The type of world that I would like to have is a world where people really are bettering themselves, because that is one way in which they find meaning in life. And what's at stake for us is, will it be the case that in 10 years we can get a significant fraction of humanity's population, billions of people, actually learning something, as opposed to either doom scrolling or just giving up on life because AI can do everything. We got to more than 100 million active users, but that's too small of a fraction of humanity.
That's Luis Von Ahn, CEO and co-founder of Duolingo, the language learning app that crossed a billion dollars in annual revenue last year. Duolingo is known on social media for its slightly unhinged owl mascot, but it was Luis himself who went viral last year over an internal AI memo that sparked intense public controversy. Today we talk through his tough lessons from that experience and what he's learned about the limitations of AI since that memo. We also talk about why he's made a strategic shift in 2026, emphasizing user growth over revenue growth even though 90% of Duolingo's users are currently non-paying. Plus, in the spirit of Duo the Owl, we talk about fun as a learning strategy, whether you're teaching languages, chess, or new business ideas. So, let's get to it. I'm Bob Safian, and this is Rapid Response.
I'm Bob Safian. I'm here with Luis Von Ahn, CEO and co-founder of Duolingo. Luis, great to have you with us.
I'm very happy to be here. Thank you for having me.
You first hit my radar when I was editor of Fast Company. You helped invent CAPTCHA and then reCAPTCHA, the security word puzzles that became kind of ubiquitous across the web. And then you launched Duolingo, which is this sort of playful but pragmatic language learning system that really took off and now it's crossed a billion dollars in annual revenue last year. Duolingo has always been unexpected. Certainly your social media campaigns have gotten a lot of attention. How important is the unexpected and playful to what Duolingo does?
I mean what we want to do is we want to teach people and we want to literally teach billions of people. That's what we want to do. Now learning something is an interesting thing. Most people will tell you they want to learn something. But what people say and what people do happens to be very different. I mean certainly trying to be a fun brand, an unhinged brand, in as much as possible because that gets people to learn. And so our belief about teaching is that it's really important that we teach you something well, but it's equally or more important that we keep you entertained while you're learning because otherwise most people just won't learn all that much.
And I mean a lot of your users are adults, so it's not just kids that you're trying to get to be playful. And it's just important for those of us who are trying to be lifelong learners.
Yeah, a lot of our users are adults. I mean, basically our users are the whole range from age, they need to know how to read. So, you know, age, call it six, seven to like 100. When we started Duolingo, there were people that would say, 'No, no, no. Serious learners are not going to want this because it has to be serious, etc.' We have just never found that making something less fun helps. It just helps with everybody. Whenever we make it more fun, even the people who call themselves serious learners end up doing more.
And the decisions you make about Duolingo, like is there ever a distinction between engagement and learning? I mean, you make it sound like it goes together. There are, you know, some parts of the tech world where there's sort of been blowback about emphasizing engagement too much.
They mostly go hand in hand because the most important thing is that people spend the time learning. If you want to learn Spanish and you're an English speaker, you kind of have to put in about 500 hours. Like that's basically it. It's kind of similar with exercising. I mean whether you're doing the elliptical or the treadmill, it probably matters a little bit, but what matters the most is that you're actually doing it a certain amount of time every day. With Duolingo, when people are learning on Duolingo, the setting is a little different than a classroom. In a classroom, I think you can frustrate quote unquote users, in this case, the learner a little more because they kind of have to be in the classroom.
It's not so easy for them to get up and walk away.
Yeah. Whereas in Duolingo, we are always one click away from Instagram or TikTok or whatever you want. So the level of frustration we can give to users has to be lower than what a classroom does and that probably makes it so that we probably are teaching a little less but we just want to have a lot more time and so in the end it actually ends up being good for learning.
I mean right now a lot of the learning that business people are being forced to do is technological, is about AI. I know that that's not been the focus of what the learning is on Duolingo, but are there things about the way we should be approaching learning about this new technology that you would take away from what you do with Duolingo? It certainly isn't often framed to us as being fun.
I think the most important thing I would say for learning anything, it doesn't have to be fun. It just has to be that it keeps people motivated. There are multiple ways to keep people motivated. With Duolingo, we've chosen mainly fun. That's the main thing we've chosen, but you don't have to do that. I mean, for example, seeing results keeps people motivated. So, in the case of learning AI, I would say that's probably a better motivator of the form like I'm going to learn AI, but the first thing that I'm going to do is make myself a dashboard or like a mini dashboard or something. But I think it just if you find the right motivation, that helps a lot.
I mean I have to ask you because last year at this time you sent out this all hands email about AI that rattled things a bit, right? No new hires unless teams showed that AI couldn't do the job and existing employees assessed on their AI use. And it really sparked this blowback on social and the stock price. You're not unfamiliar with taking risks. Was this a bigger risk than you realized at the time?
Absolutely. I did not think this was going to be controversial because internally inside Duolingo this was not controversial. We started as a technology company and I used to be a computer science professor that actually taught the AI class at Carnegie Mellon University. So we've always used AI in as much as we can. Internally this was not controversial at all. Externally, I think I was not very clear and without giving it more context. I opened it up to people thinking that what I was trying to do was to fire a lot of our employees. But that was never the intention. I mean, we've never done a layoff. We still have never done a layoff. In fact, last year when I sent that memo, we increased our number of employees, not decreased our number of employees. So I think there was that misunderstanding because there's a lot of fear that AI is going to substitute jobs etc. But the way I see it here is our employees are just way more productive if they use AI.
And so you actually want to hire more people because they can do more. It also seemed a little bit like you were sort of forcing people. You weren't making it playful to learn how to use AI. You were sort of like it was almost like a bludgeon, right? Which I guess wasn't the intent necessarily either, but I do think it's something a lot of folks are struggling with, like how do you get the people on your team who are more resistant to this new technology to get on board?
