A joint study conducted by several American universities found that as little as ten minutes of using an AI assistant can measurably reduce users’ independent problem‑solving abilities. According to a report in Gizmodo, participants who relied on the AI showed a sharp decline in performance once the tool was removed.
Study design
Researchers split participants into two groups: one group was allowed to use an AI tool when solving tasks, while the other group had to work entirely on their own. For those using the AI — specifically the OpenAI GPT‑5 model in this study — the tool was withdrawn for the final three tasks, which they then had to attempt without assistance.
Tasks tested different skills: first participants solved math problems involving fractions, then they completed reading‑comprehension exercises.
Results: short-term gains followed by a steep drop
While using the AI, participants solved a higher share of the math problems. However, when the AI was taken away for the last three questions, their performance fell noticeably: the proportion of solved problems was about 20 percentage points lower than that of participants who had not been allowed to use AI at all for the other tasks.
The AI users also skipped problems at nearly twice the rate of non‑AI users. A similar pattern appeared in reading‑comprehension tasks: AI assistance did not produce a substantial advantage in reading performance, but removing the tool led to a marked decrease in correct answers.
What influenced the decline?
The study found the biggest declines among participants who used the AI to obtain answers verbatim. By contrast, those who used the AI for brainstorming or hints did not show a comparable drop in independent performance.
The entire task session lasted roughly ten minutes, suggesting that users who initially relied heavily on the AI lost critical thinking skills within minutes.
Implications
Researchers conclude that the observed performance decline is more likely a consequence of AI‑assisted problem solving itself rather than the nature of specific tasks, since the results were similar across both math and reading tasks. The practical takeaway is that overreliance on AI tools can quickly erode an individual’s ability to solve problems independently, so indiscriminate delegation of tasks to AI is ill‑advised.
Source: Gizmodo report on a multi‑university study that used the OpenAI GPT‑5 model.
