AI Ethics & Society · Public Trust in AI

What factors most influence whether people trust an AI system?

Research and public opinion surveys generally point to a consistent set of factors shaping AI trust: perceived accuracy and reliability, transparency about how a system works and its limitations, the stakes involved in a given application, past personal or reported experiences with AI, and a sense of control or recourse if something goes wrong.

Key takeaways

Perceived accuracy and reliability of an AI system strongly shapes how much people are willing to trust it.
Transparency about how a system works, and honesty about its limitations, tends to support greater trust than opaque or overstated claims.
The stakes of a given application matter significantly — people generally demand more trust-building evidence for high-stakes uses like healthcare than for low-stakes ones.
Direct personal experience, whether positive or negative, and widely reported experiences of others both shape individual trust levels.
A sense of control or meaningful recourse if an AI system makes an error is commonly cited as important for sustaining trust.

A Consistent Set of Contributing Factors

Research into public trust in AI, including survey work from organizations like Pew Research Center, points to a fairly consistent set of factors that shape whether and how much people trust a given AI system. Rather than being driven by a single variable, trust in AI appears to emerge from an interaction among several distinct considerations: perceived accuracy and reliability, transparency, the stakes of the application, personal and reported experience, and a sense of control or recourse.

Understanding these factors individually helps explain why trust in AI varies so significantly across different applications, contexts, and individuals, rather than functioning as a single, uniform attitude toward “AI” as a monolithic category.

Accuracy, Transparency, and Stakes

Perceived accuracy and reliability is an intuitive but genuinely significant factor — people are generally more inclined to trust an AI system they believe performs its task correctly and consistently, and conversely, awareness of errors or inconsistent performance tends to undermine trust considerably. Importantly, this is about perceived accuracy, which doesn’t always align perfectly with a system’s actual, measured performance, meaning public communication and framing can meaningfully shape trust independent of a system’s underlying technical quality.

Transparency also plays a significant role, encompassing both transparency about how a system generally works and, perhaps more importantly, honesty about its limitations. Systems and companies that are candid about what an AI system can and cannot reliably do tend to fare better on trust measures than those perceived as overstating capabilities or obscuring known limitations, since discovering an overstated claim tends to damage trust more severely than an initially modest but honest claim.

The stakes involved in a given application matter enormously as well. Research has generally found people are considerably more comfortable with AI in lower-stakes contexts, such as content recommendations or basic customer service interactions, than in higher-stakes contexts like medical diagnosis, hiring decisions, or criminal justice applications, where the consequences of an AI error are more severe and where people tend to expect greater transparency, accuracy, and human oversight before extending meaningful trust.

Experience and a Sense of Control

Personal experience, along with widely reported experiences of others, plays a significant role in shaping individual trust levels. Direct positive experiences with an AI system tend to build trust, while negative experiences — particularly salient or personally consequential ones — often have an outsized effect on reducing trust, a pattern consistent with how trust tends to function for many technologies and institutions more broadly, not just AI specifically.

Finally, a sense of control or meaningful recourse if an AI system makes an error is commonly cited as important for sustaining trust over time. Systems that offer some pathway for a person to challenge, correct, or escalate a problematic AI-driven outcome tend to be viewed more favorably than systems that offer no such recourse, since the latter can leave people feeling powerless in the face of a system’s potential mistakes.

Bottom Line

Public trust in AI systems is generally shaped by a consistent set of interacting factors: perceived accuracy and reliability, transparency about how a system works and its limitations, the stakes involved in a given application, personal and widely reported experience, and a meaningful sense of control or recourse — factors that together explain why trust in AI varies so significantly across different applications and contexts rather than functioning as one uniform attitude.

Go deeper

Frequently asked questions

Does simply making an AI system more accurate automatically increase public trust?

Accuracy is an important factor, but research suggests trust is influenced by multiple factors beyond accuracy alone, including transparency, perceived fairness, and the stakes involved — meaning a highly accurate system that is opaque or used in a way perceived as unfair may still face significant public skepticism.

Do people trust AI more in low-stakes or high-stakes situations?

Research has generally found greater public comfort with AI in lower-stakes applications, such as recommendation systems or basic customer service, compared with higher-stakes applications like medical diagnosis, hiring, or criminal justice decisions, where people tend to want more evidence, transparency, and human oversight before extending trust.

Can a single negative experience significantly reduce someone's trust in AI generally?

This is plausible and consistent with how trust operates for many technologies and institutions more broadly — negative experiences, especially salient or personally significant ones, often have an outsized effect on trust compared with the more diffuse effect of many smaller positive experiences.

Sources

[1]Pew Research Center: Internet & Technology — Pew Research Center
[2]World Economic Forum — World Economic Forum

Written by Editorial Team

Last updated July 25, 2026

Get one well-sourced answer a week

No spam. Unsubscribe anytime.