AI-related skills now appear in a growing share of job postings, and postings requiring them grew 144% year over year as of April 2026. In tech specifically, around 71% of US listings now mention them. So putting AI skills on your resume is no longer optional for a lot of roles.
The problem is that almost everyone is doing it badly, in one of two ways. Either they leave it off entirely, or they add "ChatGPT" to a skills list where it sits as a generic buzzword that tells a recruiter nothing. Neither gets you an interview, and one of them actively hurts you.
This guide covers the format that works, where each piece goes, and the single rule that keeps a resume claim from blowing up in the interview.
Key takeaways
- Vague AI mentions get skipped. Recruiters are now pattern-blind to a skills list containing tool names and nothing else.
- The format that works is: tool, application, result. What you used, on what, with what measurable outcome.
- Put keywords in the skills section for the ATS, and evidence in the experience section for the human. They are two different readers with two different needs.
- The one rule: only list what you can be questioned about. Every named tool is an invitation to a follow-up.
- Quantify the outcome, not the usage. Time saved, error rate, cycle length. Not "improved efficiency using AI."
- Do not invent a prompt engineering title. Describe what you did in the job you actually had.
Why most AI mentions get skipped
Put yourself on the other side. A recruiter opens a resume and sees:
Skills: Python, SQL, Excel, ChatGPT, Claude, Copilot, Midjourney, prompt engineering
This is now the most common formulation in the pile, which means it carries no information. Worse, it tells a careful reader that the candidate has opened some products, which is not a skill. Everyone has opened some products.
Compare it with a line in the experience section:
Cut the first-draft cycle for client reports from three days to one by building a prompt workflow with Claude, with a manual accuracy check on every figure before release.
Same underlying skill. The second version is specific, it names the constraint, it quantifies the outcome, and it signals that the person knows the output needs checking. That last part is what hiring managers are actually screening for in 2026, and it is nearly absent from resumes.
Where each piece goes
You are writing for two readers and they need different things.
The skills section is for the machine
Applicant tracking systems match on keywords, so a skills section still earns its place. Group the tools under a clear heading rather than scattering them:
AI & automation: Claude, ChatGPT, GitHub Copilot, prompt engineering, retrieval-augmented generation
That is it. Do not try to make this section impressive. It exists to pass a keyword filter, and our guide on beating ATS systems covers the broader mechanics of that.
One caveat worth taking seriously: match these to the actual posting. If the job description says "generative AI" and you wrote "GenAI," a crude matcher may miss it. Mirroring the posting's own vocabulary is the entire trick, and tailoring your resume to a job description is where the real gain is.
The experience section is for the human
This is where you get hired. One or two bullets, not five, using this shape:
[Tool] + [specific application] + [quantified result]
Worked examples across different kinds of role:
Reduced support ticket triage time 40% by building a classification workflow with GPT, with human review retained for anything flagged as billing or security.
Shipped a customer-facing FAQ generator using retrieval over our existing help center, cutting content maintenance from weekly to monthly.
Used Copilot to accelerate test coverage on a legacy service from 34% to 71% in one quarter, reviewing every generated case against the original ticket.
Notice what each of these has: a number, a named tool, and a boundary. The boundary is the part that reads as competence rather than enthusiasm.
The one rule that matters
Only list what you can be questioned about.
A named tool on a resume is an open invitation. If you list five and the interviewer picks the one you have used twice, and you cannot say what it produced or where it falls down, they will discount the other four immediately. This is the most common way an AI section actively costs someone an offer.
The test is simple. For every tool on your resume, can you answer three questions without hesitating?
- What did you use it for, specifically?
- What did it produce that you had to fix?
- Where would you not use it?
If you cannot answer all three for a tool, take it off. A shorter list you can defend beats a longer list you cannot. We covered the interview side of this in how to answer "how do you use AI in your work", and the resume should be written so that answer is easy to give.
Phrases to delete
These appear constantly and all of them weaken the document.
"AI-powered." Attached to your own work it is marketing language, and recruiters read past it. Say what the thing does.
"Leveraged AI to drive efficiencies." No tool, no task, no number. Three words of nothing.
"Prompt Engineer" as a self-assigned job title on a role where that was not your title. It reads as inflation and it is easy to check.
"Familiar with AI tools." Familiar is the word people use when they mean "not really." It actively lowers the reader's estimate.
"AI enthusiast." Enthusiasm is not a skill and this now reads as a substitute for evidence.
What to write if you have not used AI at work
Plenty of people are in workplaces that restrict these tools, or in roles where they have not come up. The honest options, in order of strength:
A real project. Something you built or automated outside work, finished, that you can describe end to end. This is nearly as strong as workplace experience for junior and mid roles because the question being answered is the same.
A concrete process improvement, even a small one. Rewriting how you research, draft, or review, with a before and after you can describe. Small and true beats large and vague.
Nothing. This is a legitimate choice. An empty AI section is better than a fabricated one, and a fabricated one is very easy to expose in an interview. If the role requires it, invest a weekend in building something real rather than a paragraph in describing something that is not.
What you should not do is list tools you have opened once. That is the version most likely to be caught.
A note on honesty
There is a temptation in this area to describe intent as accomplishment, because everything is new and nobody expects you to have five years of it. Resist it specifically here, for a practical reason rather than a moral one.
AI claims are unusually easy to test. An interviewer can ask you to walk through a workflow in ninety seconds, and the difference between someone who has built one and someone who has read about it is immediately audible. Behavioral claims are hard to verify. This category is not.
Write the version you can defend. It will be shorter than you would like, and it will hold up.
FAQ
Should I list ChatGPT as a skill?
Only inside a grouped skills section for keyword matching, and only if you can discuss how you use it. On its own, as a bullet, it carries no signal. The value is in the experience section where you show what you did with it.
Is prompt engineering still a real skill to list?
Yes, as a listed skill alongside the tools, and it remains common in postings. What has changed is that listing it without an example attached no longer differentiates, because so many resumes now include it. The example is what does the work.
What if my industry does not use AI at all?
Then leave it off. Forcing it into a resume for a role that does not ask for it wastes space and can read as unfocused. Match the resume to the posting.
Do recruiters actually care, or is it a filter?
Both. It is often a keyword filter at the ATS stage, then a genuine assessment criterion later, with employer surveys placing AI fluency at the top of what they weigh. That is why the two-part approach matters: keywords to get through the filter, evidence to survive the human.
How do I quantify AI work when the result is hard to measure?
Use time and volume rather than money if the financial impact is not yours to claim. Cycle time before and after, number of items processed, coverage percentage, error rate. If nothing is measurable, describe the scope precisely instead, which is still better than an unquantified claim.
Check what your resume actually says about you
The gap between a resume that lists tools and a resume that shows judgment is usually four or five lines, and it is very hard to see in your own document.
Round Zero scores your resume against a specific job description and tells you which lines are weak and why, including where a claim is too vague to survive a follow-up. The first check is free. If your AI section is the part you are least sure about, that is exactly the kind of line it is built to catch.