Duolingo is not a free language learning app.

It’s a paid language learning game. An incredibly well designed game created to drive engagement for over 500 million registered users while icing the experience in an illusory layer of white chocolate "learning."

In this article, I take Duolingo seriously on its own terms. Not just as a language app, but as the largest gamified system ever built. Through the lens of engagement design, behavioral science, and learning theory, I’ll break down what Duolingo does brilliantly, where it quietly manipulates, and why so many users plateau despite years-long streaks. The goal isn’t to dunk on Duolingo—but to ask what it would take to turn this paid language learning game into the free language learning app its always promoted itself as.

This will help you create not just engaging and profitable products, not just products that facilitate learning, but products that holistically help users become, improving their long term quality of life.

I'm Aidan Helfant, Gamification Designer at The Octalysis Group. We leverage Octalysis gamification with behavioral science to help product & innovation leaders create user experiences that are engaging, enjoyable, and profitable while helping their customers long term wellbeing. The Octalysis Framework has been cited in over 4,000+ scholarly articles. With it, The Octalysis Group has partnered with over 175+ companies including Google, Microsoft, and LEGO, impacting more than 1 billion user experiences.

Let's dive into it!

The Genius Of Duolingo's Engagement Design

The Infamous Streak

It would be a disservice not to start with Duolingos most infamous feature: the streak...

We all know that friend who has a streak of 15,673 on Duolingo, and we all wonder if they actually work or are just playing Duolingo all day. How does this happen? As the CEO of Duolingo, Luis von Ahn, said in an interview, "If users don't show up, they can't learn anything." It's this logic that has driven the company to focus on engagement, not learning.

And the streak does a fantastic job engaging! A great CD2 streak design should balance between keeping you coming back, and avoiding you leaving because you lose your streak. Duolingo does both.

When you first join the app, you get an epic win state as you complete your day one streak and are then prompted with a CD2 consistency goal mechanic which Duolingo calls a streak goal.

This implementation intention has been proven by books like Atomic Habits and The Power of Habit to significantly increase your likelihood of committing. And boy does it work. Duolingo does everything it can to ensure you don't miss your streak the first few days of using the app. According to the Duolingo blog, users who reach a streak of 7 days of more are 3.6 times more likely to complete their language learning course.

One of the major ways they help you get to that 7 days is through a CD8 revival potion mechanic they call a streak freeze. You can equip up to two streak freezes at a time. If you miss your streak on a given day, your streak freeze "freezes" that days streak, not continuing it, but not extinguishing it either.

Streak freezes are gained by engaging in the apps lessons which encourages you to do the main desired actions to create a buffer against your precious streak.

Users are also notified an hour before their streak is lost to save it before it dies. We'll talk more about the notifications later because there is a lot to say. Finally, if worse comes to worse they also have a CD8 avoidance revival mechanic simply called a streak revival you can buy for money or diamonds, which of course is the paid currency of Duolingo—more on that later too.

Apart from helping you avoid losing your streak, Duolingo also makes you want to keep building it.

Your streak shows up as a CD5 status signal mechanic on your profile, serving as a the language learning version of driving a Mercedes-Benz to work. It's so common after someone realizes you use Duolingo to ask, "what's your streak," they should add it as a social script in Games People Play. Some people have gotten ridiculously large streaks.

Duolingo also uses a CD6 dangling reward mechanic by highlighting the potential rewards that come with getting a 7 day streak and joining the "streak society." Streak society rewards occur at major milestones (like 7, 30, 100, 365+) and include bonus streak freezes, unique app icons, profile flair, gems, and access to exclusive chests.

Everything we've talked about so far is standard streak stuff. But it's the next three mechanics that really impressed me.

Firstly, Duolingo uses a CD4 identity design mechanic through push notifications to build your identity as someone who doesn't break their streak.

Here's some examples:

• “You’re a streak legend. Don’t stop now.”
• “People with streaks like yours never skip.”
• “Future You will be proud you practiced today.”

Identity designed messages are used in app too after completing your streak:

• “You’re on fire! Keep it up!”
• “Most people don’t make it this far.”
• “Streak saved! That’s what champions do.”

