16% of voters used AI to make a choice in the voting booth: enough to “constitute a weak signal of a profound transformation in the making”: this is the conclusion of the Terra Nova think tank which published this Monday, May 18, a study on the use of generative artificial intelligence tools during the last municipal elections.
Will ChatGPT or Gemini change the casting of the next presidential election? The Terra Nova think tank published this Monday, May 18, a study showing how generative AI tools such as Le Chat or Anthropic were used by the French during the last municipal elections. If we already knew that these generative artificial intelligence (AI) tools are used, in particular by the youngest, to obtain information or information, the think tank shows that some voters have gone further than simple search for information.
This think tank, which presents itself as “ independent progressive aiming to produce and disseminate innovative political solutions in France and Europe », analyzed data from Toluna Harris Interactive produced for M6 and RTL. On March 15, 2026, the day of the 1st round of the municipal elections, “ 4,145 people registered on the electoral lists in municipalities with 3,500 inhabitants or more were interviewed online, out of a sample of 6,288 people, representative of the French population aged 18 and over ».
One in five young people use AI chatbots to find out about a program or a candidate
Among the figures to remember, Terra Nova notes that 11% of French people surveyed used AI to find out about the candidates or their programs as part of the last Municipal campaign. ChatGPT, Anthropic and Le Chat, however, come in last position in the list of information sources used for this election, far behind leaflets, word of mouth, the regional daily press, and social networks.
If 11% is an average, the detail of this proportion shows that it is mainly young people who use these tools. One in five young people (22% of 18-24 year olds and 19% of 25-34 year olds) use AI chatbots to find information on the program and candidates in their municipality, it is detailed.
Using generative AI tools to gain information is not new. But the study shows that some of those surveyed used these systems to choose which candidate to vote for. According to Terra Nova, a threshold has been crossed. There is now “ an emerging use of AI not only to obtain information but also to refine one’s choice », Underlines Jean-Daniel Lévy, deputy director of Toluna France, who wrote the study.
Also read: After the data, our intentions? How AI could manipulate our desires
“AI is making a remarkable entry into the formation of political choices”
Thus, 16% of respondents said they had used ChatGPT or its competitors to help with voting decisions:
- 7% use it to confirm their choice (comfort),
- 5% believe that AI has influenced them (they changed their mind) and
- 4% indicate that these tools helped them make a decision and make their choice.
According to the think tank, “ AI is making a remarkable entry into the formation of political choices. A significant fraction of those who have used it even say they changed their mind after consulting AI “. This phenomenon, however, remains more urban than rural ”, with a use “ significantly more common in large cities, particularly in Île-de-France and in municipalities with more than 100,000 inhabitants ».
Also read: AI will ruin the political debate, and we have no tools to protect ourselves from it
“A weak signal of a profound transformation in the making”
Likewise, this use of AI remains “ very marginal in rural communities and small towns “. Likewise, it affects more “ fringes of populations which are structurally the least informed and the most distant from politics “, like the ” people with low incomes (28% of people with the highest low average compared to 16% on average) ».
This entry of generative AI into the decision-making sphere in electoral matters “ already constitutes a weak signal of a profound transformation in the making », Write the authors of the study. In the next presidential elections, these tools could be able to “ to influence the ultimate choice of several million French people », They add. The question will then not be who uses AI, “ but who designs it, how the information is prioritized and with what safeguards »: so many elements on which AI companies, mostly American, show little transparency.
Also read: Forced to open the hood on their training data, OpenAI, Google and Mistral are dragging their feet
ChatGPT will come under reinforced control from Brussels
👉🏻 Follow tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.
