In 2026, the topic ofcheating in the classroomis no longer just about “a phone under the desk” or cheat sheets. The combination ofoff campus ai(models and services accessible via the web), increasingly discreet wearables, and fast searching through communities and social media is changing how generative AI can physically enter the classroom even during in-person tests. For teachers, the challenge is not only to “catch” improper use, but to design contexts and prompts that make any shortcut less worthwhile, and that strengthen real learning and academic integrity.
This article proposes a teaching-and-operational approach: understand how the phenomenon is changing, recognize signals without slipping into a “witch hunt,” update policies and prompts foracademic integrity ai, and rethink in-person assessments in a sustainable way. This is not about demonizing AI: the point is to move it from a clandestine tool to a transparent, declared, and assessable resource.
Why in 2026 “in-person” cheating is changing: Off Campus AI, smartwatches, and social search
For years, anti-cheating strategies have focused on surveillance, spacing between desks, collecting phones, and different versions of the test. In 2026 these measures remain useful, but they are not enough, because AI is no longer just “a website” to open: it is an ecosystem of services and channels that can be activated quickly and with little visibility.
When we talk aboutin-person assessments ai, the novelty is the “portability” of assistance: micro-consultations, rewrites, translations, guided solutions, and even suggestions on how to structure a line of reasoning. This reduces the difference between “studying beforehand” and “support during,” especially if the test measures only the final product and not the process.
Three vectors make traditional controls based only on bans obsolete:
- Wearables and accessories: smartwatches, discreet earbuds, “smart” pens, or devices that act as a bridge to a phone in a backpack. Even without obvious screens, they can enable quick input/output.
- Web apps and “off campus” models: generative AI services accessible via browser, often optimized for mobile and with concise response modes. The point is not the model’s power, but speed of access.
- Social search and communities: searching for solutions or explanations via social channels, class groups, forums, and instant sharing of photos of the assignment. AI can be “mediated” by other people or by bots in the channels.
In this scenario, talking only about surveillance risks shifting the educational relationship onto a plane of permanent control. A more effective response is to combine: (1) clear, shared rules, (2) test design that makes AI less useful as a shortcut, (3) opportunities for declared AI use in the study pathway, so students don’t perceive the only option as “hiding.”
Real risks and weak signals: what to observe without falling into a witch hunt
The main risk is not only the “perfect” answer, but the replacement of the cognitive process: AI can provide structure, intermediate steps, examples, and disciplinary language, making competencies appear that are not yet consolidated. The teaching consequence is a less valid assessment: it measures the ability to obtain an answer, not to build it.
The most frequent patterns of improper use during in-person tests (not exhaustive) include:
- Answers with a suddenly more mature or “standardized” register, not very consistent with previous work, but without a corresponding improvement in guided exercises or oral questioning.
- Correct solutions but with “opaque” steps: the trace of personal reasoning is missing, generic justifications appear, or formulas/terms not covered in class are used without being able to explain them.
- Suspicious uniformity across papers (same structure, same examples, same “elegant” mistakes), typical of a common source or shared prompts.
- Micro-consultation behaviors: repeated glances at the wrist, hands often under the desk, frequent requests to go to the bathroom, “nervous” handling of personal items.
It is important, however, to distinguish between clues and proof.ai detection schooltools (AI “text” detectors) have known limits: false positives with very proper students or with specific educational needs who use proofreaders; false negatives with reworked texts; poor reliability on short answers. Using them as the “smoking gun” can increase conflict and distrust.
A balanced approach is based ontriangulation: consistency with prior evidence (in-class work, oral exams, exercises), analysis of the process (drafts, steps, reasoning), and a brief verification conversation (“explain how you got there”). Often 2–3 targeted questions are enough to understand whether the competence is authentic, without turning the classroom into a courtroom.
Finally, watch the climate: if the perception is that “everyone cheats with AI,” normalization of the shortcut increases. Making criteria, rationales, and legitimate alternatives visible is already prevention.
Updated rules and prompts: academic integrity AI policies applicable in class
An effective policy is not a list of generic bans, but a system of expectations that makes clear what is being assessed and how the work is documented. To be applicable in class, it must be short, repeatable, and tied to rubrics and prompts.
who we are
- Three high-impact teaching uses, consistent with an academic integrity AI policy:
- Study planner and routines: have the student build a weekly plan with measurable goals (what I review, what I practice, how I check). In the assignment you can ask them to attach the plan and a brief reflection on what worked.
- In short: in 2026 cheating in the classroom changes because AI can “enter” the classroom in small and rapid ways. The most robust response is a combination of clear policies, process assessment, targeted oral checks, and opportunities for transparent AI use in the study pathway. This way integrity is not only a constraint, but a competence that is taught and trained.
