The AI Tutor That Actually Gets Paid
Duolingo built an empire on streaks, cartoon owls, and the psychology of daily habit loops. It has hundreds of millions of registered users and a stock market listing that made it a household name in edtech. But registered users and paying users are two very different things, and a growing number of learners willing to spend real money on language acquisition are starting to look elsewhere – specifically at Speak, an AI-powered conversation app that has been building a quiet, loyal paying base while Duolingo fights to convert its massive free audience into subscribers.
Speak is not trying to out-Duolingo Duolingo. It is not chasing gamification metrics or daily active user counts. Instead, it is targeting a narrower, more valuable audience: adults who need to actually speak a foreign language, not just recognize vocabulary on a phone screen. That focus is beginning to translate into real revenue traction, and the competitive pressure it is creating on Duolingo’s subscriber base is the kind that is easy to miss in quarterly earnings calls but hard to ignore once you track where motivated learners are spending their money.
What Speak Is Actually Selling
Speak launched with a single core premise: most language apps teach you to read and recognize words, but almost none of them train you to hold a real conversation. The app uses AI to simulate spoken dialogue, giving users a low-stakes environment to practice speaking out loud, get corrected in real time, and build the kind of fluency that only comes from actually opening your mouth. The product is particularly strong in Korean and English markets, where it gained early traction before expanding its language catalog.
The subscription price reflects the positioning. Speak charges more than Duolingo’s standard tier, and it makes no apology for that. The bet is that a learner preparing for a job interview, a relocation, or a professional certification is not price-sensitive in the same way a casual user playing vocabulary games during a commute is. That segment – call it the outcome-motivated learner – is smaller than Duolingo’s total addressable market, but it converts at a higher rate and churns at a lower one.
Speak’s AI conversation engine is built on large language model infrastructure that lets it adapt to a learner’s specific weak points in real time, something that static lesson decks cannot do. The experience of being gently corrected mid-sentence, then immediately trying again, is closer to working with a private tutor than completing a quiz. That is a meaningfully different product, and it is why users who try it tend to stick with it in a way that Duolingo’s data on long-term retention does not fully capture.
Duolingo’s Paying-User Problem
Duolingo’s business model depends on a conversion funnel: pull in free users with gamification, then convince a small percentage to upgrade to Duolingo Plus for an ad-free experience and some additional features. The strategy has worked well enough to sustain a public company, but it creates a structural vulnerability. The users most likely to pay for a language app are also the users most likely to have a specific, serious learning goal – and those are exactly the users Speak is designed for.
When a motivated adult learner compares Duolingo’s paid tier against Speak’s conversation-first approach, the gap in product philosophy becomes obvious fast. Duolingo Plus removes ads and adds a few extras, but the core experience is still built around gamified micro-lessons. Speak’s paid tier is the entire product. For someone who genuinely needs to speak another language, the value proposition is not even close.
The Market Duolingo Cannot Easily Defend
Duolingo’s scale is also its constraint. Building a product for hundreds of millions of users across dozens of languages and learning levels means the core experience has to remain broadly accessible and broadly entertaining. That keeps the app sticky for casual learners but makes it structurally difficult to go deep on any one use case. A user who wants to practice business Korean or prepare for a TOPIK exam is not well-served by the same lesson format designed to teach a tourist a few airport phrases.
Speak, by contrast, is making deliberate decisions about which users it wants to win. Its early concentration on Korean learners was not accidental – it was a strategic choice to dominate a specific high-intent market before expanding. The same logic applies to its focus on spoken output over reading and writing comprehension. Every product constraint is a signal about whose problem the app is actually solving.
The financial implications of this dynamic are worth sitting with. A Duolingo Plus subscriber paying for ad removal and heart refills represents a different kind of customer than someone paying for Speak’s AI conversation sessions. The latter user has a specific outcome in mind, is tracking their own progress against a real-world goal, and is far less likely to cancel when a new mobile game distracts them for a week. That difference in user quality – not just quantity – is what makes Speak’s subscriber growth a genuine competitive threat rather than a niche curiosity.
Duolingo has the resources to build something like Speak’s conversation feature, and it has been investing in AI across its product. But product roadmaps move slowly when you are protecting a core experience that hundreds of millions of users have memorized. Any significant departure from the streak-and-lesson format risks alienating the base that makes the company’s engagement numbers look good on a quarterly slide. Speak does not have that problem. It can go further, faster on AI conversation because it was never trying to be a habit loop app in the first place.
The real question is not whether Duolingo will eventually add better speaking practice – it almost certainly will. The question is whether adding a feature catches up to being built around a philosophy. Right now, Speak’s paying users are betting it does not.
