The case for AI to enhance democracy by encouraging more inclusive civic participation or producing more tangible feedback loops between government and its people has been proposed and tested numerous times. The problem that people building AI for democracy usually aim to solve is that democratic institutions lack the capacity to deliberate adequately at scale. AI, with its ability to synthesize, contextualize, and translate, seems like an obvious solution. But the problem isn’t just about expanding institutional capacity; our most recent Global Dialogue reveals there’s more to the picture.
Findings from the Global Dialogue reveal that people want AI that improves their own ability to participate in civic life, and are generally unsure about AI that makes decisions for them. 48.7% of people say one of the main things stopping them from having a say in community decisions is that their view will not change anything, and 45.8% feel their opinion matters “a little” or “not at all” in the venue that matters most to them (n=1,040).
We asked the global public about where and how they want AI to play a role in democratic processes. Our data draws a consistent line between two ways that AI could enter democratic life. In the first, AI takes over parts of the decision itself: it decides on people’s behalf or represents them in deliberation. In the second, AI improves people’s ability to exercise their own judgment: it explains, translates, summarizes, and shows them how decisions affect them, while they make the final decision.
We found the public does not prefer AI that enables democratic surrender, but wants AI that reinforces democratic empowerment. When we explore how AI can augment and empower citizens in democratic processes, we need to view the problem not just through a capacity lens, but through one of power and agency. If people don’t feel like what they say or do in the democratic sphere has any weight, no amount of technology will resolve the foundational power asymmetries.
What is considered surrender and empowerment is not an absolute distinction, and we suspect that attitudes towards AI assistance and delegation may change over time. One interesting finding from our Dialogue was that respondents did feel comfortable, to some extent, having AI generate their arguments for them (30.6%), represent their arguments (26.6%), and even have AI attend meetings in their place (9.5%). We think it’s worth keeping in mind that democratic surrender could become the default option, and that decisions around how to deploy AI in the civic sphere should keep the distinction between surrender and empowerment at the forefront.
AI as Democratic Surrender
In this present moment, we see that the rejections of full delegation are clear:
People do not want AI or humans to make community decisions on their behalf. Asked how they would most like community decisions handled, 3.7% chose an AI deciding on their behalf and 8.4% chose a human doing so.
People want to speak for themselves. 9.5% would be comfortable having an AI attend a meeting in their place, and 26.6% would let it represent their arguments (n=1,007). 13.5% would make AI-identified common ground the basis for new policy or rules.
The higher the stakes, the less people want AI involved. Even granting the premise that AI decides more accurately than humans, 70.6% say it should never set a criminal sentence and 65.8% say it should never decide the rules everyone lives under.
One of the most broadly endorsed explanations, agreed with by at least 45% of every demographic group we measured, states the principle directly: “Because we humans are the ones affected, we are the ones who make the decision, not the AI, which is not affected at all.”
AI as Democratic Empowerment
The endorsement of AI assistance over human assistance is equally consistent, however, and comes from a population that currently feels its participation is pointless:
People would be more likely to participate with the right information, and AI can help. 77.2% say they would be more likely to take part in a community decision if an AI sent them a personalized report on how it would affect their household (n=1,033). The reason, among those who said they would be more likely to participate and agreed with by at least 50% of every population group: “I need information to ensure my decision is reasonable before I participate.”
People prefer AI assistance in community affairs. When asked how community decisions should be handled, 49.7% chose an AI that helps them understand the issues while they decide. 33.0% would prefer an AI applying existing rules to reach an answer quickly over a longer in-person meeting (n=1012), and 38.0% say AI should be used to find common ground “whenever it might be useful“ (n=1018).
People prefer AI assistance across all domains in which community decisions are made: In fact, we found that across civic domains, pluralities even prefer AI assistance over human assistance. When asked about what kinds of assistance in decision-making people want in various civic domains, AI was selected more often than humans: local services (46.7% vs. 35.8%), spending local taxes or resources (47.8% vs. 28.0%), everyday administrative decisions (45.4% vs. 27.2%), and online content rules (40.9% vs. 22.2%).
Reconsidering AI in Democracy
We found a core distinction between democratic surrender, where AI fully replaces human decision-making, and democratic empowerment, where AI acts as a tool to enhance personal agency. People consistently reject institutions and technologies that reinforce surrender, especially when significant human stakes are involved. They show a strong preference for AI assistance over human assistance that helps them navigate issues and make their own informed choices.
Whether we head towards democratic surrender or empowerment is determined at multiple levels, from the design and deployment decisions made by builders, to the specific institutional arrangements that scaffold the use of AI within decision-making systems. In a world in which 36.6% of people agree that AI could make better decisions on their behalf than elected representatives, we can expect more AI systems built to help people understand and participate in civic affairs. These should be built to expand the political intelligence and agency of humans, as well as the relationships that make democracy possible, so that societies can navigate the AI transition.
This post draws on findings from Global Dialogues Round 9. 1,103 participants across 69 countries.
This survey serves as a structured global dialogue — not a nationally representative opinion poll. The distinction matters. The survey is one of the first to combine deliberative, interactive methods with global scale–gathering structured public input across a wide range of countries and contexts using a format that allows participants not only to contribute their own views but also to engage with and vote on the ideas of others. It asked people across a wide range of countries and contexts to reason through a set of genuinely complex questions about AI and democracy. It did not aim for statistical precision at a country level, but rather to allow us to form an overview of regional patterns and opinions that can help inform future policy conversations.
Participation is limited to those with reliable internet access, recruited through Prolific, and skews young (67% under 35), urban, and digitally engaged. Representativeness is tracked via the Global Representativeness Index and published for each round.
We are grateful for our team of AI & democracy practitioners who participated in an expert assembly prior to the dialogue. They helped inform the questions and overall narrative.