The genius of this mechanic can not be understated. Duolingo has realized motivation isn't just about what you desire, or what you do, it's who you believe you are. Through building your identity as someone that doesn't miss their streak, someone that learns a language, they encourage you to come back again and again, less because of the streak, but because it's who you are. Motivation builds temporary behavior. Identity build permanent behavior.

Secondly, what impressed me is Duolingo's CD5 friend streak mechanic.

Just like your personal streak, you can start a streak with a friend. It works the same but BOTH people need to complete it to keep the streak going. This adds an extra incentive for you not to lose your streak because your friend is counting on you. Even better, either friend can use a streak freeze for the other friend if they miss a day, creating a sense of collaboration and co-commitment.

This mechanic works incredibly well. Just like with Snapchat Streaks, I sometimes hear people brag to me about how high their streak with another friend is instead of their own streak.

Third and lastly, I was highly impressed by Duolingo's CD2 streak staircasing mechanic in the form of quests.

Has this happened to you? You have a 10 minute work break and realize you haven't finished your Duolingo streak. Frantically, you go to the "bathroom" and do a lesson. Once you've finished it, you complete your first quest of the day, giving you a few rewards.

And it just so coincidently happens that you make partial progress on the next two quests. Well, you think, might as well finish the second quest. And of course, finishing the second quest makes you pretty much done with the third. Pretty soon, you've spent 30 minutes, not 10, in your "bathroom break" and you need to excuse the whole affair as food poisoning.

The genius in this design is the staircasing of rewards. Once you've done one quest, you're basically there for the next, encouraging you to spend the extra ten minutes completing that too. Then the third quest is ripe for the picking.

All in all, this makes Duolingo's streak design quite impressive. It not only helps you avoid losing your streak with freezes, revivals, and notifications, but encourages you to keep it going with the streak society, identity design, friend streaks, and staircasing.

But of course, it's not all sunshine and rainbows—this is product design, not a Pixar movie.

The Dark Side of Duolingo's Streak Design

I have a friend, let's call her R. She was an avid Duolingo user. If I had a nickel for every time I heard her say, "Oh no, my Duolingo Streak," I could afford to develop my own Duolingo without a streak mechanic... Unfortunately, R, like many other Duolingo users lost her streak. And then she left the app. Forever.

The problem with a streak design that can go to zero, is that, well, it can go to zero! It doesn't matter how ingrained your language learning identity is. You lose a 1,345 day streak and you're done.

The story is as old as time. Users join Duolingo and become excited to extend their streak. A few months in, they keep using not out of excitement, but out of terror of losing their streak.

Part of this is unavoidable. If your streak didn't go to zero, there wouldn't be nearly as much of an incentive for keeping it up. The problem becomes worse when you look at the notifications.

• "Looks like you missed your Spanish lesson. You know what happens now!"
• “These reminders don’t seem to be working.”
• “You made Duo sad 😢.”

And of course, the images when you open the app are even worse. Not to mention how they incentivize you to keep your streak with money.

Notifications and images like these isn't engagement, it's shame based manipulation. Older adults can probably get through it fine. But most Duolingo users are 18-24 with one report showing 22% of U.S. learners are between 13-17. These are teens whose social motivations are highly exploitable. And Duolingo uses that.

I'm not saying streak designs as a whole are negative. It's how they are implemented.

Here's the thing: there are other streak designs out there.

In our own gamified platform for internal engagement at The Octalysis Group (yup a gamified internal system for our gamification company), we use an ember system.

The streak is called the daily spark and can be increased by one any work day. But if you miss a day, you don't lose ALL your daily sparks. Instead you miss the chance to increase your daily spark, and you lose an ember. Everyone starts with two embers and you can have up to five. At the beginning of every month you decrease your starting ember count by one if you lose two embers, keep it the same if you lost one, and increase it by one if you didn't miss a single day.

The benefit of having embers is they boost the Octapoints, the main system currency, you gain from doing desired actions. This streak design gets the best of both worlds. You're incentivized to continue your streak, because you can't build it up without doing so and you want embers. But if you miss a day or two, you don't lose EVERYTHING. You just lose one ember, which can always be gained back next month.

This design could be used in Duolingo as well. The streak could go up with an ember system that boosts XP. But it probably won't be, because engagement in the app is apparently more important than everything else, even at the expense of manipulating users...

