As the landscape of public opinion polling evolves, the emergence of artificial intelligence (AI) technologies is poised to reshape how we gather and interpret data. A pioneering French company, Naratis, is leading this shift, claiming to offer a faster, cheaper, and nearly as accurate alternative to traditional polling methods. Founded by engineer Pierre Fontaine in 2025, Naratis aims to transform qualitative research—historically a labour-intensive process—into an efficient, AI-driven operation that promises deeper insights into voter sentiment and behaviour.
The Rise of AI in Polling
Polling has long been viewed as a fundamental mechanism for gauging public sentiment, yet it faces significant challenges in the digital age. Traditional methods often rely on time-consuming interviews with small focus groups, which can take weeks to complete. Naratis aims to streamline this process through conversational AI, allowing for real-time interactions and capturing nuanced opinions.
Fontaine explains, “We don’t ask people to tick boxes—they engage in a conversation with an AI. This means we can explore not just what people think, but how they think, how they form their opinions, and even when those opinions evolve.” By leveraging AI agents that operate simultaneously, Naratis can conduct interviews at an unprecedented scale. This innovation allows the company to gather responses in under 24 hours, enabling clients to react swiftly to political events as they unfold.
A Response to Declining Engagement
Polling companies have struggled with plummeting response rates, which have dropped from over 30% in the 1990s to below 5% today, according to AI consultant Stéphane Le Brun. This decline raises concerns about the representativeness and reliability of polling data. In this context, Naratis’s AI approach emerges as a potential solution, albeit not without its own set of challenges.
Critics of AI polling point to historical inaccuracies, such as the failure to predict Brexit and Donald Trump’s victory in 2016. Fontaine argues that qualitative research is less about forecasting results and more about gaining insights into public opinion. While traditional quantitative polling has faced scrutiny, the qualitative methods employed by Naratis aim to provide a deeper understanding of voter attitudes and motivations.
The Broader Industry Landscape
The adoption of AI is not restricted to upstart firms like Naratis. Established polling organisations are also integrating AI technologies into their operations. For instance, Ipsos has begun utilising AI in market research, prompting respondents to film themselves and analysing the resulting footage. This shift allows researchers to observe behaviours directly rather than relying solely on self-reported data.
AI’s capabilities extend beyond mere data collection; they also facilitate the analysis of social media trends and the development of digital models. These digital twins and synthetic profiles can enhance the understanding of hard-to-reach demographics. However, caution remains paramount in politically sensitive polling. Firms like OpinionWay maintain a strict policy against using AI-generated data for political surveys due to trust and validity concerns.
Navigating the Challenges
While the potential benefits of AI-driven polling are clear—speed, cost-effectiveness, and richer data collection—significant risks accompany these innovations. AI systems can produce inaccurate or misleading information, a phenomenon known as “hallucination.” As a result, some industry experts argue for the necessity of human oversight in the polling process to maintain credibility and accuracy.
Trust is a critical issue in polling, and the introduction of AI-generated data could exacerbate existing scepticism. Concerns around the validity of synthetic data may lead to regulatory scrutiny, with some predicting that countries like France may eventually prohibit polls based solely on AI-generated information.
Despite these challenges, the future of polling is likely to be hybrid. AI will continue to enhance the scope of opinion research, enabling broader conversational surveys and integrating diverse data sources. However, the balance between human insight and AI simulation will be essential in maintaining the integrity of political polling.
Why it Matters
The evolution of AI in polling signifies a pivotal moment for the industry, with the potential to either restore trust in public opinion research or deepen existing divides. As economic pressures drive the push towards automation, the manner in which AI is utilised, explained, and regulated will ultimately determine its impact on the future of political discourse. The challenge lies not merely in the technology itself, but in ensuring that it serves to amplify genuine voter voices rather than distort them, thereby shaping a more informed electorate in an increasingly complex political landscape.
