AI Tutors Go Mainstream: What Duolingo's Expansion Says About the Next Wave of EdTech

For years, AI in education was mostly a subject for conference talks and pilot programs. In 2026 it is a market with real revenue, real users, and a few clear winners. The story that keeps coming up across business media this week is Duolingo, which is expanding well beyond languages, while Pearson and Khan Academy push their own AI products into schools and universities.

The scale of the shift is easy to underestimate if you only watch the headlines. Venture capital, school districts, publishers, and app stores are all pointing in the same direction at once, and that rarely happens in education, a sector famous for moving slowly. Something structural has changed in how learning products get built and sold.

Why should a business outside education care? Because edtech is currently the clearest demonstration of how AI changes the economics of a digital product: content production, subscription pricing, and user engagement all work differently once AI is involved. Let's look at what is actually happening.

Duolingo turned AI into a production line

Duolingo's expansion is the most visible example of the trend. The company published around 20,500 AI-assisted course units in the first quarter of 2026 alone, a volume of content that would have been unthinkable for its human teams a few years ago. That production speed is what allowed it to launch whole new subjects: Math, Music, and Chess now sit next to the language courses, and Chess reached roughly 7 million daily active users within a year of launch.

The financial side supports the strategy. Duolingo guides for about $1.2 billion in revenue for 2026, after growing 27% year over year in the first quarter to $292 million, with 12.5 million paid subscribers, 56.5 million daily active users, and close to 138 million monthly actives. Management openly calls 2026 an investment year: margins will suffer mid-year because AI features cost real money to run, and the bet is that user growth will pay it back in the second half.

There is also a quieter part of the story: falling inference costs. Features that were reserved for the premium Max tier a year ago are gradually being distributed to a wider audience, simply because serving them became cheaper. The company plans its roadmap around that curve, which is a discipline many businesses adopting AI still lack.

Notably, Duolingo is not squeezing more ads into the free product. Ad load stays flat, and monetization is shifting toward AI experiences, avatars, and other in-app purchases instead. That is a meaningful signal for anyone who builds consumer software.

The money behind the boom

Investors have noticed. AI education startups raised about $4.2 billion in venture capital in 2025, which was around 62% of all edtech funding that year. There are now more than 2,800 AI tutor startups operating, an 18x increase since 2023. Most of the capital is concentrated in two categories: AI tutors for students and copilot tools for teachers.

The teacher side deserves a separate mention. Tools like MagicSchool AI, Brisk Teaching, and Curipod have collectively raised over $90 million, and early 2026 brought a wave of $7-12 million seed rounds for younger players. Teachers, it turns out, adopt AI faster than students do, mostly because it saves them hours of routine work every week.

Market forecasts vary a lot, from a conservative $6.5 billion by 2030 to optimistic projections of over $17 billion by 2033. The honest reading is that nobody knows the ceiling yet, but the direction is not in question.

The adoption problem nobody has solved

Khan Academy's Khanmigo tutor shows the other side of the story. It grew from 40,000 to 700,000 active student users in a single school year, expanded from 45 to more than 380 school districts, and counts around 2 million registered users across nearly 800 districts. Impressive distribution by any standard. And yet, only about 15% of students who have access to Khanmigo use it regularly.

This is the gap that will decide the winners. Getting an AI product in front of users turned out to be the easy part; turning access into a daily habit is the hard one. Khan Academy is responding with a full product redesign rolling out this summer, built around the lessons of that first large-scale school year.

Duolingo, with its streaks, leagues, and push notifications, solved the habit problem long before it added AI, which partly explains why its numbers look so much healthier. The lesson is uncomfortable but useful: engagement mechanics are not decoration on top of an AI product, they are the product.

Pearson bets on AI readiness, not just AI tutoring

Pearson, the largest traditional player in the space, is taking a slightly different route. Alongside AI study tools that its own data says turn passive reading into active practice, the company launched AI Readiness modules aimed at the gap between what universities teach and what employers expect from graduates working with AI. Its virtual learning segment grew 21% as these platforms expanded.

Pearson has also unveiled an AI tool for educators that automates grading-adjacent routine so teachers can spend time on actual teaching. It is the same pattern as in the startup world: the fastest returns on AI in education come from saving professionals time, not from replacing them.

It is a sensible position: whatever happens to individual tutor apps, the demand for people who can actually work with AI tools keeps growing, and someone has to teach and certify that skill.

What this means for your business

Three lessons travel well beyond education. First, AI is most valuable as a production capability: Duolingo did not just add a chatbot, it rebuilt how content gets made, and that is what unlocked new product lines. When AI cuts the cost of producing your core asset, whether that is courses, product descriptions, or support answers, the sensible move is to expand the catalog, not just trim the team.

Second, users will pay for AI features that feel personal, but they will not tolerate a worse free experience. Duolingo's decision to keep ad load flat while selling AI-powered extras is a cleaner monetization path than most companies choose, and its subscriber growth suggests customers reward it.

Third, distribution without engagement is a vanity metric. If only 15% of the people who get your AI feature actually use it, the feature is not done yet, no matter what the press release says.

How to approach AI features in your own product

If you are planning to add AI to a web or mobile product, the edtech experience suggests a fairly concrete sequence. Start with one workflow where AI removes real friction for the user, not the one that demos best. For an education app that was practice exercises; for an e-commerce site it might be search, and for a service business, intake and scheduling.

Then measure habitual use rather than sign-ups. Khanmigo's 15% figure exists because someone was honest enough to measure weekly returning users instead of celebrating the registration count. Set the same bar for your own feature before you scale it.

Finally, treat running costs as a first-class part of the business model from day one. Duolingo schedules feature rollouts around the falling price of AI inference; a small business can do the same on its own scale by launching AI features for paying customers first and widening access as costs drop.

Final notes

AI tutoring has crossed the line from promising to profitable, but the field is still wide open: even the leaders are struggling with engagement, costs, and the question of what students actually need versus what is fun to build. The next couple of years will likely sort the 2,800 startups into a handful of durable companies and a long list of acquired features.

For businesses watching from other industries, edtech in 2026 is a useful preview of their own next few years. The companies that treat AI as infrastructure rather than a marketing checkbox are the ones pulling ahead, and that pattern will not stay confined to education.