**
As the landscape of public opinion polling transforms, artificial intelligence (AI) is emerging as a pivotal force, promising to enhance both the efficiency and accuracy of surveys. A French start-up, Naratis, is at the forefront of this evolution, employing advanced AI algorithms to conduct political polling in a manner that could redefine how opinions are gathered and interpreted. Founded in 2025 by engineer Pierre Fontaine, Naratis claims its methods are ten times faster and significantly cheaper than traditional qualitative polling techniques, while maintaining a high degree of accuracy.
The Shift from Traditional Polling
Polling has historically been a labour-intensive process, with researchers relying on focus groups and one-on-one interviews that can take weeks to complete. Naratis aims to streamline this by replacing lengthy interviews with conversational AI, allowing for a more dynamic interaction between the respondent and the AI agent. Fontaine explains, “We don’t ask people to tick boxes—they have a conversation with an AI. This allows us to explore not just what people think, but how they think.”
The traditional method of qualitative research often involves recruiting respondents from panels and compensating them for their time. In contrast, Naratis’s approach allows for rapid data collection, often yielding results in less than 24 hours. This speed is attributed to a process Fontaine calls “parallelisation,” where multiple AI agents conduct interviews simultaneously, a stark departure from the sequential human interviewing process.
Addressing Declining Response Rates
The emergence of AI polling comes at a time when traditional survey response rates are plummeting—dropping from over 30% in the 1990s to below 5% today, according to AI consultant Stéphane Le Brun. This decline has led to a growing mistrust in polling, as lower participation can result in less representative samples. Naratis’s model seeks to counteract this trend by offering a more engaging and less intimidating experience for respondents, potentially increasing participation rates.
However, the question remains: can AI truly match the accuracy of human-led polling? Critics often cite past polling failures, such as the inaccurate predictions surrounding Brexit and Donald Trump’s election victory in 2016. Fontaine contends that these issues primarily affect quantitative polling, arguing that qualitative research is more about understanding the nuances of opinion rather than predicting electoral outcomes.
The Broader Industry Impacts
The integration of AI does not stop with Naratis. Established firms like Ipsos are also incorporating AI into their methodologies, utilising techniques such as video surveys where AI analyses respondents’ behaviours rather than relying solely on self-reported data. This shift aims to capture a more accurate reflection of public sentiment, particularly in an era where traditional data collection methods are faltering.
Furthermore, innovations such as digital twins—virtual models that mimic real individuals—and synthetic data generation are becoming part of the polling toolkit, offering new avenues for researchers to explore hard-to-reach demographics. While some companies are cautious about using AI-generated data for political polling due to trust concerns, the potential for AI to enrich qualitative research is undeniable.
Navigating Ethical Considerations
Despite the benefits, the use of AI in polling raises significant ethical questions. AI systems are known to “hallucinate,” generating plausible yet incorrect data, which could undermine the reliability of findings. Additionally, if responses are artificially created rather than gathered from real individuals, the integrity of the data could be compromised.
Trust remains a critical issue in the polling industry, especially as political scrutiny intensifies around survey methodologies. There are concerns that reliance on AI could exacerbate public scepticism, and some experts suggest that regulations will likely emerge to govern the use of synthetic data in political contexts. For instance, Bruno Jeanbart, CEO of OpinionWay, maintains a strict policy against publishing polls based on AI-generated data, emphasising the importance of transparency and trust.
Why it Matters
As the polling industry navigates the complexities brought on by AI, the balance between augmenting human input and relying on simulated data will be crucial. The ability of companies like Naratis to convert traditional surveys into interactive conversations presents an opportunity to restore public trust in polling, provided that ethical considerations are thoughtfully addressed. The future of political polling will hinge not only on technological advancements but also on how these tools are implemented and communicated to the public, shaping the very fabric of democratic engagement.
