Artificial intelligence has been heralded as the solution to one of education's toughest problems: making math relevant and engaging for students who see the subject as disconnected from their lives. Yet after years of investment and experimentation, a troubling gap has emerged between the promise of AI-powered personalization and what actually works in classrooms. The core issue isn't that AI can't personalize lessons, but that it often personalizes them in ways that undermine real learning.

Why Does AI Struggle to Create Realistic Math Problems?

When teachers use large language models like ChatGPT or Gemini to create math assignments tailored to student interests, the results can be surprisingly disconnected from reality. A student interested in concerts might receive a homework problem asking them to calculate how many "pins" audience members are wearing at different points during a performance, a metric nobody would ever actually track. Another student interested in music might be asked about a concert where sound levels reached 400 decibels, a physical impossibility.

The problem runs deeper than silly examples. AI excels at knowing about interest areas, whether that's K-Pop, Minecraft, or TikTok influencers. But it struggles with something far more important: connecting those interests to how math is actually used in professional and real-world contexts.

"AI is very good at knowing about interest areas. It knows all about Vampire Diaries or K-Pop Demon Hunters. It's great at that, but it's not very good at connecting a topic to an academic area in a meaningful way," explained Candace Walkington, a professor in the teaching and learning department at Southern Methodist University in Dallas.

Candace Walkington, Professor of Teaching and Learning, Southern Methodist University

Walkington, who received a grant from the National Science Foundation to explore personalization in math, has been working on solutions, including a specialized "realism bot" designed to review AI-generated math problems before they reach students. But even with these safeguards, the challenge persists: creating assignments that genuinely incorporate student interests while asking them to solve problems that would actually occur in the real world.

What Does Effective AI-Assisted Math Teaching Actually Look Like?

The most successful examples come from experienced teachers who use AI as a tool to enhance their existing craft, not replace their judgment. Al Rabanera, a teacher at La Vista High School in Fullerton, California, demonstrates this approach. When teaching rate of change, a statistical concept, Rabanera used an AI tool to help him create an assignment connecting the topic to something his students genuinely cared about: the job market.

Rabanera prompted the AI to generate U.S. Department of Labor data showing correlations between education level, gender, and median weekly income. Then he worked with the AI to brainstorm questions that would help students dig into the numbers using the statistical strategies they were learning. When students analyzed the data and discovered the gender pay gap, the lesson became visceral and meaningful. One female student immediately grasped the real-world implications: "Whoa, Mr. Rab! I'm gonna get paid less 'cause I'm a girl?"

"When I design these types of lessons without AI, it takes me hours. I've done it before, looking at all the trends and themes, grouping and coding them, and then sharing it back with the kids," said Al Rabanera, noting that AI has offered him a tool to "deepen and refine" the kind of lesson he always strives to deliver.

Al Rabanera, Math Teacher, La Vista High School

This approach works because Rabanera brings domain expertise and pedagogical judgment to the process. He knows what math concepts matter, how they connect to the real world, and what will resonate with his students. AI accelerates the research and brainstorming, but the teacher remains the architect of the learning experience.

How to Create Realistic AI-Personalized Math Lessons

• Start with Real-World Context First: Before asking AI to personalize a lesson, identify a genuine real-world application of the math concept you're teaching. This anchors the personalization in authenticity rather than letting AI chase student interests into unrealistic territory.
• Use AI for Research and Brainstorming, Not Problem Generation: Leverage AI to gather data, identify trends, and brainstorm question ideas, but review and refine the actual problems yourself to ensure they make mathematical and practical sense.
• Validate Problems Against Reality: Before using an AI-generated math problem with students, ask yourself: Would anyone actually need to solve this problem in the real world? If the answer is no, revise or discard it.
• Connect to Student Values, Not Just Interests: The most powerful personalization connects math to what students care about achieving or understanding, not just what they like to watch or listen to.

Why Schools Are Pulling Back on AI Personalization Features

The challenges with AI-generated personalization have become apparent even to early adopters. Khan Academy, an organization that pioneered the integration of generative AI into student learning through its Khanmigo chatbot, made a significant decision: it removed a feature designed to incorporate students' personal interests into tutoring sessions. The reason was sobering. The company wasn't seeing clear benefits in either students' academic progress or engagement.

This retreat signals an important lesson for the broader EdTech sector. Personalization that sounds good in theory, and that AI can technically deliver, doesn't automatically translate into better learning outcomes. The gap between what technology can do and what actually helps students learn remains stubbornly wide.

More than half of teachers, 55 percent, cite poor student engagement in academics, including math, as a significant challenge, according to an EdWeek Research Center survey of 729 educators conducted from January 28 to March 5. And more than a third, 36 percent, report that students are less engaged in math than in other subjects.

The research is clear on what works: linking math concepts to students' interests and showing how those concepts apply in the real world can make an abstract subject come alive. But AI, left to its own devices, often creates the illusion of personalization without the substance. The technology knows what students like, but it doesn't always know how to teach them mathematics in a way that sticks.

As schools continue experimenting with AI in the classroom, the lesson from math education is becoming clearer: the most valuable role for AI isn't replacing teacher judgment, but augmenting it. Teachers who understand their subject matter, their students, and the real world can use AI as a powerful research and brainstorming partner. But without that human expertise guiding the process, AI personalization risks becoming an elaborate way to trick students into learning something that doesn't actually matter.