October 14, 2026
People who lose AI access perform worse than those who never had it: a study of 1,222 participants
Across three experiments with 1,222 participants, AI first raised performance, but once it was switched off, people fell behind those who had worked without it.
When the assistant was taken away, the AI group did worse than people who had solved the tasks on their own all along.
With AI, the same group had first scored higher. The gain lasted only while the assistant was at hand and disappeared along with it.
How it was measured. Three experiments, 1,222 participants: people were randomly split into groups, some got AI and some did not, then the assistant was taken away and the results were compared.
Not measuring the boost. Typical studies measure how much faster a person works with an assistant. This one measured what a person is left with after it is switched off. The result was lower than for those who never had AI.
These experiments did not measure what happens to the skill after months of working this way.