Naratis AI Polling Offers 10x Speed, 90% Accuracy, Sparks Trust Debate

TL;DR: Naratis, a French AI polling firm, says its conversational agents conduct political surveys ten times faster and cheaper than traditional methods while achieving 90% of their accuracy. The claim comes as overall survey response rates have dropped from over 30% in the 1990s to under 5% today.

Naratis deploys conversational AI that interviews respondents in real time, probing answers for depth and checking consistency. This approach lets researchers explore not just what people think but how they form opinions.

The shift to AI follows a steep decline in survey participation, with response rates falling from above 30% in the 1990s to below 5% currently, making conventional polling more costly and less representative.

Unlike mass‑survey quantitative tools, Naratis focuses on qualitative research, aiming to understand opinion formation rather than predict election outcomes.

The company states its AI‑driven method is ten times faster, ten times cheaper, and 90% as accurate as human‑conducted polling.

Naratis says its agents can follow up on superficial answers, verify that respondents are not bots, and encourage deeper reflection during the conversation.

Industry data show that the overall response rate for surveys has fallen from over 30% in the 1990s to less than 5% today.

Proponents argue the speed and lower expense enable near‑real‑time insight for campaigns and may reduce the tendency to give socially desirable answers to a human interviewer.

Skeptics caution that AI systems can hallucinate, fabricating plausible but incorrect responses, and that reliance on synthetic data could undermine confidence in political polls.

Established firms such as Ipsos and OpinionWay say they use AI for analysis but refuse to publish political polls based solely on AI‑generated data, insisting on human validation.

Watch for upcoming side‑by‑side studies that compare Naratis outputs with traditional polls and for any regulatory frameworks governing AI use in political research.