When you use AI to get work done, do you actually get better at it, or just borrow its skill?

This explores whether working with AI leaves you with skills you keep, or whether it just gets the task done while you stay the same. The corpus leans toward delegation. In one study, workers using generative AI did much better on content tasks, but when they did similar tasks alone afterward, they showed no improvement Does AI assistance help workers learn lasting skills?. Another line of research calls this an 'exoskeleton'. You produce skilled-looking work while the AI is present, then drop back to baseline when it's removed Does AI assistance build lasting skills or temporary abilities?.

The likely cause is what the AI takes over. Learners without AI ran into more errors and fixed them on their own, and they retained more skill. AI-assisted learners handed debugging to the AI, and those who debugged most with AI scored lowest on skill tests Does AI assistance remove a core learning channel through error work?. But the tool alone doesn't decide the outcome. In a randomized trial of developers learning a new library, six ways of using AI split into two groups. Three low-engagement patterns scored 24-39% on a quiz. Three that included active comprehension steps scored 65-86% Does AI assistance actually harm the way developers learn?. It's the same tool, and what changes is whether you're still doing the thinking.

The timing matters too. AI productivity gains show up when people apply skills they already have. When workers used AI to learn something new, the gains disappeared and learning suffered When does AI actually boost worker productivity?. So AI tends to amplify what you know, and it's weakest at the moment you're trying to acquire something.

Most people don't expect the next part: you probably can't feel the difference. Researchers call it the LLM Fallacy. People fold fluent AI outputs into their sense of what they can do, and come to believe they have skills they don't Do AI-assisted outputs fool users about their own skills?. It's a self-perception error, separate from hallucination or over-trusting the AI. It happens whether or not the output is accurate, so making AI more accurate or forcing verification won't fix it. What helps is making clear who contributed what How does AI-assisted work reshape how people see their own abilities?. The error thrives on smooth output, because the seam between you and the machine disappears.

Even help that is correct has costs. Well-meant AI suggestions can break the cognitive immersion that reasoning depends on, and you then have to rebuild your focus Does AI assistance always help reasoning or does it carry hidden costs?. What people ask for also often differs from what the AI does. In 40% of 200,000 Bing Copilot conversations, users wanted information or writing help while the AI coached and advised Why does AI default to coaching instead of doing?. The short version: by default AI delegates the task, and it builds skill only when you keep the hard part for yourself.