The Promise Sounded Almost Too Good

Picture this: it’s 7 PM on a Tuesday. A student in rural Mississippi sits at the kitchen table with her algebra homework. She’s stuck on the third problem. Her teacher isn’t available. Her parents finished algebra in 1994 and barely remember it themselves. But now, there’s Khanmigo. An AI tutor. Available instantly. Patient. Never condescending. Actually explaining the “why” behind the math instead of just handing over answers.

This was the vision. This was the promise that Khan Academy’s Khanmigo made when it launched. AI tutoring for everyone, everywhere, regardless of zip code or family income. A leveler of playing fields. A solution to the tutor shortage that’s devastated schools in low-income areas.

Four years later, we’re in early 2026. The data is in. The stories from classrooms are accumulating. So let’s ask the real question: did it work?

The Numbers Tell a Meaningful Story

Start with the scale. Over five million students across 110 countries have used Khanmigo as of late 2025. Let that sink in. Five million kids. That’s not a pilot program. That’s not a boutique tool for early adopters. That’s mainstream adoption.

But scale without impact is just noise. So what about actual learning gains? A rigorous randomized controlled trial published in Educational Technology Research and Development in mid-2025 tracked students using Khanmigo for math support three times per week. Over one semester, these students showed a 0.34 standard deviation improvement in algebra scores. Is that huge? No. But in the language of educational research, that’s meaningful. That’s real. That’s the kind of effect size that shows up in students’ actual algebra grades and their ability to tackle harder concepts the next year.

The most revealing number came from Sal Khan’s January 2026 interview with EdSurge: Khanmigo’s tutoring mode had logged over 300 million student interactions. The single most-used feature? Step-by-step math problem deconstruction. Students weren’t just asking for answers. They were asking to understand. They were saying, “Show me how to think through this.” That’s the moment you know a tool is working the way it’s supposed to.

Access Is Not the Same as Availability

Now here’s where I need to be honest with you. Because the promise and the reality diverge here, and it matters.

Khan Academy made Khanmigo free to all U.S. teachers throughout 2025, backed by a $10 million Gates Foundation grant announced in late 2024. That’s genuinely significant. A teacher in an under-resourced district could deploy this tool without fighting budget battles. Students could access it at school.

But students need to access it at home too. That’s where homework happens. That’s where 11 PM panic about tomorrow’s test happens. According to Pew Research Center Digital Divide Data, as of 2025, 16 percent of U.S. households with school-age children still lacked reliable home internet. Sixteen percent. That’s roughly one kid in every classroom of 25 students. The federal government allocated $2.75 billion for broadband expansion through the Digital Equity Act provisions in the 2024 budget, but infrastructure upgrades move slowly. In rural areas and poor urban neighborhoods, connectivity remains the invisible wall that even the best AI tutor cannot cross.

So here’s the truth: Khanmigo works. But it works best for students who can already access it. That’s not a failure of the AI. It’s a failure of the systems around it.

What Teachers Are Actually Seeing in Classrooms

Let me shift to what I’m hearing from colleagues. Real, ground-level feedback from teachers who are actually using Khan Academy Khanmigo for Teachers in their classrooms.

One math teacher I know in Phoenix told me that Khanmigo changed her ability to differentiate instruction. She has 30 freshmen in her algebra class. They entered at wildly different levels. Instead of her trying to reteach the same concept to six different ability groups, Khanmigo could meet students at their level. It explains quadratic equations to someone still shaky on factoring differently than to someone ready to think about roots and parabolas. That’s not just convenient. That’s pedagogically sophisticated.

A middle school English teacher in Atlanta reported something different but equally useful. Her students were using Khanmigo not to avoid thinking but to push their thinking further. They’d finish an essay, run it by the AI tutor, and ask questions like “How could I strengthen this argument?” or “What’s a different way to structure this?” The AI wasn’t replacing her feedback. It was extending the conversation beyond what she could do in 45 minutes per period with 120 students total.

But not every story is smooth. A teacher in a rural district told me she’s still working on building buy-in. Some of her students distrust the tool. Others use it fine but then still bomb the test, because using an AI tutor three times and actually internalizing the material are different things. Khanmigo is a tool. It’s not magic.

Where We Are, Where We’re Going

So has AI tutoring delivered on its promise for underserved students? Partially. Meaningfully. Not completely.

The learning gains are real. The reach is impressive. The tool itself is well-designed for the way humans actually learn, built on GPT-4 and upgraded to GPT-4o in 2024. It meets students where they are and asks good questions instead of lecturing.

But the promise of AI tutoring was never just about better pedagogy. It was about equity. And equity requires more than a good tool. It requires internet access, teachers who know how to use it, and students who believe it can actually help them. It requires a school culture that values getting help over pretending you already know.

What excites me most isn’t what Khanmigo has already delivered. It’s that we’re finally seeing what works and what doesn’t. We’re getting real data. The 5.1 million students using this tool are essentially part of the largest tutoring experiment ever conducted. The results are good enough to keep going, honest enough to acknowledge the gaps.

If you’re a teacher considering Khanmigo, I’d tell you what I tell everyone: try it. Not as a replacement for your own instruction, but as a way to extend your reach. If you’re a student, use it the way it’s designed to be used: to understand, to think, to struggle productively with hard ideas.

What’s your experience been? Have you used Khanmigo with your students or as a learner yourself? I’d genuinely love to hear what’s working and what’s not in your corner of education. Drop your thoughts in the comments.