Calibrating Trustworthiness in GenAI

Abstract

Generative AI, or “GenAI”—a type of artificial intelligence that can create new content, including text, images, music, and videos, by learning from existing data—is a constantly changing and improving technology gaining widespread use around the world. According to McKinsey’s 2024 Global Survey on AI adoption, 65% of professionals reported their organizations regularly using GenAI, up from 33% the year prior. With GenAI no longer a new tool, and one with user adoption continuing to increase year over year, the Ad Council Research Institute (ACRI), in partnership with Google, set out to understand what the American public knows and feels about GenAI in 2025. Who’s familiar with GenAI, and who uses it? How do they feel about its role in work and at home? How much does the public believe in its usefulness and benefits? What messaging (explanations and in-product statements) are most helpful for users? The survey also includes special focus on how much trust the American public puts in AI-generated content, also known as “results”— the content generated by AI in response to a natural-language prompt, or set of instructions. Do people think about hallucinations and other mistakes GenAI can make? And what in-product statements would help people better calibrate trust in GenAI and its outputs? We surveyed 1,500+ people across the United States spanning ages, genders, races/ethnicities, regions, and household income representative to the U.S. Census to explore how much or little trust Americans have in GenAI results, their perceptions around its mistakes, and the explanations and in-product statements they’d find most helpful and informative when using GenAI tools.