Khan Academy has launched new visual features and teacher controls for its AI assistant, Khanmigo, to address low student engagement and improve learning outcomes. Developed in partnership with Google.org, these updates attempt to change passive textbook-style tutoring into an interactive experience. Educational research suggests that parents and teachers must play a key role to prevent students from relying too much on AI.
What Happened
In an official announcement, Khan Academy detailed a series of updates powered by Google's Gemini models. The primary addition is interactive visual diagrams. When a student struggles with a geometry or physics concept, Khanmigo generates a responsive diagram instead of just text. Students can drag lines or shift variables, and the AI responds to these changes.
For teachers, the update integrates Gemini into the 'Practice My Knowledge' tool. Educators can now generate multiple-choice questions in seconds. The system requires teachers to review and edit questions before sending them to students. This addresses previous limitations where Khanmigo tools for classrooms generated practice questions but offered teachers no way to edit or control them.
The Bigger Picture
Khan Academy's internal testing shows progress. According to data published by the AIMS Collaboratory, Khanmigo’s new interface drove a 6.1 percent improvement in "next-item correctness," which is a metric measuring if a student can answer the very next question independently.
However, researchers warn against relying solely on this metric. As we previously reported in our analysis of how Khan Academy is redesigning its tutor, sequence models that predict next-item actions often track statistical patterns rather than genuine cognitive growth. A study by the Wharton School highlighted a "productivity paradox" in AI learning. Students using unrestricted AI help saw practice scores jump by 48 percent, but their actual exam scores fell by 17 percent unless the AI used strict cognitive scaffolding (The Learning Standard).
Low engagement also remains a major bottleneck. A two-year experimental study by the National Bureau of Economic Research across 18 Tennessee middle schools found that while assigning Khanmigo raised math achievement by 1.3 national percentile ranks, these gains were statistically identical to students practicing on Khan Academy without AI. The researchers discovered that the median student only messaged the AI tutor on a third of their practice days, and in only 17 percent of sessions where they made a mistake. Internal estimates reveal that only 15 percent of eligible students actively chat with the tutor.
There are also quality concerns with AI-generated material. A study evaluating thousands of GPT-4-generated math questions found that only 68.74 percent were ready for immediate classroom use. Nearly a third required editing or had to be discarded. Research on educational mathematics also shows that while top AI models perform well with visual diagrams, their accuracy drops to near zero if visual context is missing.
What This Means for Families
Interactive visuals work well when integrated correctly. For example, a study in Scientific Reports showed that adaptive technologies with real-time feedback lead to deeper subject mastery. Another study on concept-oriented science frameworks found that visual tools that model invisible processes improve student motivation and achievement.
However, AI is not a set-it-and-forget-it tutor. Because students tend to ignore AI assistance or use it to get quick answers rather than learn, parents and teachers must actively guide its use.
What You Can Do
To help students succeed, parents and teachers should provide active oversight instead of letting them use AI unsupervised. Educators need to carefully review all AI-generated questions, since roughly one-third of AI content requires adjustments to meet school standards. Adults should also encourage students to use the new visual sliders and interactive graphs in Khanmigo rather than asking the chatbot for quick text-based answers. Finally, to counteract the productivity paradox, check whether students are using AI as a shortcut and ask them to explain the concepts in their own words.