A recent study presented at the ACM CHI 2026 conference challenges the prevailing assumption in artificial intelligence that faster response times are always superior. While latency is typically viewed as a technical hurdle to be minimized, researchers from the CHI 2026 conference in Barcelona found that delaying AI responses can actually enhance user perception of intelligence and thoughtfulness. The findings suggest that the timing of an AI reply significantly shapes how users evaluate the system's output, often independent of the actual quality of the content provided. Conducted by Felicia Fang-Yi Tan and Professor Oded Nov, the study recruited 240 participants to complete various tasks using a chatbot. The research design required the system to respond at different speeds, offering answers after two seconds, nine seconds, or up to twenty seconds. Participants were divided into two groups based on task type: creation tasks, which involved brainstorming and drafting, and advice tasks, which required evaluation and recommendations. The study also utilized probabilistic AI models, which differ fundamentally from traditional deterministic software where outcomes are fixed and predictable. Because conversational interfaces mimic human interaction, users naturally interpret pauses as social cues, such as deep thought or deliberation. The results revealed that response speed did not significantly alter user behavior. Participants prompted the system at similar rates regardless of whether they waited two seconds or twenty. Instead, interaction patterns were driven by the nature of the task. Users engaged in more back-and-forth refinement during creation tasks, while advice tasks resulted in more focused exchanges. However, timing had a profound impact on perception. Participants who received immediate two-second responses consistently rated the answers as less thoughtful and less useful. Conversely, those who experienced longer delays viewed the same responses more favorably, attributing the pause to the system taking care to consider the question. This phenomenon highlights a subtle aspect of human psychology where people apply social expectations to machines. In human conversation, a quick reply can seem impulsive, whereas a measured delay suggests reflection. The study indicates that users project these same cues onto AI, interpreting a delay as a sign of higher quality deliberation. The researchers propose that rather than treating every millisecond of waiting as wasted time, designers should consider the concept of positive friction. This involves intentionally managing latency to promote cognitive benefits, such as encouraging users to reflect on the input or output before reacting. However, the study also raises critical ethical considerations. If users equate longer response times with superior quality, they may place undue trust in slower systems, even if the underlying logic or accuracy has not changed. This dynamic suggests that AI systems could be designed to manipulate user perception through timing, creating a risk of misleading users about the system's actual capabilities. The researchers conclude that as AI becomes more integrated into daily workflows, it is essential to determine whether timing should be designed to shape perception and, if so, whether users should be informed when delays are intentional. The findings suggest that the pursuit of speed alone may overlook the complex ways humans perceive and interact with intelligent systems.