Three years after launching Khanmigo, an AI-powered tutoring chatbot designed to revolutionize education, Khan Academy founder Sal Khan acknowledged a sobering reality: the revolution hasn't materialized. Many students didn't engage with the tool, and those who did often found it frustrating rather than helpful. The experience reveals a fundamental gap between AI's theoretical potential and its real-world adoption in classrooms .

In summer 2022, OpenAI leaders Sam Altman and Greg Brockman approached Sal Khan with early access to GPT-4, the advanced language model that would power ChatGPT. Khan Academy built Khanmigo as a specialized chatbot designed to guide students through problems without simply handing them answers. Khan became an evangelist for the technology, predicting in a widely viewed 2023 TED Talk that AI could give every student on the planet "an artificially intelligent but amazing personal tutor" .

He even cited a seminal 1984 study suggesting that individualized tutoring could turn average students into academic standouts. The vision was compelling, and Khan Academy appeared on "60 Minutes" showcasing early adoption at Hobart High School in northwest Indiana. But behind the scenes, something unexpected was happening: students weren't using the tool as intended .

What Happened When Schools Actually Used Khanmigo?

Kristen Musall, a geometry teacher at Hobart High, gave Khanmigo a try when it first launched. She appreciated its encouraging, teacher-like tone, but quickly noticed a problem. Students found the chatbot frustrating because it wouldn't give them direct answers, and when it did make mistakes, the experience felt unhelpful. "If students don't engage with the material enough to know what they're looking for, then an AI like Khanmigo doesn't necessarily help," Musall explained .

"For a lot of students, it was a non-event. They just didn't use it much," said Sal Khan, founder of Khan Academy.

Sal Khan, Founder, Khan Academy

Musall no longer uses Khanmigo in her classroom. She observed that enthusiasm for the tool came primarily from administrators rather than teachers. A few advanced students did use AI to learn new topics independently, but more students were using it to find answers, creating what Musall described as "a massive headache for teachers." Nationally, a majority of teenagers report that AI-powered cheating is at least somewhat prevalent in their schools, according to a recent Pew survey .

She

Peggy Buffington, Hobart's superintendent, took a more optimistic view. She acknowledged there was a learning curve for students to ask Khanmigo questions effectively, but said they've improved over time. She emphasized that schools need to prepare students to use AI responsibly, and she views Khanmigo as preferable to commercial products students might use on their own. However, even Buffington acknowledged that many students won't take advantage of the tool or don't know how to use it .

Why Didn't Students Ask the AI for Help?

The core issue, according to Khan Academy's chief learning officer Kristen DiCerbo, is that AI can only respond to what students ask. "Students aren't great at asking questions well," DiCerbo noted. She was initially hopeful that AI would personalize instruction to students' needs and interests, but that hasn't happened. "So far I am not seeing the revolution in education," she stated .

"Students aren't great at asking questions well," explained Kristen DiCerbo, Chief Learning Officer at Khan Academy.

Kristen DiCerbo, Chief Learning Officer, Khan Academy

Khan himself offered an apt analogy for why the technology underperformed. Imagine he walked into a classroom, sat in the back, and waited for students to seek help. "Some will; most won't," he said. That's been the experience with AI tutoring. The technology doesn't necessarily motivate students to learn or help them fill knowledge gaps needed to ask meaningful questions in the first place .

This limitation reveals a deeper truth about AI in education: the technology works best when students already know what they're looking for. For struggling learners or those without foundational knowledge, an AI tutor sitting passively in the background offers little benefit. The tool requires student initiative that many learners simply don't have .

How Khan Academy Is Rethinking Its AI Strategy

Rather than abandoning AI entirely, Khan Academy has restructured its approach based on what it learned from Khanmigo's limited adoption. The organization recently announced an overhaul that pairs AI with other offerings. A spokesperson explained that the change was made because "students were not seeking out Khanmigo's help as much as we had hoped." Now Khanmigo is integrated directly into the learning experience, available as students work through specific problems rather than as a separate tool they must seek out .

A recent study found that when teachers used Khan Academy to help students practice academic content, their classes made slightly faster learning gains. However, lower-performing students saw few if any improvements from Khan Academy, even before Khanmigo was introduced. This suggests that AI tutoring, like many educational technologies before it, may benefit some students while leaving others behind .

Steps to Integrate AI Effectively in Classrooms

• Embed AI Within Existing Workflows: Rather than asking students to seek out AI help separately, integrate it directly into the learning process so students encounter it while working on specific problems.
• Pair AI With Human Instruction: Khan emphasized that "our biggest lever is really investing in the human systems." AI works best as a complement to teacher-led instruction, not as a replacement.
• Teach Students How to Ask Questions: Since AI can only respond to what students ask, schools need to explicitly teach students how to formulate questions and seek help effectively.
• Monitor for Misuse: Schools must establish clear guidelines about appropriate AI use to prevent students from using tools primarily to find answers rather than to learn.

What Does This Mean for AI in Education?

Khan's acknowledgment that AI hasn't delivered the promised revolution is significant. It's an early indication of the limits of AI to drive massive learning gains, a goal that has eluded various educational technologies for decades. While Khan remains optimistic about AI's potential, he's also become more realistic about its role. "I just view it as part of the solution; I don't view it as the end-all and be-all," he said .

The evidence base for AI in education remains "extremely limited," according to an overview paper released recently. AI is still poised to shake up American education in multiple ways, from making cheating easier to reshaping how teachers approach their work. But the path forward requires more than just deploying powerful technology; it requires rethinking how that technology integrates with human teaching and student motivation .

For educators and policymakers watching the AI revolution unfold, Khan's experience offers a crucial lesson: transformative technology requires transformative thinking about how students learn, not just how machines can teach.