Students use AI to cheat by embedding hidden instructions in prompts that cause the AI to include a specific quotation, making it detectable but difficult to prove and enforce consequences.

Current plagiarism detectors and AI output analyzers do not detect hidden prompt manipulations that instruct AI to include specific quotations, and there is no automated way to prove cheating for disciplinary action.

Severity

4High

Willingness to Pay

none

Added

May 20, 2026

Workarounds Described

embeds hidden white-font instructions in a prompt to include a specific quotation

Implied Software Gaps

AI prompt forensics tool that detects hidden instructions in student prompts

App Concept

PromptGuard AI

PromptGuard AI is an anti-cheating tool that analyzes student-submitted prompts for hidden instructions or 'Trojan horse' elements designed to manipulate AI responses. It flags suspicious prompts before they are processed, and provides a cheat-detection report with evidence for admin and parents.

Key Features

Real-time scanning of prompts for hidden text instructions (e.g., white font, zero-width characters)
Contextual analysis to detect instructions asking for specific quotations or structured outputs
Automatic generation of a forensic report detailing detected manipulations
Integration with popular AI platforms like ChatGPT, Claude, and Gemini via API or browser extension

Target Users: High school and college instructors and academic integrity administrators in the education industry

Revenue Model: Freemium with $49/mo pro tier for unlimited scans and advanced reporting

Want to go deeper?

Sign up to save ideas, run AI analysis, and track opportunities in your personal workspace. Founding members get full access.

Join Beta

Solutions (0)

Discussion (0)

No comments yet