Yeah. I mean the good news here is at Duolingo we just hire a lot of people who are pretty young. I mean we hire a lot of people who are straight out of college and we have just not found a lot of resistance here. Internally we have this thing we call the golden rule of AI usage which is you have to use AI for the benefit of our learners. Like everything we do with AI should be for the benefit of our learners. So, for example, if it helps you be more efficient at putting out more features, that's for the benefit of our learners. If we put out a feature that helps our learners learn better because they're now, for example, interacting with an AI to practice conversation, that's for the benefit of our learners. Sometimes when we use AI, we're able to save costs, but that is not the goal of our usage of AI. It has to be about helping our learners versus we're going to use AI to save 10 million bucks. That's just not all that motivating.
And I just want to make sure everyone, because there have been some reports about you kind of walking back some of the things you said in that memo. You're clearly still a believer in AI. There's no doubt from you that this is the direction the business has to continue to go.
I'm a big believer in AI but that comes with some, you know, it definitely comes with asterisks and learnings. I mean for example my initial memo said that we were going to evaluate every employee on their usage of AI. I don't think that was right. Many people came to me and said look for the job that I'm doing I'm finding that I'm just using AI for AI's sake because you're going to evaluate me on that but it's not because I actually think that for this particular thing we can do it better. And I think ultimately in the case of performance reviews what you should evaluate is how much that particular employee is contributing to the company. It turns out for most of our employees using AI helps them contribute more to the company. That is true. However there may be cases, projects or particular roles where it may just not help all that much. And so saying blanket statement we're going to evaluate you on the usage of AI was not needed. And so we've removed that. So that's kind of one important thing. The other important thing that I think is important to mention when it comes to AI is that we're trying to use AI as much as possible, but we really don't want to decrease quality. For some things, AI is quite ready to do high-quality stuff. For some things, it's just not.
And so we're not going to decrease quality just for the sake of using AI. So where are you seeing AI not able to deliver that kind of quality?
In a lot of places. For example, we hire a lot of artists and designers. And our app is very, you know, high craft on design, etc. We're just not seeing AI get to the level of creativity or the level of polish that our top people have by any means. The other place where I think it's just not the highest quality, one of the biggest problems I think with AI is that it demos really well. So what do I mean by that? It's just like look, it can write a story and if you see one story it probably wrote a really good story like the one story they showed you and my god it wrote a story. But when you go and in our case we may need to write 1,000 different stories for people to learn a language. Then you'll find that like I don't know 20% of the things were just pure slop. Whenever we scale a lot with things with AI, we have to really be careful that slop doesn't get through. So, we do a lot of things to try to do that. And if the quality is just not high enough, even though AI is really nice in that it can do it pretty fast, we just don't go for it.
Yeah. I mean, I've heard some of that with coding, too. Like, it can be terrific in writing code and then it can produce stuff that like, well, that just doesn't work. And you need to dig into that to test all of those uses.
Yeah. For the longest of times, if you would read Twitter, the vibe there was, you probably should just fire all your engineers because AI can code way better than any human. Forget about engineers. That is just not true. Internally, you would see also that your engineers were all using AI, but somehow you hadn't seen a speed up in engineering progress, like the number of features we put out every month or whatever was just not increasing. And when you went to see why that was, you would see that, oh, it turns out in some cases it does something amazing. And in the other cases, just like you said, it kind of doesn't work, but even worse, it's so hard to figure out why it didn't work that you spend about as much time as you saved on the cases that it worked figuring out why the other thing didn't work. So, this was a little bit of a paradox that was going on. It's getting a little better because coding agents are getting better. Yeah, you can find a project with one engineer that maybe was 10 times faster, but overall this is just not yet true.
In 2026, you've sort of signaled a strategic shift for Duolingo, prioritizing user growth over profitability, which seems like another risk, especially for investors. So, why make this move and why now?
Yeah, at the beginning of the year we decided, I decided that this year we would really spend it trying to work much more on user growth rather than revenue growth. And the reason for that is there's a number of reasons, but probably the biggest one is I really think that over the next few years, the way most people learn is going to change. I mean, AI is just going to allow us to teach significantly better than we have in the past. And we're already seeing it. Our product today in terms of teaching is way better than it was four years ago. We do studies to see how well our product is teaching. It's just getting better every year. So I think there's a huge opportunity in front of us and because of that whenever you see that there's a huge opportunity in front of you I think what you want to do is just capture as much of the market as possible. The other thing that happened was that our user growth particularly towards the end of 2025 did slow. It doesn't mean that we're not growing. We're still growing. We're still growing exponentially but it did slow. And I sat here and I thought, wow, at a time when we should be actually accelerating user growth, it seems to be slightly decelerating, I really don't like that. So, we're going to make that change. And that's what happened. Now, of course, whenever you go out to the stock market and you say this year we are going to lower our estimates for revenue, the stock market, no likey.
No. But I mean we knew this was going to happen.
We knew but I really do believe that this is the right thing to do for the long term of the business.
And just so I'm sure that I and our listeners understand your business model. Most of your users are free users already supported by ads that come after a lesson. Most of your revenue is from premium users or premium fees. And so user growth means more emphasis on the free level.
Yeah, that's the main thing that helps us grow users. So about 90% of our monthly active users are not paying us and about 10% pay us. But the amount of money we make is flipped. So about 90% of the amount of money we make comes from the 10% that pay us. So most of the money we make comes from subscriptions. Increasing the fraction of people who pay is actually not that hard for us. And we can do it at any point in time. If at the end of a lesson I give you one ad versus if I give you two ads, that makes a huge difference in the number of people that pay.