I don't dislike Duolingo. I'm simply sad that a company this much in the limelight for gamification is using tactics like this. But there are still many other great things about the apps engagement. Let's move on to the actual language learning and how that's engaged starting with the language journey.

The Language Journey

Possible the most well known feature aside from the streak, is the language journey of Duolingo. Through an epic line of sections, units, and individual lessons, the promise of Duolingo is by the end you will have mastered your chosen language. We will get to how well it does that later on, but it's an awesome promise.

This is a great example of the CD2 progress world manifestation mechanic in which progress is actually shown visually in the world. As users complete lessons, individual circles are flipped over and the next lesson unlocks.

This linear unlocking path makes the next desired action incredibly clear and gives a steady sense of progress as the user continues on their language learning journey.

This is complimented by a CD7 milestone unlock mechanic. As users progress throughout their language learning journey, they unlock new lessons, units, and sections with new material. This drives users to continue learning out of the curiosity of what will come next.

The whole language journey is given extra meaning with the CD2 language score which represents how well the user has mastered that target language according to the ABC system. As users go from unit to unit this score increases creating a constant sense of progression.

Another great feature is CD2 section jumping. When users begin learning a language they are assessed for how well they know that language and put on the language journey based on their results. Unfortunately, the app placed me well below where I felt I should be for my Spanish level. So I used to the jump section feature to skip way ahead to section seven as long as I could complete a test.

It's awesome that I could do this. The problem is the test was way too easy, and I ended up in a section that was actually above where I should have been. This tension of ease of engagement versus genuine learning is something we'll come back to again and again.

What does the language journey actually include?

Each section has 10s of units comprised of nine lessons. The nine lessons are always the same:

1. Vocab/Grammar/Conjugation
2. Vocab/Grammar/Conjugation
3. Story
4. Vocab/Grammar/Conjugation

• A Reward Chest

1. Video Call (Duolingo Max Only)
2. Radio
3. Practice Past Concepts
4. Video Call (Duolingo Max Only)
5. Unit Review

This CD7 Engagement Variation mechanic keeps things fresh and interesting. Whenever you get bored of one lesson type, there's a new type waiting. Each lesson type is meaningfully different enough it doesn't feel like you're doing the same thing over and over again. quite genius really.

The chest midway through each unit is also super well designed. Using a CD6 timed event mechanic, the chest sometimes pops up early in your unit journey with a timer. If you get the chest before that timer, you'll have a CD7 Mystery Box mechanic give you the chance of upgrading the chest to a higher rarity with better rewards. The combination of these two mechanics pushes you to go through your learning journey.

As we know from behavioral science, variable rewards are more motivating than non-variable rewards. And yet, people aren't willing to expend too much effort for a reward they don't even know what is. Duolingo navigates this masterfully because you have some idea of what the rewards will be beforehand—streak freeze, diamonds, or a timer prolonger. And getting to the chest isn't so much effort (maybe 30 minutes) that you aren't willing to push for a random reward.

All of the lesson build up to a CD1 Boss Fight mechanic in the form of the unit review which takes everything you've learned from the last eight lessons and combines them. This creates a epic feeling of accomplishment, the one problem being the unit reviews aren't nearly hard enough to fully give the feeling, for me at least. Maybe I'm just a Duolingo pro who knows.

The last great feature in this language journey is the CD2 year in review mechanic. This is a mechanic you see across many apps like Discord and Spotify which summarizes your journey in the app over the last year. Duolingo shows:

• How many lessons you did in your highest target language
• What language score you reached
• How much XP earned
• How much time you spent learning
• Your longest streak
• Your highest league

Lastly and most importantly, Duolingo gives you a personalized user type based on your summative journey. Unfortunately, I didn't use the app much last year and so I probably got the worst one.

Ouch.

That's not what is genius though. It's the personality off this title and the app as a whole that is so well designed. Say what you will about Duolingo, it's created an exceptional brand. Not a brand in the sense of its logo, or its notifications, but how they all come together to create a relationship to the user.

Users of Duolingo feel a connection to Duo, the owl mascot, and the app as a whole because of this conglomeration of things. They start to associate parts of their day with the app, becoming micro triggers for using it more.

Duolingo stops being a simple app on one's phone, and becomes a lifestyle, an identity.

Once again though, the language journey isn't perfect. Here are the major issues.

