Collective Intelligence Project（CIP）於5月21日至6月1日針對1,000名多元成人進行的一項新調查顯示，全球56%的受訪者信任AI聊天機器人會為其最佳利益行事——超越了對民選代表和企業的信任。約50%的受訪者偏好以AI輔助決策支持來處理涉及地方服務（如稅務、許可證或執照）的社區選擇。紐約大學心智、倫理與政策中心主任Jeff Sebo將此趨勢歸因於聊天機器人的持續可用性、對個別使用者的專注關注以及其被感知的中立性，而政治人物則常顯得疏遠、缺乏回應且自私自利。

根據Pew Research Center的研究，美國民眾對政府的信任數十年來持續偏低，僅有17%的美國人信任政府「會做正確的事」，相較之下，OECD國家約有40%的受訪者表示高度信任政府。CIP的研究顯示，公眾對AI的信任是經過校準的——在諮詢和資訊功能方面信任度較高，但在涉及自主決策時則較低。僅有3.7%的受訪者願意讓AI代替其做出實際決定。當被問及誰會更公平地權衡其利益時，近40%選擇了AI，而約31%選擇了目前的決策者。紐約大學Courant研究所的Hasan Aljabbouli指出，政府信任度下降、AI的持續可用性及其被感知的公正性是關鍵驅動因素。

該研究還發現，超過80%的受訪者在日常生活中注意到AI的存在，約60%的人每天個人使用AI系統。然而，專家警告AI可能放大錯誤資訊，且私密、個人化的聊天機器人對話中的偏見可能比社群媒體演算法中的偏見更難被偵測，尤其是在選舉前夕。哥倫比亞大學機械工程教授Hod Lipson指出，選舉決策很複雜，因為選民通常已有既定立場。Aljabbouli強調AI供應商在確保準確性、平衡性、透明度和問責制方面負有日益增加的責任。Lipson預測這些信任數字將持續增長。

A new survey by the Collective Intelligence Project (CIP), conducted May 21–June 1 with 1,000 diverse adults, reveals that 56% of global respondents trust AI chatbots to act in their best interest—surpassing trust in elected representatives and corporations. Approximately 50% prefer AI-assisted decision support for community choices involving local services such as taxes, permits, or licenses. Jeff Sebo, director of the Center for Mind, Ethics, and Policy at New York University, attributed this trend to chatbots' constant availability, focused attention on individual users, and perceived neutrality, while politicians often appear distant, unresponsive, and self-interested.

Trust in government has remained low for decades in the US, with only 17% of Americans trusting the government "to do what is right," per the Pew Research Center, compared to approximately 40% across OECD nations. The CIP study shows public trust in AI is calibrated—high for advisory and informational functions but lower where autonomy is involved. Only 3.7% of respondents would let AI make actual decisions on their behalf. When asked who would weigh their interests more fairly, almost 40% chose AI versus about 31% who chose current decision-makers. Hasan Aljabbouli of NYU's Courant Institute cited declining governmental trust, AI's continuous availability, and perceived impartiality as key drivers.

The study also found over 80% of respondents notice AI in daily life, with approximately 60% using an AI system daily. However, experts caution that AI could amplify misinformation, and bias in private, personalized chatbot conversations may be harder to detect than in social media algorithms, particularly before elections. Hod Lipson, professor of mechanical engineering at Columbia University, noted that election decisions are complicated because voters typically have their minds already made up. Aljabbouli stressed the increasing responsibility on AI providers to ensure accuracy, balance, transparency, and accountability. Lipson predicted these trust numbers will continue to grow.