People Overestimate Others' AI Use by Nearly 80%, Study Finds

When you receive an email that feels unusually polished or uses repetitive sentence structures, you might wonder whether the sender used AI to write it. According to new research, you're probably overestimating how much artificial intelligence your colleagues actually use.

Cognitive scientist Zoe Purcell conducted two studies examining more than 1,000 adults to understand perceptions around AI-assisted communication. Participants viewed realistic scenarios in familiar applications like Gmail that demonstrated varying levels of AI involvement, from none to substantial enhancement across text, audio, and video formats.

The findings reveal a striking perception gap. When asked to rate on a 100-point scale how likely others were to use AI in similar interactions, participants averaged around 71. When rating their own likelihood of using AI, that number dropped to approximately 40—a difference of nearly 80 percent.

Purcell's team also discovered that disclosed AI use was viewed more favorably than hidden use, according to details first reported by AI Watch.

Why it matters

This perception gap represents more than curiosity about technology adoption. It signals a default stance of distrust that can slow collaboration, increase verification overhead, and damage workplace relationships. As AI writing tools become ubiquitous, leaders must address the trust deficit before it becomes embedded in organizational culture.

The trust deficit carries real costs

The tendency to assume others use AI more than they do reflects a broader pattern psychologists call asymmetric feedback. Research by Detlef Fetchenhauer and David Dunning demonstrated this phenomenon through economic games involving 56 strangers. Participants predicted 52 percent would prove trustworthy, but the actual rate was 80 percent.

The mechanism is straightforward: when you trust someone and they disappoint you, you receive immediate confirmation that trust was misplaced. When you withhold trust, you never create an opportunity to learn whether trust would have been warranted. Over time, this feedback loop trains people toward suspicion.

Stephen M. R. Covey frames trust as an economic factor rather than merely a soft skill. High-trust environments enable faster decisions and lower transaction costs. Low-trust environments require additional verification, fact-checking, and second-guessing—all of which drain productivity.

Rebuilding trust in AI-era communication

Experts recommend three specific interventions when suspicion arises:

First, recognize the moment suspicion forms. Noticing the thought before acting on it creates space for more deliberate responses.

Second, apply your own standards to others. Consider your own reasons for using or avoiding AI assistance, then extend that same reasoning to colleagues.

Third, construct a generous interpretation. Develop at least one plausible, charitable explanation for the communication you received and treat it as a legitimate possibility.

These practices extend beyond detecting AI use to addressing the underlying tendency toward distrust in professional relationships. As artificial intelligence becomes more integrated into workplace communication, the ability to maintain trust while adapting to new tools will increasingly separate high-performing teams from struggling ones.

The research was conducted by cognitive scientist Zoe Purcell and first reported by AI Watch.