Case Study: Perlego strengthens AI launch readiness with Global App Testing

Perlego uncovers 3 critical AI failures before launch with Global App Testing

Perlego, a digital learning platform, needed to independently evaluate its new AI Research Assistant feature before its full launch. The challenge was to ensure the AI could effectively guide student research without inappropriately doing their work for them, resist manipulation, and handle sensitive situations safely. Perlego engaged Global App Testing's AI GroundTruth service to conduct this human-led stress test.

Global App Testing implemented a structured evaluation where ten UK-based testers interacted with the AI as real students would. The testing revealed the AI's strengths in book discovery and abuse handling but also identified critical risks, such as a tendency to erode its boundaries during multi-turn conversations. These findings provided Perlego with a prioritized list of fixes, giving the team the confidence and evidence needed to improve the feature significantly before its public launch.

Perlego