> For the complete documentation index, see [llms.txt](https://examind.gitbook.io/v1/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://examind.gitbook.io/v1/our-approach/cheating-modes.md).

# Understanding Cheating Modes

Our analysis of how students cheat groups today's methods into a handful of categories. The landscape has shifted toward **AI-assisted cheating**, but several long-standing methods remain widespread — and any integrity strategy designed before modern AI is worth re-evaluating.

## AI-assisted cheating — the fastest-growing threat

Since modern large language models arrived, AI-assisted cheating has become the most rapidly evolving category. It isn't one thing — students can reach AI in several ways during an exam, for example:

* AI built into the browser or the device they're testing on.
* A **second device** (like a phone) used to look up or generate answers. Because it never touches the exam device, this is one of the hardest vectors to catch.
* Emerging vectors on the horizon — AI-native browsers, smart glasses, and on-device AI agents.

The takeaway: strategies built for the pre-AI era need to be re-examined against live AI use.

## Other persistent methods

* **Contract cheating** — someone other than the student does the work: impersonation, sharing account access, or letting someone else operate their device.
* **Leaked content** — questions or answer keys shared after an exam (for example, on sites like Chegg, Course Hero, or Quizlet), or leaked ahead of time.
* **Copy cheating** — copying another student's answers in real time (over the shoulder, via messaging, or relayed by an early finisher).
* **Unauthorized resources** — notes, a second monitor, or switching to other tabs or apps to look things up.

## Where EXAMIND's design helps

EXAMIND's [dynamic question engine](/v1/examind-platform/build/question-builder/dynamic-questions.md) gives every student a unique variant, so **shared, leaked, and copied answers lose their value** — a structurally strong defense against the answer-sharing, copying, and leaked-content family.

Design alone, though, doesn't neutralize AI-assisted, contract, or second-device cheating. Those call for complementary layers — most importantly [online proctoring](/v1/our-approach/online-proctoring.md) for remote, high-stakes exams.

## Match your defenses to your setup

Where the greatest risk lies depends on how and where students take the assessment:

* **Testing center · institution devices · in-person proctoring.** The environment is locked down, so the main risk is **leaked content** — questions circulating between sections or semesters. [Dynamic questions](/v1/examind-platform/build/question-builder/dynamic-questions.md) address this directly: every student gets a unique variant, so leaked answers are worthless and you don't have to rewrite exams each term.
* **Testing center · students' own devices · in-person proctoring.** The main risk becomes students using **AI on the test-taking device itself**. EXAMIND's [tactical deterrents](/v1/examind-platform/deliver/tactical-deterrents.md) help control this during the attempt.
* **Remote exams on students' own devices.** The hardest case: students could use **AI on any device**, in view or not. Here [online proctoring](/v1/our-approach/online-proctoring.md) becomes essential alongside EXAMIND's assessment design.

**The risks compound.** Each setup carries the risks of the more controlled ones plus its own — a remote exam faces leaked content *and* on-device AI *and* off-device AI. Match your defenses to your setup.

{% hint style="info" %}
Designing an integrity strategy for your specific setup? [Talk to us](mailto:support@examind.io) — we'll help you choose the right combination of controls.
{% endhint %}