is aimed at the wrong thing. Employers expect to teach you the tools. What they test is whether you can think with them.

The consensus: list your AI skills on your CV

The argument goes:

→ every job now mentions AI → so list ChatGPT, Claude and Copilot under "skills" → and you'll beat the filter

And there's data behind it. LinkedIn says US jobs requiring AI literacy grew 70% year on year (LinkedIn).

Students have clearly heard the message. In a Prospects survey of 5,000 students and graduates (ISE):

  1. 73% had used AI somewhere in the application process
  2. 46% had used it to write CVs or cover letters
  3. 95% believed employers use AI to shortlist

So everyone is using it, and everyone is writing it down. Which is exactly the problem.

What graduate employers actually say about AI skills on a CV

Look at what recruiters worry about:

  1. They don't trust the claim. 67% of ISE employers worry graduates use AI to misrepresent their skills, up from 52% a year earlier (Prospects Luminate)
  2. They can't see it in practice. GMAC's survey of 621 recruiters across 39 countries found AI tool skills are among those graduates are "least prepared to demonstrate" (The Tribune)
  3. They'd rather train it. Hence that 70% vs 37% gap

A line on a CV is cheap. Showing it is rare. That gap is your opening.

What AI hiring tests actually check

Interviews are turning into live AI tests, according to FT reporting this month:

→ McKinsey has added AI use to junior consultant interviews since late 2025. It says it's assessing judgement and how candidates use AI, NOT prompting skill or how many tools they know. Candidates can even bring their own tools → Michael Page now asks for "three ways you would use AI" in the role, or asks candidates to improve a weak prompt → ResiDesk had a marketing candidate build a landing page with Claude, then spent about 40 minutes discussing that output

Notice what none of them test: whether you've heard of the tool.

They test:

  1. can you spot what the AI got wrong?
  2. can you explain your process out loud?
  3. can you connect it to the actual job?

bad CV line: "Proficient in ChatGPT, Claude, Gemini" good CV line (example shape): "Used Claude to code 40 survey answers into themes, then checked them against the raw data and fixed two misreads"

The second one gives an interviewer something to dig into for 40 minutes.

What uni won't teach you about AI in hiring

→ uni AI policies focus on misconduct, so you learn to HIDE AI use, not explain it → group projects rarely ask you to show your working with a tool → careers advice still says "add a skills section", when employers are moving to live tasks

The critical-thinking move: 4 questions before you write "AI skills"

  1. What did I actually make? A tool name isn't evidence. An output is
  2. What did I check? Name one error you caught. That's the judgement McKinsey says it's testing
  3. Could I do it live in 30 minutes? If not, practise until you can
  4. Is it relevant to THIS job? Michael Page's question is about the role, not about AI in general

What to do this month

  1. Rewrite one CV bullet from "tools I know" to "thing I made, plus what I checked"
  2. Practise the "three ways" question for one job you're applying to. Say it out loud, timed
  3. Do one live task with AI, then write half a page on what it got wrong. Our job search toolkit has prompts to start from
  4. Keep an error log. We suggested the same habit to law students in our legal AI toolkit, and it works in any field

More on the hiring side in our HR and recruiting hub.

so TLDR: employers will teach you the tools. They're hiring for the judgement, so show it, don't list it.

If an interviewer asked you to name one thing AI got wrong for you last week, what would you say?

Sources

Get the next issue by email

theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.