Google and Khan Academy announced the availability of a new set of artificial intelligence tools within Khanmigo at the start of the 2026 academic year, after moving the features from the early experimentation phase to use in actual classrooms. The tools are based on Gemini models and focus on helping students understand mathematics and science concepts, while enabling teachers to prepare training materials that can be reviewed and edited.
The announcement extends a collaboration that began earlier in the year and included a grant from Google.org and direct access for Khan Academy experts to Gemini models. Since March, six Google engineers have worked with product and engineering teams at Khan Academy as part of a six-month Google.org fellowship, with the aim of integrating the new capabilities into the Khanmigo platform, which presents itself as an assistant teacher for students and an assistant for teachers.
Charts That Respond to the Student’s Steps
Rather than limiting itself to text answers, Khanmigo can now create interactive visual aids when the system determines that a drawing could help solve a problem or clarify a concept. Examples include charts, geometric diagrams, parallel coordinate curves, and other visual elements.
Students can directly modify some elements of the drawing, such as dragging a geometric vertex or a line segment, while the system recognizes the change and adjusts the explanation or lesson accordingly. Khan Academy says its teams identified indicators for evaluating chart quality, then tested their relevance to the problem, usefulness, and clarity. It also added a feedback tool within Khanmigo to collect user opinions and improve the experience.
The organization noted early indications of increased student engagement and improved performance in answering questions after the charts were displayed, but it did not publish detailed figures or independent comparative results in the material.
Teachers Review Content Before Distribution
The second feature is a redesign of the Practice My Knowledge tool, which originally targeted the generation of multiple-choice questions. The new version allows teachers to use Gemini to plan, formulate, and verify questions, with the option to upload their own materials for the system to use as a basis.
Before sending the assignment to students, teachers can review the entire set and edit or reject questions. After students finish the work, teachers receive detailed performance reports to help them with assessment. According to the announced design, Gemini creates the drafts, but the teacher retains the final decision about what students see.
Why Does This Announcement Matter?
The practical change is not limited to adding an artificial intelligence model to an educational tool; it combines interactive individual assistance for students with a workflow that requires human review of content before it is used. This could benefit schools and educational districts that want to take advantage of automated generation without handing the system the task of publishing questions directly.
Nevertheless, the announcement’s limitations remain clear: the indicators related to engagement and achievement are preliminary, and the material does not include independent research findings, details about accuracy standards, mechanisms for protecting student data, or the geographic availability range. Human review also reduces the risk of passing along an unsuitable question, but it does not eliminate the need for teachers to check the difficulty level, content accuracy, and alignment with the curriculum.
The new features are available to educational districts that use Khanmigo, according to Google and Khan Academy, without additional details about pricing or subscription requirements.