France AI Adoption Statistics 2026: People, Work and Business

France AI adoption statistics 2026: the numbers at a glance
France AI adoption statistics 2026 show two distinct shifts. Arcep’s digital barometer found that 48% of people aged 12 and over used generative AI in 2025. France Num’s 2026 survey found that 40% of very small businesses and SMEs reported using at least one AI solution. Those figures describe different populations, different tools and different reference periods. They answer two useful questions, but combining them would produce a number that neither survey measured.
For a third perspective, the French media regulator Arcom found that AI services reached 57.1% of the population in its internet audience measurement for March 2026. That is a measure of visitors in a month. If you are assessing whether people rely on AI at work, whether a business has adopted it, or whether a particular platform leads the market, each question needs a different measure.
This guide explains what the French figures cover, where comparisons can go wrong and how to cite them. For context, see our AI adoption statistics by country and the What Is ChatGPT? homepage.
How many people in France use generative AI?
Arcep’s 2026 edition of the Baromètre du numérique reports 48% use among the French population aged 12 and over in the 2025 survey. The corresponding figure was 20% in 2023. That is an increase of 28 percentage points across the two survey years. The 2026 label names the edition of the report. For a chart or presentation, label the data point 2025 and put the publication year in the source note.
The report also gives an 85% figure for 18- to 24-year-olds. This is the share within that age group. It does not mean that 85% of French generative AI users are 18 to 24, or that the same share of young people use it at school or work. It shows that the national average masks a wide difference between age groups.

Self-reporting has a practical limitation. People may identify ChatGPT or a dedicated image generator as AI, while thinking of an automatic summary inside another product as an ordinary feature. A survey response also tells you little about frequency. A one-off experiment and an everyday writing habit are both forms of use, but they would have very different implications for an employer deciding how much training to offer.
If you need one headline statistic for general public use, the 48% figure is the clearest choice among these sources. Keep the age threshold, year and term “generative AI” together. Leaving out any of the three makes it easier for readers to misinterpret the figure as a current count of adult ChatGPT users.
What Arcom’s 57.1% internet audience figure means
Arcom’s first AI services audience barometer reports 33 million unique visitors in March 2026, equivalent to 57.1% on its measured population basis. An audience panel answers a reach question. It estimates how many distinct people visited services covered by the measurement over a specified period. It does not establish how often those visitors used a tool, whether they had an account, or whether they completed a task on it.
That distinction is why the Arcom percentage cannot be subtracted from Arcep’s figure to calculate growth. The reference month, measurement method, definition and population all differ.
Someone might visit an AI service without generating anything, while another person might use a feature inside a product outside the audience list. Both are plausible, and neither is resolved by putting the headline percentages side by side.
Audience reach is helpful if your question is whether dedicated AI services attract a large public audience. The survey is more useful if you want to describe reported generative AI use. Neither establishes the share of all French residents using ChatGPT, since both headline figures cover multiple tools or services.
How common is AI in French businesses?
The France Num barometer, published on 28 September 2026, surveyed more than 9,000 businesses. It found that 40% of very small businesses and SMEs reported using at least one AI solution, up from 26% in 2025 and 13% in 2024. The increases are 14 percentage points in the latest year and 27 points across two years. For a trend chart, use figures from this same series and label the surveyed business group.

Among SMEs specifically, the 2026 figure was 53%, up from 34% the previous year. That is a subgroup of the business population covered by the survey. It is reasonable to report both results, provided you identify which includes very small firms. Using 53% as a headline for every organisation in France would extend the survey far beyond what it measured.
The barometer also distinguishes use cases. It reports 34% using AI to generate text, voice or images, 24% using chatbots, assistants or information-search tools, 13% using it to analyse or classify documents and 11% using it to automate tasks. A firm can report several uses, so adding those percentages would double-count some businesses. Someone preparing a business case could use the breakdown to identify common starting points, such as research or drafting, while treating company-wide deployment as a separate question.
What these figures tell us about AI at work
A firm can count as an AI user even if adoption is concentrated in one team or tool. A retailer might use AI to classify product information while most employees never touch it. Conversely, an employee might use a personal account to prepare a presentation even when the business has not formally adopted an AI solution. The same workplace could look quite different in a survey of businesses and a survey of employees.
For workforce planning, look for a survey that defines the working population, asks where or why AI is used, and distinguishes occasional use from regular use. Role and sector differences matter too. A measure that includes managers, self-employed people and frontline staff may be useful for understanding the economy, but less useful for deciding which tasks need support in a specific organisation.
These three headline sources do not give a single directly comparable percentage for all French workers who use AI at work every week. If that is the number your project needs, leave it unfilled until you have a matching workforce survey. Substituting the 40% business figure might look precise on a slide and lead to the wrong decision.
Comparing France with other countries
Start with the definition and denominator. Compare the same age range, the same question and reference period, and the same type of population wherever possible. A harmonised European survey can support a more defensible cross-country comparison than a mix of national figures. Use the national data for depth, such as detail about very small firms, where the surveys capture something the common measure does not.
Our Germany AI adoption guide and UK AI adoption guide provide local context. Read the population and reference year alongside every figure. If you cannot match the measures, describe the contrast in wording rather than ordering the countries on a leaderboard.
How to cite the data in a report
A usable citation could read “In the 2025 French digital barometer survey, 48% of people aged 12 or over reported using generative AI (2026 edition, Arcep/CRÉDOC).” For business adoption, write “France Num’s 2026 barometer found that 40% of the very small businesses and SMEs covered reported using at least one AI solution.” Each line does more work than a headline by naming who answered, what counted and when the data applies.
If you quote a change, say “percentage points” when subtracting shares. Forty minus 26 is 14 percentage points. A relative growth calculation would instead divide that change by the 26% starting value. Choose the presentation your audience will understand and label it clearly.
A quick way to review a chart before sharing it is to read the subtitle out loud without the headline. It should name the population, the activity and the year. Give the source underneath and retain its wording about “reported use” or “audience reach”. When using slides or a dashboard that combines the public and business figures, put them in separate panels or clearly label each bar so a reader cannot mistake the figures for results from the same survey.
For a UK team considering work in France, the evidence can help frame questions for customers or suppliers. Ask which tools staff use, which tasks they use them for, and whether usage is informal or supported by the business. Those answers will tell you more about training or procurement than assuming a national average describes a particular partner.
Frequently asked questions
Is 48% the figure for 2026?
It comes from a report published in 2026, but the 48% refers to the 2025 survey. Use 2025 for the data point and 2026 for the report edition.
Does 57.1% mean that most French people use ChatGPT?
It describes visits to the AI services covered by Arcom’s audience measurement for March 2026. It is not a ChatGPT-specific usage rate.
Are French small firms using AI more than they were a year earlier?
France Num reports 40% among very small businesses and SMEs in 2026, compared with 26% in 2025. This measures whether firms reported at least one AI use, not how deeply it is integrated.
Can I rank France alongside the UK and Germany using these numbers?
Use a common survey with matched questions, populations and dates to rank countries. The national figures in this guide are useful for understanding France, but they are not a common yardstick.

















