47% of job seekers say they've actively built AI skills in the past six months, up from 41% a year ago, and 30% taught themselves with no employer support at all, according to iCIMS' September 2026 workforce report. But 61% of those same workers describe their AI proficiency as limited to general-purpose tools like ChatGPT. Most hiring processes still can't tell the difference, and that gap is where a lot of disappointing AI-skills hires start.
Workers are outpacing employer training, and the numbers show it
iCIMS' report, drawn from 3.1 million platform users and 691 million candidate profiles plus a separate survey of 1,000 US job seekers, found candidates are building AI fluency largely on their own initiative. Employer-provided AI training hasn't kept pace: roughly 1 in 6 workers report getting any at all, a share that hasn't moved in a year.
What "AI skills" actually means to a candidate versus a job posting
The self-reported numbers hide a wide range of actual capability. Of workers who say they've built AI skills, 61% describe their proficiency as limited to general-purpose tools like ChatGPT, Copilot, or Gemini. Only 18% report prompt-engineering skills, and just 17% report machine learning or model-development capability. Meanwhile, 45% of job seekers say they've encountered generative AI requirements in postings for roles they'd consider. "AI skills" on a resume and "AI skills" in a job description are frequently not describing the same thing.
Why job postings keep asking for something they aren't testing for
Even with AI skill requirements showing up in nearly half of relevant postings, AI-related roles still make up only about 4% of total US hiring demand. Most companies aren't hiring AI specialists, they're asking general roles to have some baseline AI fluency, without a shared internal definition of what that baseline is or a way to verify a candidate actually has it beyond a checked box or a keyword on a resume.
Workers are outpacing employer training. Job postings are outpacing both. That is the market talking.
Trent Cotton, Head of Talent Insights, iCIMS
The practical takeaway: if "AI skills" appears in a job requirement, someone in your hiring process should be able to say exactly which tier of skill the role actually needs, and how the interview verifies it. Otherwise it's a keyword, not a qualification.
Building a rubric instead of a keyword requirement
- Define which AI-skill tier a role genuinely needs, general-purpose tool fluency, prompt engineering, or model development, before writing the job posting, since these are very different bars and most candidates only clear the first one.
- Replace self-reported AI proficiency with a short, structured work sample or scenario question scored against that tier's rubric, rather than taking a resume claim at face value.
- Treat employer-provided AI training as a real recruiting lever, not just a retention perk. 42% of job seekers say it makes an employer more attractive, and 14% say they'd accept lower pay for it.
- Audit current job descriptions for generic "AI skills" language and either sharpen it to a specific, testable requirement or remove it if the role doesn't actually need it.
The bottom line for hiring leaders
Candidates are building AI fluency faster than employers are training for it or screening for it, and that mismatch is currently running on the honor system. The hiring teams who define what "AI skills" actually means for a given role, and build a real way to check it in the interview, will make better hires than the ones still treating it as a resume keyword.