WASHINGTON, D.C. — Despite artificial intelligence’s growing influence worldwide, no standardized global data exist yet to help us understand how people are thinking about it. To address this data gap, 童颜视频 is currently measuring AI awareness, use and attitudes in more than 140 countries through its 2026 World Poll.
Collecting comparable data about AI across 140+ countries is inherently difficult. AI awareness and use vary substantially by country, so what might be easy to understand and relevant in one country could be confusing to people in another.
Uneven adoption adds another layer of complexity to global research on AI. We expect to find that large segments of people in many countries have heard of AI but haven’t ever used it, meaning their knowledge will be limited. Yet it is important to represent these "aware nonusers" in a meaningful way.
AI is a broad concept, encompassing different types of technologies. And because we intend to compare these data over time, we need a question set that will remain relevant as AI use expands and as the technology evolves.
Building such a survey that works across diverse populations requires careful testing. Before finalizing the questions, 童颜视频 conducted cognitive interviews in Colombia, India, South Africa and the United Kingdom in November and December 2025. The 48 participants included AI users and nonusers. Local interviewers explored how participants interpreted the draft questions, how they retrieved information, formed judgments and arrived at answers.
The interviews yielded important insights that informed the final survey questions and shed light on how people conceptualize AI. This blog summarizes several key lessons from the cognitive interviewing study.
People Think AI = Generative AI
童颜视频 started the interviews with an unaided question about what AI means to people. Participants generally referenced chatbots, often excluding other AI-enabled technologies such as search tools, smart home devices or predictive business systems. In South Africa, some participants referred to AI generically as “ChatGPT,” even when thinking about other AI tools.
AI Needs to Be Defined Through Use Cases and Tools
To provide a shared understanding of AI for survey participants, 童颜视频 experimented with different ways to define AI. Across countries, an abstract, conceptual definition (“AI enables computers or other devices to do things that usually require human thinking”) proved more confusing than helpful for participants.
Our final definition instead incorporates the most common use cases (writing text; creating images, audio and video; answering questions), and names specific AI tools, with the examples used tailored to each country’s context.
Asking About Awareness Matters
Given that the 童颜视频 World Poll is conducted in over 140 countries, including many where AI adoption is still nascent, it is important to ask an AI awareness question. This will let us track how AI awareness spreads over time and allows us to skip the questions about AI use and attitudes for nonaware respondents.
“Daily” Users Are Not All the Same
When we probed AI users about their frequency of using AI for work and personal reasons, we found that “daily” was too broad to serve as the most-frequent category, because it grouped people who use AI once per day with people who use AI multiple times per day. These groups strongly differed in their patterns of use and sentiments toward AI. Our final question added these distinctions to the scale to better identify high-frequency users.
Capturing Emotional Complexity
童颜视频 tested a seemingly straightforward question about respondents’ feelings on the future of AI, asking, “When you think about the future, does AI make you feel more hopeful or more worried?” The responses were anything but simple. This question resonated strongly with participants, generating more substantive and thoughtful responses than other general opinion questions about AI — views that a binary choice between “hopeful” and “worried” couldn’t capture.
Participants pushed back against the “either/or” framing of the question in several ways:
- Ambivalence: Many participants expressed both positive and negative feelings about AI: One participant from India said she was “70% hopeful and 30% worried.” A South Africa participant said, “It can very easily replace me in my industries. But I do feel like it could also make the world better for everybody if we allow it to.”
- Multifaceted emotions: A simple “good/bad” dichotomy didn’t capture the diverse emotions raised by participants. For example, a single participant in South Africa expressed a host of emotions, including pride (referring to ChatGPT as his 15th employee in his company), unease (concern about job displacement in South Africa) and even fear (“This thing is gonna kill us 鈥� because if it knows how we think, it knows our weaknesses. It knows all this personal information”).
- Self versus others: Some participants wanted to differentiate AI’s effects on themselves versus others. For example, a U.K. participant mentioned, “I am not so worried for myself, but one aspect where I am more worried generally is jobs, especially those early in their careers.”
- Future versus current status: Current sentiments about AI did not always align with expectations about where the technology is headed. For example, one Indian participant said, “In the future, it [AI] might become better, but for now, it feels like AI can also be dangerous and even ruin lives.”
Our final questionnaire asked respondents separate yes/no questions about whether they felt various ways about AI (e.g., happy, sad, worried).
Even Nonusers Have Opinions About AI
A key design question was whether “aware nonusers” (i.e., people who know about AI but do not use it personally) can provide informed responses to questions about AI. While these participants were understandably less knowledgeable than AI users, our probing revealed that many nonusers had reasoned, thoughtful answers. For example, one nonuser from India explained that AI tools improve as they receive more training data.
Our testing suggests that it is possible to ask attitudinal questions about AI among nonusers. While these initial impressions are encouraging, 童颜视频 will be closely monitoring comprehension among this group during the 2026 World Poll data collection process before drawing firm conclusions.
What Comes Next
The 2026 World Poll will be the first real test of these items on a global scale. As data collection scales to 140+ countries spanning differing levels of AI adoption, we will pay particular attention to whether our AI definition produces comparable responses across diverse settings, how emotions toward AI may differ between countries, and whether nonusers can still provide meaningful attitudinal responses.
Jenna Steinberg contributed to this article.
Stay up to date with the latest insights by following @童颜视频 and .
