Klassy is built on peer-reviewed research from Harvard, Google DeepMind, and the National Council of Teachers of Mathematics — and tested extensively for child safety before any student interaction.
Click each category to explore the research, safety architecture, pedagogy, and gamification that power Klassy.
A 2025 randomized controlled trial found that students using a pedagogically engineered tutor learned significantly more in less time than students in active in-class instruction — with effect sizes between 0.73 and 1.3 standard deviations.
Read the studyA 2025 RCT across five UK classrooms found that question-driven tutoring matched or outperformed expert human tutors on every learning outcome — with zero harmful content instances across 3,617 reviewed messages.
Read the studyThe National Council of Teachers of Mathematics is explicit: culturally responsive math instruction is not optional. Standards-based teaching alone is insufficient to achieve equitable outcomes.
Read the position statementEvery student message is assigned a content score before it reaches the tutor. Inputs that fall below the safety threshold end the session and log the event.
The tutor is scoped to discuss only the active lesson — never personal topics, off-topic content, or anything outside the Klassy curriculum.
Every tutor response passes through a separate scoring pass against a content safety rubric before it is delivered to the student.
Reviewed by safety auditors. Zero instances of harmful or risky content. Factual error rate of just 0.1%.
Klassy's learning model is grounded in question-driven pedagogy — the tutor never gives answers. Instead, it asks carefully sequenced questions that guide students to discover solutions themselves.
Our CRA progression (Concrete → Representational → Abstract) is the structural backbone of every lesson. A 2025 meta-analysis of 30 studies found a CRA effect size of 0.9965.
Read the meta-analysisA 2024 synthesis of 15 peer-reviewed studies found that integrating cultural context into math instruction narrows achievement gaps for diverse learners.
Read the CRMT researchA 2023 systematic review found that game-based learning in mathematics affects both conceptual understanding and student motivation — not just surface engagement.
Frontiers in PsychologyA 2024 review found that game mechanics increase intrinsic motivation and self-confidence, particularly for students with low-to-moderate confidence in math.
Frontiers in Education