Critical thinking is rarely measured using a single indicator, but a recent figure raises the question differently: according to an Ipsos survey for EPITA published in February 2026, 92% of French students have used generative AI for their studies. This massive adoption rate far precedes the educational frameworks intended to train discernment. Comparing the speed of adoption of these tools with the pace of deployment of training in digital common sense helps to better understand the extent of the gap.
Generative AI and Critical Thinking: The Gap Between Use and Training
The problem does not lie in the tool, but in the time lag between its diffusion and the capacity of institutions to respond. On one hand, nearly all students integrate generative AI into their daily practices. On the other hand, structured training paths for critically analyzing content produced by these systems are only in their early versions.
| Indicator | Adoption of Generative AI | Training in Digital Critical Thinking |
|---|---|---|
| Rate of Diffusion | Massive since 2023, mainstream use in 2026 | Mandatory Pix AI course starting from the 2026 school year |
| Target Audience | All levels, from middle school to higher education | Students in 8th grade, general and technological 10th grade |
| Institutional Oversight | No general access restrictions | Guidelines from the ministry: frugal use, teacher expertise |
This table highlights a specific point: institutional training arrives three years after the mainstreaming of the tool. Students have appropriated generative AI without a critical framework, and initiatives like Pix AI currently only cover certain classes in secondary education.
In this context, developing one’s common sense in the face of data produced by artificial intelligence involves an approach that Le Bon Sens illustrates well: returning to the fundamentals of reasoning to assess the reliability of information, regardless of its mode of production.
Critical Thinking in the Face of Artificial Intelligence: Three Skills to Strengthen
Critical thinking applied to AI-generated content is not limited to checking whether a text was written by a machine. It involves specific skills that traditional philosophical approaches do not fully cover.
Evaluate the Source When There Is None
A text produced by a language model does not cite its sources in a traceable manner. The critical skill here is to identify the absence of a verification chain. When a student copies a generated response without comparing it to a verifiable publication, the process of critical analysis is short-circuited at the root.
Spot Apparent Coherence Without Factual Basis
Generative models produce fluid and structured texts, creating a credibility bias. A well-written paragraph appears reliable, even when it contains fabricated information. Common sense, in this case, requires separating form from content, a skill that traditional philosophy workshops develop little.
Distinguish Assistance from Substitution
Using AI as an exploration tool (rephrasing a question, synthesizing a corpus) constitutes a reasoned use. Delegating the production of a complete reasoning to it eliminates the very exercise of critical thinking. The Ministry of National Education also emphasizes the use of teachers’ professional expertise as a safeguard and recommends not using AI when a less costly solution suffices.
Pix AI Pathway and School Oversight: What Changes with the 2026 School Year
The Pix AI pathway becomes mandatory starting from the 2026 school year for 8th grade and general and technological 10th grade students. This initiative marks a regulatory turning point: for the first time, training in digital critical thinking enters a mandatory and measurable framework.
The guidelines published by the ministry establish three operational principles:
- A frugal use of AI, to limit environmental impact and avoid systematic dependence on the tool.
- The prohibition of using AI when a less ecologically costly solution is available, which requires justifying each use.
- The maintenance of teacher expertise as the primary reference in evaluating content and reasoning.
However, these guidelines do not cover higher education, where the rate of generative AI use is the highest. The gap between secondary education, now regulated, and higher education, still largely unregulated in this regard, constitutes a blind spot.
Common Sense and Digital Data: Beyond Classical Philosophy
Descartes defined common sense as the natural ability to distinguish the true from the false. This definition remains valid, but its application has changed terrain. When information arrives in the form of generated text, aggregated data, or automated summaries, common sense requires verification reflexes that intuition alone does not provide.
The educational approaches identified by the National Education Scientific Council (CSEN) emphasize a often overlooked point: critical thinking is not a general disposition one possesses or lacks. It is a set of contextual skills. A person may demonstrate excellent discernment in their professional field and accept without question fabricated information in a domain they do not master.
This specificity makes training in critical thinking more complex than a simple methodology course. It requires identifying situations where common sense is lacking, not just reminding that it should be present.
The deployment of Pix AI and the ministerial directives on the frugal use of artificial intelligence in education establish an initial framework. The data that remains to be monitored is that of higher education: as long as 92% of students use generative AI without mandatory critical pathways, the gap between practice and discernment will continue to widen.