Two years ago, merely 45% of adults had genuinely utilized generative AI; today, this figure has surged to 73%. Logically speaking, the more user-friendly a tool becomes, the greater the trust people place in it.

However, the other half of the findings from Prophet's 2026 Consumer Research paints a different picture: overall enthusiasm for the technology has waned by 7% compared to two years prior, and the number of individuals willing to entrust AI with daily decisions has plummeted by 30%.

Increased usage, yet diminished trust—this diverging trend has emerged as an insurmountable obstacle for all AI applications.

01. The Diverging Paths: More Users, Fewer Believers

This phenomenon is not isolated. A survey conducted by Fractl and Search Engine Land in the second quarter of this year revealed that the percentage of consumers who believe AI search outperforms traditional search plummeted from 82% to 54% within a year—a staggering 28 percentage point drop.

In the same study, the number of individuals stating that 'heavy AI use by brands diminishes my trust' doubled from 20% to 40%, reaching 54% among Gen Z.

The most paradoxical fact is that 70% of respondents admitted to using AI search more frequently than a year ago, while only 3% reported using it less. Typically, the curve of technological adoption trends upward, but the curve of trust does not follow suit.

Two years ago, people were thrilled about AI composing poems and creating paintings; now, they meticulously scrutinize its errors.

PwC's 2026 AI Predictions Report assessed this trend, stating that competitive advantage no longer lies with those who adopt AI first, but with those who manage it effectively and utilize it in ways that instill confidence in consumers.

Behind this assessment lies a fundamental business truth—users pay for capabilities, but they remain loyal only if they trust the provider.

02. Trust is Not Uniform: The More Critical the Task, the Less Willingness to Delegate to AI

The disparity in trust becomes even more apparent.

The 'China Netizens' AI Cognition Survey Report,' released in May this year by the Digital Economy and Legal Innovation Research Center at the University of International Business and Economics and the Digital Economy Research Center at Renmin University of China, dissected user attitudes:

68.51% are willing to trust AI for daily life arrangements, and 67.47% for educational support—both exceeding 60%.

However, when it comes to professional advice, the percentage drops to 46.69%, and for handling complex affairs, it plummets to just 29%.

In financial services and the judicial field, the number willing to entrust AI falls below 20%, at only 18.42% and 17.64%, respectively. Overall, 60% of users remain 'skeptical but open' about AI.

In other words, people are willing to delegate low-risk routine tasks but swiftly reclaim control when the stakes are high.

Experian's 2026 Identity & Fraud Report corroborates this caution: 31% of consumers already utilize AI tools for online shopping and transactions, but only 21% are willing to let AI book travel for them, and just 17% for financial decisions.

This caution is not unfounded. Data from the U.S. Federal Trade Commission reveals that consumer-reported fraud losses reached $15.9 billion in 2025, up 27% year-on-year, with AI face-swapping and generative phishing emerging as new tools for criminal activity.

The trust gap is often measured in tangible financial terms.

03. Investment Made, Yet Progress Stalled

The corporate landscape tells an even more intriguing tale. McKinsey's 2026 'State of AI' survey, encompassing over 1,700 practitioners across 97 countries, found that 89% of organizations utilize AI in at least one business function, and 44% have deployed AI company-wide—up 6 percentage points from the previous year. By this metric, AI adoption is undeniable.

However, the other half of the same report reveals that only 37% of respondents attribute operating profit growth to AI, nearly identical to the previous year; merely 6% truly qualify as 'high performers' reaping significant value from AI.

Personal efficiency gains are tangible—80% say AI boosts productivity, and half say it aids in making better decisions—but these benefits largely remain confined to individual workstations and fail to translate into company profits.

McKinsey's '2026 AI Trust Maturity Survey' delved deeper, revealing that the top barrier to scaling intelligent agent AI is neither technical limitations nor regulatory uncertainty, but concerns about safety and risk.

Technology is no longer the bottleneck—trust is. This conclusion would have been unimaginable two years ago when debates centered on whether models were intelligent enough; now, the question is whether we dare to truly relinquish control.

Trust becomes the bottleneck because it is highly asymmetric: it takes countless promises fulfilled to build, but only one failure to collapse.

Gartner's research this year found that in customer service scenarios, after a single negative experience, only 27% of users are willing to try a chatbot again—meaning 70% abandon it after one failure.

49% of respondents are willing to accept chatbots, but only 7% actually used one in their most recent support request.

This is the classic 'leaky bucket' scenario: every new feature loses a batch of users, and the wider the deployment, the harder it is to retain them. More alarmingly, AI is transitioning from a 'tool' to an 'agent.'

Gartner predicts that by 2030, decisions made by AI agents without human oversight could result in $100 billion in asset losses. In Thales' 2026 Digital Trust Index, 77% of respondents express concern about AI assistants acting on their behalf online.

They do not oppose AI itself—they oppose being obstructed by AI without understanding its actions. When AI begins signing documents, placing orders, and making decisions on behalf of humans, trust becomes not just an experience issue, but a liability issue.

04. Trust is Not Elusive—It Has a Blueprint

The good news is that trust is not abstract—it has a clear formula: transparency, controllability, traceability, and an escape route.

Gartner provides quantified answers: by 2026, organizations that robustly implement transparency, trustworthiness, and security are expected to witness a 50% increase in AI model adoption and effectiveness.

The same research found that 87% of customers believe businesses must offer a human escalation option when using generative AI—the highest consensus across all surveys.

Users never sought to reject automation; they simply desire a safety net at critical moments.

Regarding transparency, businesses fall significantly short. Fractl's survey shows that 84% of consumers want AI-generated text clearly labeled, with even higher demands for video (91%), images (90%), and audio (87%); in reality, only 20% of marketers say their companies always disclose AI use, while one-third never do. This disparity itself breeds distrust. Chinese users place greater emphasis on regulations.

In the same 'China Netizens' AI Cognition Survey Report,' 72.11% of respondents deemed AI governance 'urgent' or 'very urgent,' with over half advocating for stronger regulation through legislation and judicial practice.

The 'Measures for Identifying AI-Generated Synthetic Content,' implemented on September 1, 2025, respond to this expectation—requiring machine-generated content to disclose its origin is a prerequisite for rebuilding trust, not a burden.

05. Conclusion

Looking back, the AI industry competed on parameters, computing power, and deployment speed over the past two years—that was the question of 'can it be used?' The next question is 'dare we use it?'

When everyone utilizes AI, the true scarcity is no longer capability, but reassuring certainty.

Whoever builds trust first—by being transparent, returning control to users, and allowing room for error—will secure the next round of AI adoption. In this business, slowing down can actually accelerate progress.

- End -