As artificial intelligence large model technology deeply permeates various industries, the public's attitude toward AI is undergoing a qualitative shift from "trying out" to "deep scrutiny." The recently released "2025 China Internet Users' AI Cognition Survey Report" points out that China's awareness and penetration rate of AI large models have ranked among the global leaders. However, at the same time, the public's focus on AI risks has shifted significantly, moving from traditional concerns about "privacy" and "unemployment" to deeper anxieties about "their own cognitive abilities."

I. Domestic AI Penetration Rate Significantly Leads, but Faces "Deep Understanding" Challenges

Survey data shows that the penetration rate of AI large models in China reached over 50% in 2025, with only 3.64% of respondents having "never heard of" AI, far below the international average of six countries (22.5%). In work, study, and daily life scenarios, the AI usage rate of Chinese internet users exceeds 40%, leading other countries and major economies.

image.png

However, experts point out that high penetration does not equate to high deep understanding. The current cognitive situation shows a "middle group expansion, deep understanding group shrinkage" feature, reflecting the dilemma of knowledge update lag for ordinary consumers under the rapid iteration of AI technology. Currently, the public tends to view AI as a "partner" rather than a simple tool, and their attitude toward technology has moved away from early enthusiasm and panic, tending toward rationality and objectivity.

II. Trust Divergence: "Hard to Put Down" in Daily Life Scenarios, "Cautious" in Serious Fields

The report reveals a clear trust divergence phenomenon: Chinese internet users place high trust in AI in personal frequent life scenarios, but are cautious in professional service fields.

  • Low Trust Range: Only less than 20% of respondents are willing to trust AI in financial services (18.76%) and the judicial field (17.43%).

  • Deep Analysis of Reasons: Experts believe that AI applications in fields such as healthcare and finance have extremely low "tolerance for errors," and the current level of data governance and sharing is not yet aligned with high-level demands. Additionally, the country's regulation in these high-risk areas is relatively strict, thus affecting public psychological expectations and trust benchmarks.

III. Shift in Risk Concerns: Be Wary of "Cognitive Degradation" and "Emotional Dependence"

Compared to 2024 data, the public's macro-level concerns about AI have undergone a profound change:

  • Old Risks Receding: The proportion of concerns about "personal privacy leakage" and "job positions being replaced" has clearly decreased. This indicates that society's acceptance of the "human-machine collaboration" work model is gradually increasing, and the anxiety in the job market is being alleviated by rational cognition.

  • New Concerns Emerge: The survey shows that more than 45% of respondents list "AI overuse leading to their own cognitive ability degradation" as the top risk. In addition, AI hallucinations (36.37%) and "emotional dependence" generated in interactive scenarios (24.08%) have also become new focuses of public concern.

IV. Governance Path: Legislation and Regulation Become a "Soothing Pill"

Facing the complex challenges brought by AI applications, the public's consensus on risk response is becoming increasingly clear. The survey shows that 54.89% of respondents consider "strengthening legislation and judicial practice in AI regulation" as the most effective way to cope.