AI tutoring is moving away from the chatbot model of instant answers toward a more deliberate approach: AI that guides students through problem-solving while keeping teachers informed about what's happening in real time. This shift reflects emerging evidence that the most effective AI learning tools scaffold instruction, meaning they provide just enough support to keep students thinking without removing the intellectual work that leads to actual learning (Source 1, 2).

What Does the Research Actually Show About AI Tutoring?

Stanford's National Student Support Accelerator and AI Hub for Education recently reviewed the evidence on AI tutoring and found something surprising: the strongest results come from tools that support human tutors rather than replace them. Studies of AI-only tutoring showed substantial engagement problems, with some students barely using the tools at all. A randomized trial involving 6,997 students in Tennessee found that a guard-railed AI tutor improved students' accuracy after mistakes and reduced the attempts needed to return to a correct answer, though students worked more slowly and evidence of retained learning one week later was modest.

The distinction matters because students now have easy access to general-purpose AI tools like ChatGPT that can produce completed solutions instantly. The question educators are grappling with is not whether students will use AI, but how to design AI experiences that actually teach rather than shortcut learning.

How Are Schools Designing AI Tools That Keep Students Thinking?

At Lehigh University, researcher Zilong Pan is developing StemPal, a suite of educational AI agents designed around a core principle: provide the right amount of support at the right time so students can continue thinking and solving problems themselves. The platform includes three specialized tools:

• MathPal: Provides guided problem-solving hints, step-by-step instruction, and opportunities for students to learn mathematical concepts through guided practice rather than receiving completed solutions.
• StatPal: Supports statistical learning through contextual hints and guidance during data analysis, helping students understand concepts rather than just getting answers.
• HackPal: Provides programming hints and debugging assistance intended to help students understand coding concepts rather than simply repairing their code for them.

MathPal, the most extensively studied component, operates as a browser extension compatible with digital math platforms. When students ask questions, they receive explanations, practice problems, strategies, and step-by-step guidance without the solution being revealed immediately. The design is informed by principles of metacognition, which means it encourages students to think about their own problem-solving process, not just the answer.

"I don't want AI to simply give students an answer. The goal is to provide the right amount of support at the right time so students can continue thinking, questioning, and solving the problem themselves," explained Zilong Pan, faculty member in the College of Education's Teaching, Learning and Technology program at Lehigh University.

Zilong Pan, Faculty Member, Teaching, Learning and Technology Program at Lehigh University

Usability testing with 78 high school students who used MathPal in their classrooms for a month revealed that students found it especially useful when working through difficult problems because it broke the process into manageable steps. Students also reported that MathPal could provide support when a teacher was helping another student, introducing them to different ways of organizing and approaching problems. A particularly important finding: students became better at communicating with AI, learning to formulate more precise questions to receive more useful responses.

Why Are Teachers Getting Data Dashboards Alongside Student AI Tools?

The tools are designed with teachers in mind, not just students. StemPal collects information about student interactions and progress and presents processed information on a teacher dashboard, creating opportunities for educators to see where students struggle, what kinds of assistance they seek, and how they approach problems. This reflects a central principle behind the research: AI should augment teachers rather than limit their role in instruction.

When researchers examined 48 ninth-grade Algebra I students who used MathPal over a 14-week semester, the students generated 1,214 conversational threads with the system, providing researchers with a detailed picture of how students seek assistance from an AI learning tool. The study found that students most often approached MathPal for help with solutions or computations, while MathPal most often responded with strategic scaffolding, guiding students through the steps of solving a problem. Researchers also identified four distinct patterns of student-AI interaction, suggesting that students do not all use an AI tutor in the same way.

"A student who repeatedly asks how to take the next step may need a different type of support than a student who asks conceptual questions about why a mathematical principle works. AI learning systems could become more effective by recognizing those differences and adapting the type of scaffolding they provide," noted researchers studying MathPal interaction patterns.

Zilong Pan and collaborators, Lehigh University

Teachers in usability testing recommended features that would allow them to control when students could access MathPal, upload their own instructional materials, and more easily categorize and analyze student interactions. Researchers concluded that teacher-in-the-loop mechanisms are essential for AI tools to remain aligned with classroom goals while allowing educators to intervene when necessary.

How Is This Different From General-Purpose AI Chatbots?

The market is already reflecting this shift. MagicSchool, the leading AI education startup, combines roughly eight million educator signups with major district deployments, student usage, administrative tools, and nearly $63 million in funding. A recent deployment with Hillsborough County Public Schools followed a pilot involving more than 5,000 educators across 233 schools. The district then expanded the partnership to more than 14,000 teachers, principals, coaches, and school leaders serving a system of more than 218,000 students. During the pilot, 94% of participating teachers called the platform useful, and 65% said it saved them at least one hour a week.

Schools are currently buying specialized AI platforms because giving every teacher and student an unrestricted chatbot does not solve the problems a school district has. A district must decide which students can use AI, what teachers can see, which curriculum the AI should follow, how conversations are moderated, what happens when sensitive content appears, and whether student data meets privacy requirements. Those needs have created room for companies whose product includes an AI model but extends well beyond the chatbot itself.

The shift reflects broader changes in how students are already using AI. A nationally representative Gallup study of 2,232 U.S. public-school teachers found that 60% used AI for their work during the 2024-25 school year, including 32% who used it at least weekly. Regular users estimated saving 5.9 hours a week. Among students, 80% of surveyed undergraduates across 15 countries had used generative AI for their studies, according to Chegg's Global Student Survey of 11,706 undergraduates. When students were stuck on a concept or assignment, 29% said they went to generative AI first, ahead of free online resources at 24%, friends or peers at 15%, course materials at 14%, and professors or teaching assistants at 8%. That shift from 10% in Chegg's 2023 survey to 29% in 2026 shows how quickly student behavior has changed.

What Skills Do Students Need to Use AI Effectively?

As AI becomes more embedded in learning, a new educational need is emerging: AI literacy. Knowing how to ask good questions, evaluate the response, and determine what you still need to understand is becoming an important learning skill. Students need to be active participants in that process, not passive recipients of whatever an AI system produces.

"Students are going to use AI. Education has to move beyond asking whether they should use it and begin thinking carefully about how we design AI experiences that help students learn. There is a major difference between an AI system that completes a problem for you and one that helps you develop the skills to complete it yourself," stated Zilong Pan.

Zilong Pan, Faculty Member, Teaching, Learning and Technology Program at Lehigh University

This emerging focus on guided learning and teacher oversight represents a fundamental shift in how the education sector is approaching AI. Rather than viewing AI as a replacement for human instruction, schools are increasingly designing systems where AI and teachers work together, with AI providing targeted support and generating data that helps teachers make better instructional decisions.