Blind spot 01
US over-representation
Most benchmark studies focus on American users. Yet behaviours, cultural contexts and attitudes to technology differ significantly. What holds true in San Francisco doesn't necessarily hold in Lyon or Bordeaux.
AI Observatory
A long-term research programme.
What is happening with AI is not just another wave of technology. It is a deep reshaping of our relationship with information, with knowledge, and between individuals and organisations. Understanding it is what makes it possible to act.
Why this observatory
The available data has three blind spots.
Blind spot 02
Rapid obsolescence
The AI landscape moves so fast that findings published before 2024 may already be partly out of date. The behaviours we saw 18 months ago have mutated. A frozen data point is not a reliable compass.
Blind spot 03
Self-reported data
The vast majority of studies rely on surveys: people are asked what they do. The problem: people don't always know what they really do. Only in-situ observation can capture unconscious trade-offs, the contexts in which AI gets called upon, and the red lines nobody puts into words.
What's in it for you?
From insight to decision, to design.
01 · Understand
Feed your strategic thinking right now
These insights shed light on concrete questions for your business: how are your audiences already using AI in areas that matter to you? Which behaviours are being reconfigured? What does this change for your digital ecosystem, your content strategy, your customer relationship?
02 · Map
Identify what you're missing to decide
Our data is a solid first building block. It also helps map the blind spots, what we don't yet know about your specific audiences, your sector, your internal practices, and pin down exactly what information is missing to make truly informed decisions.
03 · Design
Produce that information and design the best solutions
Targeted sector surveys, ethnographic studies of your specific audiences, strategic design workshops, design of business-specific AI services: we can produce that missing data with you, then co-design from it the schemes, strategies and services that will let your company define its role in the age of AI agents.
Our approach
Two anchors: academic rigour and French fieldwork.
Strand 1
International literature review
A systematic review of 50+ sources published between 2022 and 2025. Academic studies, industry reports and market analyses, selected for their methodological rigour. A critical reading that highlights population biases and limitations. Continuously updated.
Strand 2
French ethnographic studies
A qualitative study conducted in successive waves with a panel of 15 participants split into two groups:
- Regular users (12 participants), frequent users aged 25–39 with varied profiles.
- Power users (3 participants), intensive experts who help anticipate the behaviours the rest of the population will adopt within 12 to 24 months.
What ethnography allows that surveys don't: observing real practices, understanding the context in which AI is called upon (or deliberately set aside), and grasping the implicit trade-offs people never spell out.
First insights
What we're already observing.
A few findings from our first ethnographic wave.
Insight 01
“AI isn't replacing experts, it's filling the blind spots.”
In every area we observed, AI steps in where the system leaves gaps. Patients turn up to appointments better informed, more precise, sometimes more critical. The relationship with the doctor is evolving, but the doctor remains. What changes: the level of expectation, and the nature of what people expect from professionals and institutions.
Insight 02
“People make fine-grained trade-offs, not binary choices.”
They delegate what feels low-risk: logistics, writing, summarising. They keep control over anything touching their physical, emotional or financial integrity. That isn't irrational: it's a sophisticated trade-off logic that self-reported data simply cannot capture.
Insight 03
“The search engine monopoly is being challenged.”
Among regular users, Google use is down by a factor of two or three, reduced to ultra-specific queries or source-checking. Brand content is absorbed, synthesised and reinterpreted by agents. Users get an answer without ever visiting the website.
Among our current lines of research
What we're digging into now: sovereignty at the level of everyday use
Observation work under way
unknowns is opening a new line of observation: the real place of sovereign AI tools in French people's daily lives. The debate on digital sovereignty usually plays out at the level of summits, regulation and infrastructure. Our hypothesis is that it plays out just as much (perhaps above all) in a series of micro-gestures: which tool you open in the morning, which reflex you let settle in, at what point you choose familiarity over any other criterion.
This is precisely what the available data doesn't look at. We measure intentions, market shares, opinions on sovereignty. Nobody observes the gesture itself: the gap, if there is one, between what people say they want and the tool they actually reach for, without thinking.
We don't have an answer yet. It's ground we're just beginning to explore, with the same method as the rest of the observatory: observe real practices before commenting on them. Because before trying to shift a reflex, you first have to understand it.
Bibliography
50+ sources classified, sorted, annotated.
Over time we are building a bibliography of academic papers and industry reports on AI usage. Each source is annotated, dated and set against our own field observations.
Browse the bibliography →Continuously updated · Hosted on Notion
FAQ
What we're asked most often.
Field observation shows that real-world use is more nuanced than the discourse suggests: people don't hand everything over to AI, they make fine-grained trade-offs. They entrust AI with what feels low-risk (summarising, writing, logistics) and keep control over anything touching their physical, emotional or financial integrity. unknowns' Observatory of AI Usage documents these behaviours in France, through ethnography.
Among regular users, reliance on Google is falling sharply, reduced to very specific queries and source-checking. Brand content is increasingly absorbed and rephrased by AI agents, which deliver an answer without the user ever visiting the website. For brands, visibility is no longer just about traditional SEO, but about how AI cites and rephrases their content.
Yes, and it's a blind spot in the available studies: most focus on American users, whose cultural contexts and attitudes to technology are different. What holds true in San Francisco doesn't necessarily hold in Lyon or Bordeaux. That's why unknowns runs its own ethnographic observatory, rooted in French fieldwork.
It's a long-term research programme on how artificial intelligence is actually used in France. It rests on two pillars: a critical review of the international literature and a French ethnographic study, continuously updated. Its purpose: to observe what people really do with AI — not what they say they do — to help organisations make decisions.
Presentation on request
Want to go further?
We present the observatory's findings on request, live, in an interactive format designed to surface the concrete implications for your business. Three themed blocks, discussion breaks, and questions that speak directly to your sector.
Request a presentation →