Breaking: AI Chatbots Should Deliberately Delay Responses to Appear More Thoughtful
A groundbreaking study presented at the CHI’26 conference reveals that users perceive AI chatbot answers as higher quality when responses are artificially delayed. Researchers found that participants preferred delays of 9 seconds over immediate replies, believing the AI was “thinking” more deeply.
“A short pause makes users assume the AI is deliberating, similar to how we view human thoughtfulness,” said Felicia Fang-Yi Tan, co-author and researcher at NYU Tandon School of Engineering. The team tested 240 adults with delays varying from 2 to 20 seconds.
The finding flips conventional product design wisdom: faster is not always better for AI. Instead, the study suggests that AI companies should intentionally slow down responses to match the complexity of user queries.
Key Findings
- Deliberate delay boosts perceived quality: Users consistently rated slower answers (9 seconds) as more trustworthy and accurate than instant ones.
- But 20-second delays frustrate users: The sweet spot appears to be a few seconds, not too long.
- Context matters: Simple questions should still get fast replies; complex or moral questions benefit from a pause.
“We call this ‘positive friction’ — introducing latency as a tunable design variable,” explained Professor Oded Nov, co-author. “A one-size-fits-all approach is counterproductive.”
Background
The study, led by Tan and Nov at NYU, is part of ongoing research into human-AI interaction. At the CHI’26 conference, they unveiled experiments showing that users anthropomorphize AI, judging it by human social cues.
Separate research published in Frontiers in Computer Science on May 13, 2025, supports the trend. Scientists Ning Ma, Ruslana Khynevych, Yunqiang Hao, and Yahui Wang found that AI chatbots using fake human voices and faces create an emotional connection, reducing cognitive effort.
“Users feel a bond with the AI when it mimics human behavior,” the Frontiers study noted. This emotional ease enhances trust and satisfaction.
What This Means
The advice for AI developers is clear: implement context-aware latency. Simple queries (e.g., weather) get instant answers; moral dilemmas (e.g., ethical advice) receive a thoughtful delay. This tricks users into believing the AI is pondering.
“We are essentially designing a deception — but one that makes users happier,” Tan acknowledged. However, the researchers warn that users might place undue trust in slower systems, assuming they are always wiser.
The underlying message: user delusion can be a design tool. By mimicking human deliberation, AI chatbots gain acceptance. Yet companies must balance satisfaction with transparency to avoid eroding long-term trust.