Bad Design In The Language Journey

To start with, match madness, rapid review, and legendary aren't engaging at all. Match madness and rapid review are optional lessons that shown up on the side of the main language learning path.

Once unlocked, you can do them to get up to three stars. Which does, well nothing. They literally just give you more XP. It's like Duolingo saw what Angry birds was doing with their three star system and just copied it. Great design doesn't come from blindly copying and pasting features from one app to another.

Great design emerges from how all the features fit together holistically.

At least match madness is a lesson type you can't get in the normal lessons, giving some CD7 for doing it. The problem is it's not good. You just match words with their translation for two minutes... It's one of the least engaging games in the app while being the least effective for language learning.

Rapid review is just a normal vocab/grammar/conjugation practice lesson but with a timer attached. The timer adds some CD6 for completing the lesson quicker, but the XP reward doesn't make up for the added pressure and difficulty of doing it compared to a normal lesson.

Worst of all, both Match Madness and Rapid Review just feel like distractions from actually engaging in the main learning journey. They just give XP, while not progressing in you the actual path.

These features would be much more engaging if they felt like compliments to the learning journey rather than distractions. Make the games more different from the main path. Make the rewards higher. Make them harder. Make it so they boost the main path in some way. Perhaps every star on the side lessons boosts XP gained from the main path.

Legendary is also a weak feature.

Legendary allows you to repeat a lesson you've already done for more XP, and a gold plated lesson visual.

The gold is cool sure, but you're literally just doing the same lesson over again! It feels like a cheap way to drive engagement for end game users. The legendary lessons are made to be slightly harder than the normal ones because they're legendary, but it's not nearly enough to make it feel like an actual challenge. Don't worry, we'll get to the lessons themselves in a bit.

Just like Match Madness and Rapid review, getting Legendary doesn't do anything unique to your learning journey. It feels like a distraction from learning again. If you got unique rewards, more XP, or more challenging lessons I might do it. But as things are, I've done Legendary once, and never again.

Another issue with the language journey is it's too rigid.

The language journey is the same for everyone, regardless of your language learning goals, or your life situation. This can, for example, lead to you having to learn about communicating in Higher Ed even if you're a 35 year old product or innovation leader. Sure you can jump skip these lessons, but it's janky.

The larger problem is Duolingo doesn't personalize your language journey to what you actually want to learn in your target language. What priority do you have for speaking, listening, reading, and writing? Doesn't matter because you're going to get all four equally.

What if you didn't care about reading or writing at all. You just want to be able to talk to people in the language? Or you want to learn to read and write but not speak or listen as much? Nothing you can do.

Of course, I understand this would be really hard to code in the app. It makes sense Duolingo hasn't been able to find a solution for this yet. But for a company thats main claim is it will teach you a language for free, this lack of personalization will show up as a glaring hole more and more as we get to the learning design.

Speaking of learning, let's explore the engagement design in the individual lessons and games on the language journey.

Language Lessons & Games

Language lessons and games! Apart from the streak, this is likely what keeps you coming back to Duolingo over and over again. And there's a reason. They're pretty fun.

Every lesson has a CD2 progress bar mechanic showcasing how close you are to the end. This gives you a sense of development and direction. And since most lessons are under 10 minutes, it's not a big ask to get through.

Every correct answer adds one to a CD3 Row Boost mechanic. Every five correct answers in a row Duo pops up on the screen with a lightning bolt to celebrate your perfect responses. I love this mechanic. The one thing that would make it better is if it clearly showed how much your XP is being boosted for perfect responses. A win state that is not perceived is a dead win state.

If you get one of the games really close there is a CD8 So Close, Second Chance mechanic which lets you try again without losing your Row Boost. This is a nice touch which makes you feel like Duo is trying to help you, not work against you. It promotes a growth mindset because you know even if you fail, as long as you're close to failing you'll get another chance.

Now for the games themselves, starting with the vocab/grammar/conjugation lessons.

Here's all the ones I could find ordered by engagement:

Great:

• Speak & Translate Sentence
• Translate This Sentence (Box Choices)

Good:

• Speak & Translate Words Flashcards
• Complete The Sentence (Fill In The Blank)
• Select Correct Translation (Three Choice Pairs)
• Complete The Chat
• Read And Respond

Meh:

• Tap What You Hear
• Repeat The Phrase

We'll get to learning later, so for now the ratings above are just about engagement. Overall, the games are good.