Oral simulations: practice short explanations, definitions, examples, and counter-questions. If you then adopt oral checks during the test, the simulation reduces anxiety and increases fairness.If we want to reduce improper AI use during tests, a powerful lever is to offer a legitimate and guided channel to use itOperationally, you can ask students to use AI in a declared way during preparation (for example to generate questions, summaries, or outlines), but to bring to class only analog evidence: concept map, worked steps, personal notes. If you want to experiment, you can
- and define a short assignment with mandatory disclosure. To explore the educational setup and the project philosophy, you can also consult
- .
- The key step is to make explicit the distinction between: AI as study support (allowed and traceable) and AI as substitution for performance during a test (not allowed). When this distinction is clear and practiced, the appeal of clandestine off campus ai use during the test decreases.
A useful practice is “double evidence”: (1) a short, targeted written output, (2) a mini oral explanation or a process note. In this way AI can also be used during the study phase, but during the test personal mastery emerges.
How StudierAI can help teachers and students: AI planners, quizzes, and oral simulations for transparent use
If we want to reduce improper AI use during tests, a powerful lever is to offer a legitimate and guided channel to use it
start for freeand define a short assignment with mandatory disclosure. To explore the educational setup and the project philosophy, you can also consultwho we are
Three high-impact teaching uses, consistent with an academic integrity AI policy:
- Study planner and routines: have the student build a weekly plan with measurable goals (what I review, what I practice, how I check). In the assignment you can ask them to attach the plan and a brief reflection on what worked.
- In short: in 2026 cheating in the classroom changes because AI can “enter” the classroom in small and rapid ways. The most robust response is a combination of clear policies, process assessment, targeted oral checks, and opportunities for transparent AI use in the study pathway. This way integrity is not only a constraint, but a competence that is taught and trained.
- Oral simulations: practice short explanations, definitions, examples, and counter-questions. If you then adopt oral checks during the test, the simulation reduces anxiety and increases fairness.
- Operationally, you can ask students to use AI in a declared way during preparation (for example to generate questions, summaries, or outlines), but to bring to class only analog evidence: concept map, worked steps, personal notes. If you want to experiment, you can
- and define a short assignment with mandatory disclosure. To explore the educational setup and the project philosophy, you can also consult
.Three high-impact teaching uses, consistent with an academic integrity AI policy:The key step is to make explicit the distinction between: AI as study support (allowed and traceable) and AI as substitution for performance during a test (not allowed). When this distinction is clear and practiced, the appeal of clandestine off campus ai use during the test decreases.
A useful practice is “double evidence”: (1) a short, targeted written output, (2) a mini oral explanation or a process note. In this way AI can also be used during the study phase, but during the test personal mastery emerges.
How StudierAI can help teachers and students: AI planners, quizzes, and oral simulations for transparent use
If we want to reduce improper AI use during tests, a powerful lever is to offer a legitimate and guided channel to use itOperationally, you can ask students to use AI in a declared way during preparation (for example to generate questions, summaries, or outlines), but to bring to class only analog evidence: concept map, worked steps, personal notes. If you want to experiment, you canstart for freeand define a short assignment with mandatory disclosure. To explore the educational setup and the project philosophy, you can also consultwho we are
Three high-impact teaching uses, consistent with an academic integrity AI policy:
- Study planner and routines: have the student build a weekly plan with measurable goals (what I review, what I practice, how I check). In the assignment you can ask them to attach the plan and a brief reflection on what worked.
- In short: in 2026 cheating in the classroom changes because AI can “enter” the classroom in small and rapid ways. The most robust response is a combination of clear policies, process assessment, targeted oral checks, and opportunities for transparent AI use in the study pathway. This way integrity is not only a constraint, but a competence that is taught and trained.
- Oral simulations: practice short explanations, definitions, examples, and counter-questions. If you then adopt oral checks during the test, the simulation reduces anxiety and increases fairness.
Operationally, you can ask students to use AI in a declared way during preparation (for example to generate questions, summaries, or outlines), but to bring to class only analog evidence: concept map, worked steps, personal notes. If you want to experiment, you canstart for freeand define a short assignment with mandatory disclosure. To explore the educational setup and the project philosophy, you can also consultwho we are.
The key step is to make explicit the distinction between: AI as study support (allowed and traceable) and AI as substitution for performance during a test (not allowed). When this distinction is clear and practiced, the appeal of clandestine off campus ai use during the test decreases.
In short: in 2026 cheating in the classroom changes because AI can “enter” the classroom in small and rapid ways. The most robust response is a combination of clear policies, process assessment, targeted oral checks, and opportunities for transparent AI use in the study pathway. This way integrity is not only a constraint, but a competence that is taught and trained.
