AI, a new cheating tool or a driver of exam transformation?
The study by Ipsos for the online assessment platform Theia[1],, carried out from 25 March to 3 April 2026 with a representative sample of 1,000 young people aged 18 to 25 in France, shows just how much artificial intelligence has become part of students’ daily lives. In detail, 80% of respondents say they use generative AI at least once a week, while 41% use it daily. Among men aged 18 to 19, 55% say they use these tools every day.
This widespread use goes far beyond just school and affects personal and professional life too. The survey shows that this change is completely transforming how they see learning, with 77% thinking we now need to learn differently, focusing on reasoning, analysis, and the ability to cross-check information rather than just memorizing.
AI is becoming a work tool… but also a way to cheat
The study also highlights a phenomenon already well established in schools, namely the use of AI to bypass assessment rules. More than one in two young people, or 54%, admit to having used it to cheat on homework or exams. This practice is especially common among students, with 59% admitting to having used these tools fraudulently, and men are also more likely to admit to it.
Exams seen as unsuitable
This major shift in how things are done comes with a no-nonsense questioning of the testing system. According to the survey, 63% of young people think exams don’t really reflect their true level. Plus, 58% of respondents believe that getting a degree isn’t essential to succeed professionally, showing a bit of mistrust in the value of diplomas without questioning the worth of studying itself.
To make exams fairer, several ideas are widely supported by respondents: 86% want topics that focus more on reasoning and intellectual approach, 82% support strengthening continuous assessment, 78% are in favor of exams with allowed resources, and 74% prefer teaching students how to use AI rather than banning it. Despite this strong push for modernization, young people still value some traditional benchmarks since 85% think it’s essential that a human continues to mark papers, especially in subjects requiring analysis or judgement. The criteria considered most important for ensuring a fair exam are also very traditional, relying on the same conditions for everyone, supervised exams, transparent marking schemes, anonymous papers and preventing cheating.
Computer-based homework is losing credibility. The study also highlights a strong mistrust of certain assessment formats, especially since the rise of generative AI. Desk and paper homework are still largely seen as the most reliable way to reflect students’ actual level, with 69% of young people ranking them among the most relevant formats, just ahead of oral exams with 64% favorable opinions. Ipsos’ survey ultimately shows a strong expectation among young people, who aren’t calling for the end of traditional exams, but want assessments that are better adapted to AI, which is now an everyday part of life.
And what about oral exams?
So, many teachers now choose to hold knowledge tests orally despite the extra workload; you can immediately see if the student really knows their stuff. But recently, a new idea has emerged, which is how to use AI to prepare for an oral[2] exam.
Actually, 80% of success in an oral exam comes down to preparation, not just the day of. The problem is that most students revise for their oral exams like they would for written ones — by going over their notes. An oral exam is best prepared by speaking, being asked questions, and structuring your thoughts out loud. AI can become a training partner available 24/7. Generally, for any oral exam, you can structure your answers using the STAR[3] method:
• Situation: set the context,
• Task: what is the problem or question,
• Action: what was proposed or demonstrated,
• Result: what conclusion or impact.
In the end, AI is pushing us to rethink exams and knowledge tests, as student-written work is losing credibility. Oral exams seem like an interesting route, especially since AI-based tools help students get ready for these kinds of tests.
[1] https://www.theia.fr/etude-ipsos-bva-theia-jeunes-examens-ia/
[2] https://innovaweb.fr/fr/blog/preparer-oraux-examens
[3] https://cparfey.fr/wp-content/uploads/2025/10/methode-star.pdf