The feedback mechanics trigger the desired actions clearly and there's a lot of embedded win states to make getting it right satisfying. How else do you think Duolingo would have 500 million registered users?

Take the Speak and Translate Sentence game. A character says something in your native language or learning language. And then you translate, either through typing or speaking out the sentence. It's deceptively simple, which is exactly what makes it so engaging. Small, bite sized games which give a sense of progress. And the challenge is meaningfully difficult enough to be engaging, even if you make a mistake.

Speaking of mistakes, the CD3 Mistake Feedback mechanic called "explain my answer" is fantastic. Every time you get something wrong, you can use this mechanic to get in depth explanations for why you got the game wrong. This is feedback at its best. Not outcome feedback, not informative feedback, corrective feedback.

This feature is remarkable. But of course... If you want the personalized explanations, you'll have to pay for Duolingo Max.

Another great feature is a CD3 Scaffolding mechanic which scaffolds mistakes you've made multiple times so you have help with fixing them. This works alongside the fix mistakes feature at the end of all lessons where you go back and re-do any games you got wrong. Both these features make you feel like Duo is on your side, promoting a growth mindset.

Once you complete a lesson, you receive three pieces of outcome feedback:

• XP
• Lesson Mastery
• Time

This creates a meta game loop inside the mini games throughout each lesson. With these feedback mechanics, you want to gain more XP, higher lesson mastery, and less time; a perfect representation of how feedback mechanics trigger desired actions.

Love it or hate it, after every lesson you receive a mega boatload of reminders.

These include adding push notifications if you haven't, starting friend streaks, checking out the leaderboard, and a CD5 Social Reward mechanic. If you share your lesson completion with a friend, you get diamonds. Basically you're doing Duolingo's marketing for them.

Honestly, I don't have much problems with the reminders themselves—they have to be somewhere. My issue is HOW MANY of them there are. I remember after one lesson I got five reminders. Five.

Now, we need to talk about my favorite lesson type in the ENTIRE app: Call Lily.

This is the most engaging and effective learning lesson of all. Hands down. Calling Lily is a feature with allows you to have a real conversation in your target language with an intelligent speaking Chatbot. It's a truly marvelous leap in language engagement, learning, and accessibility. After a few weeks of using the app, the call Lily feature quickly became my go to method for learning Spanish.

It's got everything. Personalization, actual application, feedback after the call for how you can improve your speaking. It's miraculous. I'd find myself calling Lily while making dinner, or while walking outside, or right before going to bed.

She can do anything you ask her to do. If you want practice on the conditional in Spanish, ask her to give feedback on your usage in every one of her responses. If you want to talk about how cold it is in Ithaca New York, you can! If you want to talk to her about the article you're writing on Duolingo, you can!

But of course, there's a catch. Call Lily is a Max only feature... We'll get to my whole thing regarding paid versus in just a bit.

The last two great mechanics I want to shout out are a CD2 Badges mechanic and a CD2 Personal Records mechanic.

A great badges system gives rewards for both performance and labor. Two users who have both been using the app for a year should not have the same badges as each other. That is a clear sign of linear game play.

Duolingo does a marvelous job navigating this. You get badges for labor actions like gaining more XP as well as performance actions like how many perfect lessons you've gotten. You even get badges for personal stuff like doing lessons in the morning versus at night. This creates a healthy mix of development simply for acting while also incentivizing performance. And it adds more to the personal relationship you have with Duo.

One other great thing about the badge system is how they stack on top of each other. Some gamified systems just add badge on badge on badge, even if they're awarding the same action. This creates a garbage heap of badges that create overwhelm instead of pride. Duolingo stacks badges on top of each other (generally out of ten) so you can get higher and higher badge levels without feeling like your eyes are being attacked.

Finally, personal records are displayed on top of your profile counting your longest streak, perfect lessons, XP, and highest league. By tracking the users highest growth metrics, Duolingo not only points the direction for what the user should focus on, but creates a feeling of epic accomplishment from their entire history using the app.

All in all the lessons and games are quite well designed. Feedback mechanics trigger clear desired actions with engaging win states throughout. Duolingo scaffolds and helps you fix mistakes after lessons and outcome feedback, badges, and personal records create a feeling of development and accomplishment.

