At first, AI chatbots can make people feel more productive, confident, and even smarter. But many users report a later hangover: memory lapses, duller writing or coding skills, and a reduced ability to solve problems or remember facts without AI’s help.
Susette Brooks, a marketing director in Philadelphia with an MFA in nonfiction writing, began using AI reluctantly and then let it draft some of her emails. The workload lightened, she said, but “I felt my writing and thinking decline as it was happening. I had a harder time recalling information and synthesizing ideas.”
What Brooks experienced is cognitive offloading: using external tools to reduce the mental effort required for a task. Calculators and calendars are forms of offloading, but tools like Claude and ChatGPT do that work at an unprecedented scale: they can do math, remember appointments, and write essays or presentations for you.
Cognitive offloading isn’t inherently bad — it helps humans be productive — but outsourcing more cognitive tasks can make you forgetful or weaken certain skills, a process sometimes called de‑skilling. The term has resurfaced to describe perceived declines in competency linked to widespread AI use. De‑skilling can affect critical thinking, writing, creativity, and memory by making it harder to acquire and retain those abilities.
Key takeaways
- AI can boost productivity while quietly reducing independent capability.
- Offloading tedious tasks is less likely to erode core skills.
- The long‑term fix is not less AI but more intentional use of it.
Young people are especially worried. A Gallup and University of Pennsylvania survey found 68 percent of Gen Z adults in the US fear offloading too many tasks will prevent them from building skills; similar shares worry AI discourages critical thinking and social learning. A Pew Research Center study in June reported 37 percent of US adults under 30 expect AI to have a negative personal effect. Still, more than half of Gen Z Americans use AI weekly.
That paradox—knowing AI may harm skills but using it anyway—stems from human tendencies to avoid taxing cognitive work, says Daniel Willingham, a psychology professor at the University of Virginia: “Our mind is really set up to save you from having to think. Thinking is what you do when you’re in trouble.”
Experts, however, say you can use AI to save time while protecting your cognitive abilities. Their advice centers on intent and strategy.
Too much AI use can erode skills over time
Derek Salyers, a content creator in Ohio, noticed his output had become what he called “AI slop,” and that he was less mentally present. “When you get used to the AI making the decision, you stop pausing to actually think about things,” he said. “I wasn’t forgetting things because my memory got worse; I was forgetting because I was never really thinking in the first place.”
Studies show similar patterns in classrooms. One study of nearly 1,000 high school students found those using ChatGPT scored 48 percent higher on a math test while they had access to the tool, but later—when tested without the chatbot—their scores were 17 percent lower than students who never used ChatGPT. Longitudinal data from UC Irvine also showed students spent less time and did better on text‑based questions after ChatGPT’s introduction, yet performed worse on offline tests without AI help.
These findings suggest short‑term gains from offloading often come with long‑term costs to independent thinking and skill retention.
Evan Risko, a psychology professor at the University of Waterloo and co‑author of foundational research on cognitive offloading, notes that offloading can extend imperfect memory systems but risks compromising performance when it replaces important mental processes. Basic cognitive abilities like working memory and selective attention are resilient, he says, but learned skills—arithmetic or piloting, for example—require acquisition and practice; AI can undermine both.
Brooke Macnamara, a psychology professor at Purdue who co‑authored the paper “Is AI making us stupid?,” warns people may not notice skills decaying because AI makes their performance look good. “They may not realize that if the AI tool were unavailable how low their independent performance might be,” she said.
Be intentional about which tasks you hand to AI
Before using AI, ask whether you should. Anjali Singh, a human‑AI interaction researcher at the University of Texas at Austin, says outsourcing tasks like writing removes opportunities to practice independent thinking and improve writing skills. “We keep outsourcing these kinds of skills to AI over time and do not engage in those activities ourselves, then the de‑skilling occurs,” she said.
Counterintuitively, it’s often best to use AI for tasks you’re already good at. A working paper by MIT economist David Autor and colleagues gave AI tools to 133 patent lawyers; while all lawyers saw short‑term gains, younger lawyers lost those gains when tools were removed, whereas experienced lawyers improved afterward compared with peers who never had AI access.
A Middlebury College experiment found two user types among undergraduates who wrote essays with or without AI: augmentation users, who used AI to develop their thinking, and automation users, who let AI do the work. The augmentation group performed best when later tested without AI.
“Using AI to deepen or develop one’s own thinking on or understanding of a topic, rather than letting AI do all of the thinking, seems like the best way to use these tools,” said Lauren Richmond, a cognitive science professor at Stony Brook University.
Tell your chatbot to treat you like a student
AI can overwhelm learners with immediate, dense answers that short‑circuit reflection. Older research methods—reading books or using search engines—often paced learning, giving time to process. “We are not getting these moments to reflect,” Singh said.
If you’re trying to learn, set up your AI to help you practice rather than just supply answers. Use custom instructions in ChatGPT or equivalent personalization features in Claude to say you want to build skills and be challenged. Study‑focused tools—ChatGPT’s study mode, Claude for Education, or Google’s Gemini for Students—are designed to act more like tutors by asking questions, prompting problem solving, and encouraging you to arrive at solutions yourself.
Slow down and embrace friction
AI’s speed is seductive: tasks like prototyping code can take minutes instead of days. But rushing increases the risk of offloading critical thinking. Joei Chen, a former brand director, described heavy workloads pushing people to dump work into AI and rely on its judgment: “So you just throw stuff into AI and say, ‘What do you think?’”
Faster output doesn’t always mean better results or happier workers. Slowing down restores agency and preserves thinking skills. Make AI interactions collaborative: request multiple draft options, compare them, and decide which is best. Comparing outputs activates critical thinking, judgment, and metacognition—thinking about your own thinking.
Advait Sarkar, an AI researcher at the University of Cambridge, says comparison improves judgment and awareness of knowledge gaps. Metacognition repeatedly came up in expert advice: many users let it fall away when chatbots stream content.
Matt Beane, an associate professor at UC Santa Barbara and author of The Skill Code, calls an unquestioning approach to AI “your worst enemy.” To avoid that trap, he and other experts urge users to introduce friction: experiment aggressively, ask for feedback, hints, examples, explanations, practice questions, and other prompts that engage your thinking.
A practical test for whether you’re offloading too much is simple: stop using AI for a time and try the task yourself. Spend a day writing emails without AI, use traditional search for research, or build an app with your own code. You may find AI made you worse at something. That’s not permanent—you aren’t dumber than before—and you can rebuild skills by reading, writing, practicing arithmetic, or simply thinking more often.
Source: vox.com
