In the fast-moving world of artificial intelligence, speed is often seen as a clear advantage.
Companies compete to make AI systems respond faster and faster, reducing the delay between a question and an answer.
But new research suggests that faster is not always better—and that a short pause might actually make AI seem more thoughtful.
A study presented at the ACM CHI 2026 Conference explored how response time affects the way people experience AI.
The research was led by Felicia Fang-Yi Tan and Oded Nov, who asked 240 participants to complete everyday tasks using a chatbot.
Some tasks involved creating content, such as brainstorming ideas or writing text. Others focused on advice, like evaluating decisions or giving recommendations.
The researchers designed the system to respond at different speeds. Some participants received answers almost instantly, within two seconds.
Others had to wait longer, up to nine or even twenty seconds. Importantly, the quality of the answers did not change—only the timing did.
The results were surprising. The speed of the response did not significantly change how people used the AI.
Participants asked a similar number of questions and interacted with the system in similar ways, regardless of whether the answers were fast or slow. Instead, their behavior depended more on the type of task.
Creative tasks led to more back-and-forth interaction, while advice tasks involved fewer, more focused exchanges.
However, response time had a strong effect on how people judged the AI. Those who received very fast answers often rated them as less thoughtful and less useful. In contrast, people who waited longer tended to see the same kinds of responses as more careful and valuable.
This effect is linked to how humans interpret pauses in conversation. In everyday life, a quick reply can seem rushed, while a delay suggests someone is thinking carefully. People appear to apply these same social expectations to AI systems, even though they know they are interacting with software. A short pause can make the AI feel more intelligent, even if nothing about the answer itself has changed.
The findings suggest that delays, often seen as a flaw, could actually be used in a positive way. Designers sometimes call this “positive friction,” meaning small slowdowns that encourage reflection. Instead of trying to eliminate every moment of waiting, developers could think about how timing shapes the user’s experience.
At the same time, this raises important ethical questions. If people trust slower responses more, they might assume the answers are better—even when they are not. This could lead to misplaced confidence in AI systems. It also raises the question of whether companies should be transparent about how response timing is designed.
Overall, the study challenges a simple assumption: faster is always better. In the case of AI, a little delay might not be a problem—it might actually make the experience feel smarter, more human, and more trustworthy.