But of course, there's a lot that could still be improved.

Bad Design In The Language Lessons & Games

Most of the games in the Duolingo lessons are great.

The biggest problems come with Tap What You Hear and Repeat The Phrase games. These games are just plain meh.

Tap what you hear supposedly helps teach you listening skills, but because Duo gives you a word bank right under the game, it's easy to put all your focus on listening for the words rather than what they actually mean. It could just be me, but I find the AI voices difficult to understand when they talk to fast, so I almost always have to use the turtle speed (which for accessibility reasons is awesome they include). But the turtle speed goes full circle and makes it too easy taking me out of my engagement Goldilocks Zone.

Repeat the phrase is easily the worst game in the entire app. It's so mind numbingly easy you could do it before falling asleep. If you want to practice speaking, actually speak. Don't repeat after someone without having to translate anything in your head.

The more I used Duolingo, the more I noticed the same sentences showing up again and again throughout units.

This was fine once or twice, but by the fourth, fifth, or even sixth time, I stopped actually translating or reading the sentence and started recognizing the thing outright. Unit reviews became less challenges and more recognition assembly lines.

Once I noticed this I began looking at the sentences more analytically and was baffled by some of the things I found.

I have no issue with AI being integrated into Duolingo. Of course there are environmental, social, and economic concerns that must be addressed but ignoring AI completely is too much of a drawback to be practical for a product company.

But if AI is used it should be used well. Slop sentences like these are unacceptable, albeit hilarious.

It becomes even worse in the story and radio lessons of the learning journey.

Sometimes the stories and radio lessons repeat across units with minor variations. I've heard of users hearing the same whale watching story with a slight variation over ten times. In addition, the story and radio lessons themselves are way too easy. Most of the games simply ask what the story or radio is talking about with boxes, or ask a yes or no question. Often times, most of the choices have almost no relevance to the story or radio so you can make a blind guess and still get it correct.

This AI slop is sad, because if integrated effectively, AI could vastly improve all the lessons and games.

AI could check your responses and give personalized feedback based on your entire experience in the app. For example, imagine the read and respond game, but instead of picking from box choices, you were actually able to speak or type a response live. Then, the AI would check how well that responds to the games sentence and count it as a success or failure. Apply this logic to all the games.

This would be revolutionary.

It's like having a mini call Lily feature in every part of the app. It would be especially helpful for allowing multiple answers to the same question. One thing which frustrated me greatly the more I used the app is how it forces you to translate in a certain way. In any language their are many ways you can say the same thing, but because Duolingo doesn't have AI integrated into every game, they force you to translate it their way. AI could navigate this by reading your response and seeing how well it aligns with the spirit of the translation.

Yes, integrating AI could be impractical cost wise, but it would increase engagement and learning radically.

Another design choice which has many Duolingo fans raving is the new energy system.

Before the energy update, all users had five hearts. Anytime you got a game wrong, you lost one heart. If you lost all your hearts, you couldn't do lessons anymore, and either had to wait for them to come back (one heart every five hours), or watch an ad to gain some more.

In general, users liked this system because it promoted skill and learning, especially for free users. Even if you were on the free version, as long as you didn't get five questions wrong, you could use the app without too many distractions.

The new energy system takes away all that.

There are some test groups with different energy numbers, but this is one of the most common versions I kept seeing. Free users start with 25 units of energy. Each exercise in a lesson uses one unit of energy, right or wrong. So if you have a lesson with 17 exercises and you get them all correct that will use 17 units. You gain a random number of units between 1 and 5 at each 5 at each row boost. So for a 17 exercise lesson that would be at 5, 10 and 15 in a row.

You can watch ads to refill between lessons. Standard is one 5-30 second ad gives you between 1-5 units of energy. Every lessons uses energy but some use less. Stories for example don't really have individual exercises and they have fewer questions. You could do 5-7 stories in a row without running out of energy, but could usually only do 2-3 normal lessons in a row before you have to refill.

Basically what this means is for free users, no matter how well you play, you can only do 2-3 regular lessons before you are forced to wait for your energy to recharge (1 per hour), or watch LOTS of ads.

Users are furious.

Duolingo's central promise is free language learning education for everyone. This system breaks that promise. It's not free language learning if you can only do 2-3 lessons a day (20-25 minutes) before you're forced to stop, buy more energy with diamonds, or buy Super or Max which give unlimited energy.

