MagicSchool
MagicSchool is an AI operating system for schools. It brings together tools that help teachers plan, assess, differentiate instruction, and provide feedback. Students engage with AI through teacher-developed and guided experiences.
What it looks like in action
What the students do:
When assigned to a room by their teacher, students work independently on a teacher-created activity at their own pace. With the Class Writing Feedback feature, a student drafts a piece of writing, submits it, and gets feedback aligned to their teacher's rubric, which the student uses to revise. Students can also interact with teacher-calibrated chatbots to practice targeted skills, talk through concepts, or interview historical figures or fictional characters. Feedback can be read aloud to students, and multilingual learners can work in their home language.
What the AI does:
On the teacher's side, the AI uses teacher-uploaded documents and/or prompts to produce first drafts of the material: a leveled version of a text raised or lowered to a target reading level, a lesson plan, a quiz, a rubric, or a set of practice items. It can adapt the materials to specified needs (e.g., for students with IEPs or multilingual learners) and translate them into other languages. The AI scores student work against a teacher-provided rubric, returning suggested scores and rubric-based comments, and creates student groups based on their performance. On the student side, the AI functions as a custom chatbot. In addition to providing feedback, it can carry on a back-and-forth conversation, responding to what the student writes.
What the teacher does:
Teachers use MagicSchool to make and adapt the materials they use to teach. Starting from documents they upload or prompts, they generate draft materials, which they edit before use—because the AI produces a starting point rather than a finished product. Teachers review and edit what the AI produces before using it, including the scores it suggests when grading student work against their rubric, and refine their setup based on the results. Teachers can also create custom chatbots for students to interact with, setting specific learning objectives, boundaries, and time limits. While students work, teachers can watch what's happening from a dashboard that shows each student's activity and step in to support individual students.
What the leaders do:
School and school system leaders control which of MagicSchool's tools their teachers and students can use, turning specific features on or off. Leaders can also see how teachers and students across the system are using MagicSchool through system dashboards to help shape implementation.
Reach
––
Students
8 million+
Teachers
13,000+
Schools
––
School systems
Use
Recommended use
None provided
Actual use
TBD—pending data
Product dimensions
Subject
All subjects
Grade span
K–12
Curriculum alignment
Standards-only: MagicSchool generates from the materials a teacher or school system gives it, adapting and differentiating from uploaded documents or prompts, and content can also be tagged to state standards; the teacher supplies the curriculum connection.
Design scope
Multi-function use
The way AI interacts with students
- Both students and AI asks questions
- Rubric-scored or written feedback (beyond right/wrong)
- AI gives feedback directly to students
Teacher guidance
Offers a range of uses (no specific teacher nudges)
Social vs. solo use
Solo by default — primarily independent, small groups possible
Teacher override
Teacher can override the AI scoring
Product adoption
Both—teacher-adopted and system-adopted
Dashboard scope
Teacher and leader dashboards available
Dosage recommendations
Does not provide dosage recommendations
Users' take on wins and limitations
There are clear trends in what users (i.e., teachers, students, instructional leaders) see as the wins and limitations of each product. For each trend below, we've included one illustrative quote from our user interviews.
Wins
MagicSchool scores student work against the teacher's rubric, so instead of starting from zero the teacher is checking and adjusting scores that are already done. Because it judges writing against the criteria rather than against a teacher’s expectations for a particular student, some teachers report it helps check their own biases.
"It does take a little bit of the bias out of it, because you're not like, 'oh, this is so good for [Student Name]… I'm so proud of him' and then giving him a three when it's really sold a one….It keeps that level there. And then it saves so much time, because it's already graded and you're just kind of checking that you agree with it, instead of starting from zero."
— Teacher, 5th Grade ELA
Teachers can upload grade-level work and have MagicSchool level it for students with IEPs and multilingual learners, so those students have the scaffolds they need and are able to work through the task independently.
“Students with IEPs usually get instruction below grade level, but now I can give them access to the grade-level curriculum….My special education students need access to grade-level work, even if it is tricky and hard. They have the right to that, and, now, I can give them access to it with scaffolds and supports."
— Teacher, 4th Grade Special Education
Rather than using generic writing feedback, teachers program the tool around what they are actually teaching.
"We'll put in our rubrics, and we'll add, like, other information that we want it to do…our kids, their writing's really wordy, so we talk a lot about being more concise—the economy of language."
— Teacher, 8th Grade ELA
Limitations
MagicSchool's writing feedback is reliable on grammar, mechanics, and whether a rubric criterion is met, and it does not push students equally on the content of their thinking. Teachers report advanced students, in particular, find less to act on.
