The new study followed 26,811 Chinese high school students for 30 months. The study measured how generative AI affected their learning.
Good news: Students using AI saw homework scores increase by 18% and completion time decrease by approximately 30%. Better scores are done faster. These are results any teacher should be happy to see.
The bad news: These same students saw their closed-ended test scores drop by 20% within six months. Entrance exam scores fell by 18% to 24%, with the biggest drop occurring after about two years of AI use.
About 80 percent of student AI users demonstrated outsourcing thinking results. They completed the tasks very quickly. They did their homework well. But then, when the AI tools weren’t there, they underperformed.
These findings are also reflected in work settings. A 2026 survey by IT firm GoTo found that 39 percent of workers say their reliance on AI has already undermined their skills. Same dynamic. Various settings.
This raises a question that every parent, manager, and professional should ponder: What happens to your ability to think and achieve your goals when AI does the thinking for you? goals?
Effort is an important variable for learning
Psychologists call this cognitive load. This is what happens when you outsource your thinking to conserve your mental energy. Using a calculator or setting notes are all forms of cognitive overload, freeing up mental bandwidth for higher-order thinking. The strategy is detrimental when the task is specifically designed to promote learning, such as using artificial intelligence to complete homework or wind through an on-the-job training program.
Wharton professor Ethan Mollick has studied AI adoption Summarizing the main conclusion of a Chinese student study, he said that artificial intelligence “harms learning if it impairs mental activity.”
The brain develops capacity through struggle. When a problem arises, the mind draws on previous knowledge, makes new connections, and develops its type pattern recognition it cannot be formed any other way. This reasoning is formed when the answer is not clear and your brain needs to find the answer.
Distractors gradually disrupt learning and growth.
Therefore, the highest-achieving students suffered the greatest losses. They were able to outsource their work to AI. Nothing in their homework scores indicated a problem. In fact, their homework was of high quality because they taught artificial intelligence at a high level. Paradoxically, the students in the best position to use AI were also the most negatively affected.
The invisible timeline
What makes it difficult to capture the effect of cognitive load is its latency. Homework scores looked better than ever. But the decline in learning didn’t show up in the students for months, and it took nearly two years for the full effect to take effect.
The delay creates a feedback loop that makes the AI stronger learning problem. If the output appears strong, there is no clear signal that behavior should be changed. When the gap is obvious—for example, a bad test result, a job interview, or any moment that requires you to think on your feet—the decline in learning is already imperceptibly amplified.
MIT Sloan researchers have modeled the same dynamic in the workplace, calling it the “multiple trap”: short-term productivity achievements lose experience over time. In other words, workers who use AI will achieve great results initially, but reduce the chances of long-term success with AI. Data from a GoTo study shows that 46 percent of Gen Z workers believe that their reliance on artificial intelligence is weakening their skills. This is a big challenge given that many Gen Z workers are currently building their foundational education and making decisions about their future careers.
What leads to real learning
A study in China found that while still acting strongly, students who used artificial intelligence showed minimal learning loss. Using AI as a tool rather than a crutch allowed students to master and continue their learning.
When it comes to destroying AI learning, the antidote is for humans to work alongside AI in ways that will evolve over time with human capabilities. See AI as an input for thinking, not a replacement for thinking.
Zoom out and it becomes what the story is about.”intelligence“It actually is, and where it comes from. As I described in my last book, Experienced intelligencebeing “smart” IQ (intelligence), EQ (emotional intelligence) and wisdom taken from real life experiences. Experiences shape your thinking and abilities, resulting in the development of judgment, a skill that is more necessary than ever in today’s artificial intelligence-driven world. Students who worked with AI provided themselves with a more robust learning experience when using it than those who used it to bypass the homework process.
Whether you are a parent, teacher, or manager, consider this question for yourself and those you support: Today, where artificial intelligence is helping to produce results, are you still practicing the mindset that improves your skills over time, or are you quietly dependent on the tool that produces them to produce better results?
