Duolingo's stock price has dropped nearly 50% from its all-time high in May, amid concerns that users may prefer learning languages with ChatGPT.
Alex Karp, the CEO of Palantir Technologies, has described large language models (LLMs) as raw material in the context of criticizing Silicon Valley's overhype of AI technology. In a recent speech, he stated:
One of the things Silicon Valley totally f’d up is an overhyping of the value of LLMs. An LLM is a raw material, that has to be processed. The processing of the LLM will change America and the world.
He argues that LLMs are not inherently valuable for businesses on their own and require refinement through platforms like Palantir's to deliver real-world impact, such as integrating with enterprise data and ontologies for secure, actionable AI applications.
How is that relevant to Duolingo?
Easy: Duolingo acts as the ontology for learning. It optimizes the use of LLMs to generate content and deliver lessons, based on data collected from user interactions within its app framework.
While you can definitely use ChatGPT to improve your language skills by practicing—especially if you are a medium or advanced user—it lacks the refinement necessary to become a true language learning tool.
Duolingo has been focusing on improving the user experience and engagement for years now, continuously running A/B tests and adapting the platform hundreds of times per month.
Do you believe that is worth nothing?
Do you really believe a raw LLM can deliver the same value?
One of the things we're going to be experimenting with is having a conversation that is bilingual. So if you're an English speaker learning Spanish, some of it is going to be in English, some of it is going to be in Spanish, to kind of ease you in,
explained Luis von Ahn Arellano, addressing how to increase Max penetration.
Do you believe such an experience can be replicated on ChatGPT?
Maybe you can get close, if you know how to prompt at a good level, you already have a Premium ChatGPT account for other reasons, and you have time to waste. But I guess that is not exactly the typical profile of the person who uses Duolingo to learn.
Now we have a really good metric that we're optimizing for, which is average number of words spoken per Max subscriber. And it's a really good metric because we can move it and because it exactly captures what we want people to do, which is to speak more.
And ever since we started optimizing that, I think our Video Calls just started getting better and better in the sense that they—that now the models are starting to learn that it is better to do things to keep you engaged.
If you believe Duolingo’s focus on continuously improving the app is worthless, think again.
In Q2 2025, the Daily Active Users to Monthly Active User ratio improved again, signaling that engagement kept rising, as it has for years now, signaling that Duolingo’s app is improving quarter after quarter, ever since.
I do not believe that the ChatGPT experience can come anywhere close to that—unless OpenAI decides to develop a vertical focused on teaching languages. But even then, they would lack the years of experience, learnings, and improvements that Duolingo has accumulated. It would be akin to saying “Spotify will be crushed by Apple Music and Amazon Music,” so to speak.
Duolingo delivered another strong quarter, with revenue reaching $252.3 million, up 41% year-over-year. The company raised its full-year bookings guidance to 32% growth at the midpoint, and Free Cash Flow per Share is stately rising.
These results underscore Duolingo's product-led growth strategy. As Luis von Ahn noted in the shareholder letter,
Our growth is fueled by moments of delight that keep learners coming back... We’ve already shipped dozens of rewarding interactions this year, and while they may seem subtle, they compound into a product experience that’s increasingly enjoyable.
If you haven't read my original deep dive on Duolingo, I recommend doing so before reviewing this update. It's essential reading for a thorough understanding of the company.
Is the Declining Sequential MAU Growth a Red Flag?
Q2 2025 marked the first quarter of a clear decline in sequential MAU growth.
If we carefully look at the MAU graph, we notice there is actually seasonality where MAU always drops or slows down from Q1 to Q2, which partially explains the phenomenon.
The fact that this quarter the drop is accentuated is explained by Luis von Ahn Arellano with the hit campaign run in Q1.
The main reason they went down a little bit when compared to previous quarter is, you got to remember, in Q1, we had this crazy campaign of that dead Duo. That kind of slightly inflated MAUs that may not have been DAUs because it got a lot of people that may not have been very committed to just come in.
I believe that, on top of that, simply in Q2 the marketing activities have been less effective.
As the saying goes, sometimes we win, sometimes we learn. As confirmed by the following statement of Luis von Ahn Arellano:
Last time we guided to we were going to say our DAU growth was going to be somewhere between 40% and 45%…
The reason we came towards the lower end [40%] was because I said some stuff about AI…
I didn't give enough context. And because of that, we got, particularly in our social media, we got some backlash on it.
And what that did is that the most important thing is we wanted to make our—the sentiment on our social media became negative. So we wanted to make the sentiment positive. So we stopped posting edgy posts on our social media and we started posting things that we thought would get our sentiment more positive.
And that has actually worked. By now, the sentiment on our social media channels is all very positive. But we still are not posting the extremely edgy things that go—that are more likely to go viral. So that's probably what had us come in slightly the lower end.
During the earnings call, Luis was asked about how we should think about growth and if the market is now mature, and this was his insightful answer:
In the case of maturation, we're not worried about this. I mean if you look at our DAU growth, some of our most penetrated markets are actually the ones that are growing fastest. So we don't see that like, oh, we've reached the level of penetration that allows us—that makes us so that we're going to grow slow everywhere. We don't see that. I mean there are some countries that are growing faster than others, but the penetration is not something we're worried about.
And I just want to remind you, I mean, we have—there are 2 billion people learning a language in the world. We have about 130 million active users, give or take. And so there's a lot of room there. And now we're also adding other subjects. I mean there's hundreds of millions of people that are interested or already playing Chess. The same is true for Math, the same is true for Music. So I just don't think that we're anywhere near our full TAM.
What I believe is that Duolingo’s hyper-growth journey will be a long one, and far from ending, with Daily Active Users still growing 40% YoY, and a lot of growth levers still to kick in and impact the financial numbers, such as the more recently launched courses Music, Math, and Chess.
Our new Chess course has grown faster than any other subject we’ve launched (including language learning!), with over a million DAUs as of quarter end. And that’s just on iOS and in English (we plan to expand Chess to Android and more UI languages throughout this year). We have an exciting product roadmap ahead, including better content for intermediate learners and the ability to engage in matches against others. Chess is still early, but we believe it has the potential to attract new audiences and deepen engagement with our existing learners.
As it becomes a larger slice of the cake, international growth is also going to play an important role.
Luis von Ahn Arellano on China:
We feel really good about China. It's our fastest-growing market. We've been growing a lot. We had a really incredible partnership, this time around with Luckin Coffee, where, for a couple of weeks—Luckin is like their Starbucks. It's essentially everywhere.
Another potential next engagement and retention lever could come from expanding social experiences between users. The app already includes features like the "satisfying haptic when you nudge your friends" to encourage social interaction. Looking ahead, the Chess roadmap includes "the ability to engage in matches against others," which could foster competitive play and community building, similar to how streaks and leaderboards already motivate users.
Conclusion
If anything I can infer from the MAU growth slowdown is that Duolingo may need to increase marketing spending in some specific regions such as the US, where, so far, they grew without much effort but probably now requires some more investment as the “easy reachable users” have been locked in already.
For all the rest, this was an excellent quarter, and ChatGPT concerns are unjustified in my view.
As always, here is the “Deep Dive To Date” (DDTD), that is how the DUOL -0.58%↓ stock is performing since my initial deep dive on the January 11th 2025, when the price was $318.15.
-15% DDTDSee you in the next update!
The content of this analysis is for entertainment and informational purposes only and should not be considered financial or investment advice. Please conduct your own thorough research and due diligence before making any investment decisions and consult with a professional if needed.
Great post.
Interesting