What generative AI is changing is not just what we produce, but how we behave.

When I began studying how workers were using generative AI on digital labor platforms, I expected to see a familiar story of empowerment. I assumed AI would help people write better proposals, respond more quickly, and compete more effectively for opportunities. In some ways, that was true. ChatGPT made it easier to produce polished language and lowered the effort required to participate in a highly competitive market. But the reality was more complex and, in many ways, more sobering.

What I found was that AI did not simply improve performance. It changed behavior. Workers were not just using ChatGPT to write better bids. They were also shifting into different competitive patterns, becoming faster, more active, and more aggressive in how they approached the market. That may sound like progress, but it raises a deeper question: does becoming more active actually mean becoming more effective? Not always. In my study, some of the most active bidding patterns were associated with better outcomes for less experienced workers, but worse outcomes for more experienced workers. The same behavioral shift that helped one group compete more successfully pushed another toward rejection.

This finding reveals something deeper about the relationship between AI and human judgment. We often think of AI as a productivity tool, something that helps people work faster or communicate better. But AI also changes the conditions of action. When a system can generate polished language in seconds and make a person feel more prepared to compete, it does not just improve execution. It can also reinforce behavioral tendencies that were already there. In that sense, AI does not simply support work. It can accelerate habits.

That is why the issue is not only whether AI makes people more capable, but whether it makes them more calibrated. In competitive settings, activity can look like competence from the outside. A person who responds more quickly, bids more frequently, and sounds more polished may appear better positioned to succeed. But those signals can be misleading. Increased action is not always evidence of better judgment. Sometimes it is simply evidence that the cost of acting has fallen.

This is where AI introduces a new kind of illusion. My earlier work focused on what I called the capability illusion: the appearance of competence without the foundation of real mastery. This study points to a related problem. AI can also create a strategic illusion. It can make people feel as though they are competing better simply because they are moving faster, participating more often, and sounding more persuasive. But acceleration is not the same as calibration. AI can make people more active without making them more accurate.

This is not just a labor market problem. It is also an educational one. If students use AI to generate cleaner prose, quicker analysis, and more polished answers, the immediate result may look positive. They may appear more fluent, more confident, and more productive. But the deeper question is whether they are actually thinking better or simply moving faster through tasks they do not fully understand. That is the risk education now faces: not just dependency on a tool, but dependency on a mode of behavior built around speed, fluency, and unexamined confidence.

Much of the current discussion around AI literacy focuses on prompt engineering, how to ask better questions and get better outputs. While that skill has practical value, it misses the larger point. True AI literacy is not just about using the tool well. It is about developing judgment. Students need to learn not only how to generate answers, but how to question them, verify them, and recognize when a polished response reflects fluency rather than understanding. In that sense, AI literacy should be taught less as a technical skill and more as a reflective discipline.

The goal is not to ban AI from learning. It is to make sure that education still rewards the human capacities that matter most: interpretation, calibration, self-awareness, and the ability to recognize when confidence has moved ahead of understanding.

Generative AI has already changed how we work, learn, and compete. It has lowered barriers to participation and increased the speed of output. But it has also introduced a new challenge. It can make people more active without making them more reflective. The future will not belong simply to those who use AI the most. It will belong to those who can tell the difference between movement and progress. In the age of generative AI, fluency is no longer enough. What we need is wisdom.

References

Barber, B. M., & Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics, 116(1), 261-292.

Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work (Working Paper No. 31161). National Bureau of Economic Research.

Cho, E. (2026). Generative AI, latent bidding regimes, and hiring outcomes in digital labor markets. Working paper.

Dell’Acqua, F., McFowland, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality.

Logg, J. M., Minson, J. A., & Moore, D. A. (2019). Algorithm appreciation: People prefer algorithmic to human judgment. Organizational Behavior and Human Decision Processes, 151, 90-103.

Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192.