"[We] usually can't just take MagicSchool's stuff and completely apply it on our own for students at a lower level. And on the higher end, many of my advanced students do not find it useful…helpful for catching errors, typos, and run-on sentences, but not content."
— Teacher, 5th Grade ELA
Teachers describe needing guidance on which features to use when, and some settle into a few tools rather than working across the platform.
"They have so many tools but it can be overwhelming. I don’t use all the features. Mostly what I use MagicSchool for is the student rooms.”
— Teacher, 8th Grade ELA
MagicSchool's scores can shift on a student without their work changing. When that happens it reads as arbitrary to students and erodes their trust in it, so the score holds up only with a teacher's review behind it.
"There are times where students will be typing and MagicSchool will kick them out. They go back in and it gives them a lower score but it's on the same text conversation. That's kind of frustrating for them because they're saying things like, 'I had 17 points yesterday, now I have 11.' And I'm trying to tell them it doesn't matter. The point is you're always going to improve."
— Teacher, 5th Grade ELA
Our instructional take
Instructional wins
- By directly providing feedback and by making it easier for teachers to give more feedback, MagicSchool increases the amount of feedback students get, which is key to better learning.
- MagicSchool makes it easy for teachers to set up a room using their own materials, integrating work on the platform and core learning smoothly.
- Using AI to get to a first draft rubric or presentation can save teachers considerable time, allowing them to spend more time connecting with their students and focusing their attention on students who need the most support.
- Developing students’ AI literacy is not a central question on the tour, which focuses on how AI helps students learn existing standards. Still, we were struck by how many schools value MagicSchool as a bounded way to help students begin to learn about ways to use AI, particularly to help improve their writing. Users reported this feels like a safe way to help students begin to learn about AI in a protected space.
- MagicSchool rooms and chatbots can be built either by teachers or centrally by the school system’s academic team. This creates an opportunity for system teams to train and calibrate feedback support tools in a way that could lead to more consistent feedback for students across schools than we typically see with paper-pencil instruction.
Instructional limitations
- Using AI to get to a first draft of a lesson plan, rubric, or presentation can reduce the quality of thinking teachers put into their plan for the day, undermining their readiness to support the real-time instructional decisions needed to support learning at the level of the standard.
- When teachers develop novel materials, it can pull students out of the core curriculum in ways that create gaps in learning. Ensuring this does not happen requires strong teacher support and system-level planning.
- Lessons generated for ELA can emphasize comprehension skills in isolation (i.e., instruction that does not include a text, which has not been shown to support learning) as opposed to skills in the context of a text.
- Tools that allow for quick differentiation and text releveling can lead to a reduction in the expectations for students if they are not reviewed carefully with attention to the expectations for students.
Implementation considerations
- Like all platform products we studied, MagicSchool requires leaders to be clear with teachers about acceptable use and unacceptable use, including expectations for how teachers will review AI-generated feedback and lesson materials before assigning them to students.
- If a school system plans to make MagicSchool a part of the instructional program, leaders will need to decide what parts of the curriculum it can be used to support and what parts it cannot be used to support. If they build a room (i.e., custom chatbot) centrally, they need to invest time for relevant training and quality testing of the feedback. If they delegate room building to teachers, they need to train teachers how to train the AI and check the quality of its feedback.
- Leaders need to think through their expectations for what teachers should be doing while students are in the MagicSchool rooms, working with custom chatbots.
- Leaders need to think through guidance about when MagicSchool rooms should be used in groups or pairs and when it should be used solo.
Evidence of Impact
Available evidence
No evidence yet
There is no independent, causal evidence of impact on student learning.
Studies of educator use of the platform are available: https://scale.stanford.edu/research-in-action/how-highly-active-k12-educators-are-using-ai-tools-like-magicschool
From the developer
From
MagicSchool
The following was written by the team at MagicSchool and has been edited for clarity and tone by Instruction Partners.
MagicSchool's product development centers on supporting the full instructional loop: how teachers plan, teach, assess, and respond to students. For back to school, we launched a redesigned experience with a new information architecture and personalized tool recommendations that helps teachers move through their workflow more efficiently. Our roadmap continues this focus, with investments in classroom-ready instructional materials grounded in each teacher's curriculum, tools that keep teachers in control of what the AI produces, and system-level configuration so academic teams can set expectations consistently across schools. Student safety remains foundational, with ongoing investment in moderation, visibility for teachers, and safeguards in every student-facing experience.
Product website: https://www.magicschool.ai/
Hear from the developer
Instruction Partners' CEO Emily Freitag spoke with Adeel Khan from MagicSchool AI to understand what problems they're trying to solve, how the tool works, and what they're still figuring out.