Duolingo beat every Wall Street estimate for the first quarter of 2026. Revenue rose 27 per cent year on year to 292 million dollars. Earnings per share came in at 89 cents against expectations of 76 cents. Daily active users grew 21 per cent to 56.5 million. Paid subscribers grew 21 per cent to 12.5 million. Adjusted EBITDA margin reached 29 per cent. Then CEO Luis von Ahn told investors that the company was going to slow down on purpose. Duolingo would prioritise user engagement and long-term growth over near-term monetization. Bookings growth would decelerate to six per cent in the second quarter. Full-year guidance called for 10 to 12 per cent bookings growth, 15 to 18 per cent revenue growth, and a 25 per cent EBITDA margin, all below the trajectory the stock price had priced in. The shares fell 14 per cent. The stock is now down more than 40 per cent in 2026 and roughly 80 per cent from its all-time high. Von Ahn is not concerned. He is making a bet that in the age of AI, the only thing that will keep Duolingo alive is the thing that made it successful in the first place: the daily habit.

The strategy

Von Ahn’s letter to shareholders laid out the logic with unusual directness. Duolingo’s goal is to reach 100 million daily active users by 2028, nearly doubling its current base. To get there, the company is investing in product improvements that increase engagement, even when those improvements reduce short-term revenue. The most significant change is expanding access to features that were previously locked behind the paid subscription. Longer free trials, free access to the Explain My Answer feature that uses GPT-4, and a redesigned progression system are all designed to keep users coming back daily. The calculation is that users who build a daily habit are more likely to convert to paid subscribers eventually, and that the lifetime value of a habitual user exceeds the subscription revenue lost by giving away features in the short term.

The product roadmap supports the thesis. Duolingo’s AI-powered language tutors were among the first consumer chatbot implementations that actually worked, and the company has continued to invest in AI features that make the learning experience more conversational and less mechanical. Video Call, a feature that lets users practise speaking with an AI character called Lily who responds in real time, is being expanded beyond the Max subscription tier. Speaking Adventures, a new feature that places users in simulated real-world scenarios, is designed to address the gap between app-based language learning and actual conversation. Spoken Tokens, a currency earned by completing speaking exercises, adds a gamification layer to the part of language learning that users find most intimidating. Every feature is built to increase time spent in the app, which increases habit strength, which increases the probability that a free user eventually pays.

The threat

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The urgency behind the strategy is not financial. It is competitive. ChatGPT can already conduct a fluent conversation in more than 50 languages, correct grammar in real time, and adapt its vocabulary to the learner’s level. Google’s Gemini models can do the same, integrated into the world’s most widely used search engine and mobile operating system. Google has experimented with dedicated language learning features built on its translation and AI infrastructure, and the combination of a free product, a massive distribution channel, and increasingly capable language models makes Google the most dangerous potential competitor Duolingo has ever faced. Meta’s Llama models, available open-source, allow any developer to build a language tutoring application with conversational AI at near-zero marginal cost. The technical moat that Duolingo built over a decade, a structured curriculum delivered through an addictive app, is being commodified by AI models that can generate equivalent instruction on demand.

Von Ahn has acknowledged this openly. In the earnings call, he described language learning as one of the first consumer categories where AI would fundamentally change the competitive landscape. Google’s AI agent strategy positions conversational AI as a platform-level capability available across every Google product, from Search to Android to Workspace. If language practice becomes a feature of the operating system rather than a standalone app, Duolingo’s value proposition narrows to the things AI cannot easily replicate: the streak counter, the leaderboard, the push notification that arrives at 7 p.m. reminding you that your 847-day streak is about to break. The gamification is not a gimmick. It is the product. And von Ahn is betting that it is more defensible than the curriculum.

The numbers

The disconnect between Duolingo’s operating performance and its stock price reflects a market that is not sure whether the engagement bet will pay off. Revenue growth of 27 per cent is strong by any measure. Subscription revenue of 251 million dollars, up 31 per cent, demonstrates that the paid model works. The 29 per cent EBITDA margin shows operational leverage improving as the user base scales. But the forward guidance, six per cent bookings growth in Q2 and 10 to 12 per cent for the full year, represents a deliberate deceleration that investors are not accustomed to from a company that has been one of the market’s highest-growth consumer technology stocks.

The institutional revenue story adds a complication. Duolingo’s English Test, which allows users to take a certified language proficiency exam through the app, has become a meaningful revenue stream and a competitive alternative to TOEFL and IELTS. But the test business operates on different economics from the consumer subscription: it depends on institutional acceptance, which grows slowly, and on immigration and education policy, which changes unpredictably. The consumer subscription business, which accounts for the vast majority of revenue, is the segment that von Ahn is reshaping. The question investors are pricing is whether the engagement-first strategy produces 100 million daily active users by 2028, or whether it produces a larger free user base that is harder to monetise in an environment where AI has made the core product less differentiated.

The moat

The pattern of legacy software products adding AI features to defend their market position is now visible across the technology industry. Slack rebuilt its Slackbot around AI to compete with Microsoft’s Copilot integration. Adobe added generative AI to every creative tool. Notion, Canva, and dozens of other productivity applications have embedded AI assistants that perform tasks their products were originally designed for humans to do manually. In each case, the company’s defence is not the AI itself, which competitors can replicate, but the user base, the workflow integration, and the switching costs that keep customers on the platform. Duolingo’s version of that defence is the habit loop: the streaks, the experience points, the league tables, and the social pressure that make opening the app feel like a daily obligation rather than a choice.

OpenAI has released open-source safety tools specifically designed for applications used by teenagers, acknowledging that AI consumer products face unique regulatory and safety requirements when their user base skews young. Duolingo’s user base includes a significant proportion of students and young learners, and the company’s ability to deploy AI features responsibly while maintaining engagement is both a competitive advantage and a regulatory exposure. The AI features that make Duolingo’s product more compelling, conversational AI tutors, real-time speech recognition, adaptive difficulty, are the same features that attract regulatory scrutiny when deployed at scale to younger users.

The bet

What von Ahn is doing is unusual in public markets: telling investors that the company will be less profitable in the short term so that it can be larger in the long term, and asking them to trust that the engagement metrics, not the financial metrics, are the ones that matter. The bet is that daily active users are the leading indicator and that revenue is the lagging one. If Duolingo reaches 100 million DAUs, the monetization will follow because 100 million people who open an app every day represent an audience that advertisers, content partners, and premium subscribers will pay to reach. If it does not, the company will have sacrificed years of revenue growth for a user base that AI competitors can serve for free.

The stock market’s 14 per cent verdict on Sunday night was not a rejection of the strategy. It was an acknowledgement that the strategy carries real risk. Duolingo is trading at roughly 11 times forward revenue, down from more than 30 times a year ago, a compression that reflects the market’s uncertainty about whether engagement-first growth can produce the financial returns that monetization-first growth delivered. Von Ahn has staked his company on the proposition that in a world where AI can teach anyone anything, the only competitive advantage is making people want to come back tomorrow. The owl is the moat. The 14 per cent drop is the price of proving it.