AI Agents Conducting Opinion Polls
"When you hear the word 'politician', what is the first image or emotion that comes to mind?"
The voice posing this question is young, female, brisk, and business-like, belonging to an AI agent — a computer program, essentially a string of code.
A man on the other end of the line responds. While he expresses a rather cynical view of politicians, three other AI agents simultaneously process his answers.
One AI checks whether he is answering the question, another assesses if his responses are too superficial and if prompting is needed to encourage deeper reflection, and a third verifies that the respondent is genuine — not a robot or fraudulent entity.
This polling process is conducted by Naratis, a French AI opinion poll company.
Emergence of AI in Political Polling
"The US has start-ups like Outset, Listen Labs and Hey Marvin that do AI polling like this in the commercial sphere. To my knowledge we're the first to do this for political opinion polling as well,"says Pierre Fontaine, the 28-year-old engineer who founded Naratis in 2025.
What was once the most labor-intensive aspect of opinion research is now becoming one of its most automated.
In France and elsewhere, this transition is beginning to reshape how public opinion is measured, understood, and potentially influenced.
Transforming Qualitative Research with AI
Naratis aims to revolutionize qualitative research — traditionally the slowest and most expensive form of polling — by rebuilding it around AI technology.
Qualitative studies typically involve small groups or one-on-one interviews with paid respondents recruited via panels. These interviews can take weeks to conduct and analyze. Naratis replaces this process with conversational AI.
Unlike quantitative polling, which is largely automated through mass surveys, Naratis emphasizes depth.
"We don't ask people to tick boxes - they have a conversation with an AI,"Fontaine explains.
"That means we can explore not just what people think, but how they think - how they build their opinions, and even when those opinions change."
The company claims its method is
"10 times faster, 10 times cheaper and 90% as accurate as human polling".
A study that once took weeks and tens of thousands of euros can now be completed in a day or two. Responses can often be gathered in under 24 hours, allowing clients to react to events almost in real time.
This speed is achieved through what Fontaine calls "parallelisation": instead of human interviewers working sequentially, AI agents conduct many interviews simultaneously.
Challenges Facing the Polling Industry
The rise of AI polling occurs amid significant challenges for the industry. Response rates to surveys have dropped sharply, from over 30% in the 1990s to below 5% today, according to AI consultant Stéphane Le Brun.
"The decline in response rates has made polling both more expensive and less representative, fueling public distrust,"Le Brun notes.
Assessing AI Polling Accuracy
Regarding Naratis's claim of near-human accuracy, critics may cite past polling failures, such as the inability to predict Brexit or Donald Trump's 2016 victory.
"Such problems mainly affect quantitative polling,"Fontaine argues.
He adds that qualitative research focuses less on predicting outcomes and more on understanding opinions — for example, testing reactions to a campaign slogan rather than forecasting a vote.
AI Integration in Established Polling Firms
Across the industry, established polling firms are also incorporating AI. Ipsos, for example, uses AI extensively in market research.
Instead of asking participants to describe their habits, researchers may request them to film themselves, with AI analyzing the footage. This approach allows companies to observe behavior directly rather than relying solely on self-reported data.
AI is also employed to analyze social media and experiment with "digital twins" and "synthetic people." A digital twin is a virtual model of a real individual designed to respond similarly, while synthetic data involves generating entirely new profiles based on real-world patterns.
These tools help address a persistent polling challenge: studying small or hard-to-reach groups. Researchers sometimes alternate between real respondents and simulated ones, though real people are still used to validate findings.
Caution in Politically Sensitive Polling
However, caution remains strong in politically sensitive polling. Ipsos does not use AI-generated respondents in political surveys, and other firms adopt similar policies.
At OpinionWay, AI may conduct interviews, but
"we would never publish an opinion poll based on AI-generated data,"says CEO Bruno Jeanbart, citing concerns about trust.
Advantages and Risks of AI-Driven Polling
The benefits of AI-driven polling are evident: it is faster, cheaper, more flexible, enables richer data collection, and allows researchers to respond quickly to events.
It may also reduce certain biases, as people can be more candid with a machine than with a human interviewer, especially on sensitive topics. This may explain why in France, opinion polling has consistently underestimated support for the far-right.
Nonetheless, significant risks exist. AI systems can "hallucinate," inventing plausible but incorrect answers. They also tend to produce "common sense" responses shaped by typical perceptions, which runs counter to polling's purpose of capturing actual opinions.
Synthetic data raises deeper questions. If responses are generated rather than collected, what is truly being measured? How should such data be interpreted?
Trust is another major concern. Polling is already subject to political scrutiny and regulation. The introduction of AI, especially in generating data, could intensify these concerns. Jeanbart expects that countries like France may eventually prohibit publishing polls based on synthetic data.
Human Oversight Remains Essential
Even AI advocates acknowledge its limits.
"The goal is end-to-end automation, but today it would be unsafe and socially unacceptable to remove humans entirely,"says Le Brun. Human oversight remains essential for validating results and assuming responsibility.
The Future of AI in Polling
For now, the most probable future is a hybrid one. AI will continue expanding polling's scope, enabling large-scale conversational surveys, integrating social media data, and delivering faster insights.
Techniques like digital twins and synthetic data may find niche applications, particularly in market research.
However, in political polling, the distinction between augmenting human data and simulating it is likely to remain critical. Companies like Naratis are betting that the real transformation lies not in replacing respondents but in changing how they are heard — turning surveys into conversations and conversations into data at unprecedented scale.
Whether this shift restores trust in polling or further erodes it will depend less on the technology itself than on how it is used, explained, and regulated. What is clear is that economic pressures will continue to push the industry toward greater automation.
