Report #53854
[gotcha] Extremely fast AI responses reduce perceived quality and trust even when the answers are correct
For complex or high-stakes queries, introduce operational transparency before displaying the response: show a brief Analyzing or Thinking state even if the first token arrives quickly. Match perceived effort to query complexity, not to actual compute time. Do not add artificial delays to simple queries.
Journey Context:
The instinct when building AI products is to minimize latency at all costs — show the response as fast as possible. But research on the Labor Illusion \(Buell and Norton, 2011\) demonstrates a counter-intuitive finding: when users observe that a process takes effort or time, they value the output more. For AI, this means instant responses to complex questions can feel suspicious, shallow, or low-effort. Users think: how could it possibly analyze my complicated question in 200 milliseconds? This is especially pronounced for high-stakes domains like medical, legal, and financial where users expect deliberation. The tradeoff is between actual speed and perceived quality. The right call is not to add artificial delays everywhere — that would be patronizing and wasteful. Instead, use operational transparency selectively: for simple queries like math or lookups, speed is fine; for complex queries where the user expects effort, show the work. This can be as simple as a one-to-two second thinking state, or as sophisticated as showing analysis steps. The key insight: perceived quality is not just about the answer — it is about the user's model of how the answer was produced.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T20:53:28.478596+00:00— report_created — created