It's not just the pay to win nature that is frustrating. The energy system breaks the game loop. No longer is skill as primary a factor in how long you can play. With the heart system, you could become so good you could avoid getting the paid version. But with energy, skill doesn't have nearly as much of an effect, diminishing your incentive to answer questions correctly. Either way you lose energy. Even though you get some energy back for doing games correctly, it's not enough to outpace the energy drainage.

The pay wall frustrations don't just come from the energy system—the best features in the app are unavailable for free users. Call Lily, in my opinion the greatest part of Duolingo, is completely unusable for users on the free plan. Explain my answer. Unavailable. Role plays. Unavailable.

I have no issue with Duolingo having paid plans. It needs to if it's going to survive. But when your central mission is to have free language learning for everyone, an energy system which allows for only twenty minutes of use, and no access to the most valuable features of the app feels a galaxy away from that mission.

With all this talk about bad design, lets move onto the last engagement design aspect the app does incredibly well before we talk about the learning design.

Leaderboard & League Ranks

Leaderboards are hard to get right. Most gamification designs simply slap a top ten list on the experience of some sort, wipe their hands, and call it a day.

The problem with leaderboard designs like that is that destroy motivation for everyone not on the leaderboard. If you’ve just joined the app and you see “EpicManz” at position 10 with 10,746 points, there isn’t even a flicker of hope you could ever make that.

Duolingo's design navigates these problems masterfully.

Let's break down this analysis by the previous project manager at Duolingo to understand why.

Users are ranked on a leaderboard only with users in their league rank, of which there are 10: Bronze, Silver, Gold, Sapphire, Ruby, Emerald, Amethyst, Pearl, Obsidian, and Diamond. In addition, they're only ranked with active users who engaged with Duolingo on the day you engaged with it. This ensures not only that you're ranked with people around your skill level, but also that you're playing against actual people.

League ranks are fought for over a week, with the fight ending on Sunday night. Duolingo chose Sunday night because they found many users struggle to remain engaged on the weekend, losing their streak and falling off the app. By putting the league rank ending on Sunday, there's not a huge incentive to continue using the app over the weekend, keeping many would be streak losers on the app.

The top 10 users on the leaderboard rank up, the bottom five are demoted, and the users with rankings on 11-15 are kept in the same rank. Within the top 10, the top three in the leaderboard on Sunday are given increasingly valuable chests, rewarding those that try extra hard even in the top 10.

Once you reach Diamond, the highest league rank, you compete in special tournaments. If you get top three in those tournaments, you get a special badge for your profile.

Overall, this creates a sensational engagement system, for both average and hardcore users. It's not uncommon for users to log on Sunday night and fight viscerally over the top ten ranking with Mr. Jawz from who knows where. There's even a special rapid review lesson users can do on the leaderboard page. This lesson gives more XP but is also timed and can use a lot of your time prolongers, diamonds, and XP boosters if you want to boost it.

Even with all this great engagement design, there's still one major gripe I have with the league rank system.

Bad Design Of The League Rank System

It's pay to win...

In the early ranks, it's quite easy as a free user to make it higher and higher up the leagues. Even with the energy system, simply logging on for twenty minutes a day should be more than enough to out perform most other users. But once you get to Ruby, Emerald, and Amethyst, it's no longer easy to play as a free user.

While most of the other users are Super and above, you're forced to engage with the app watching countless ads in order to engage in lessons. On top of that, you can only do 2-3 lessons a day before you run out of energy. Of course, you can watch ads to get energy back or use diamonds to recharge but both of those take time or money.

Even in those lessons, you don't have the same diamond access to afford timer prolongers and essential to getting through tough timed lessons. In other words, if you want to compete with the paid users, you have to spend double or even triple the time for the same XP.

I get why Duolingo has paid tiers. But the disparity in competition ability between the free and paid tiers is so great it's not even fair. I suggest two solutions.

One, get rid of the ability to buy diamonds.

Any app which makes the competition pay to win gets rid of the fun in the competition. We've seen this happen in Clash of Clans, Shadow of War, and so much more.

Two, change back to the heart system.

The heart system made skill much more important in determining your league rank. Instead of forcing free users to stop using the app after 2-3 lessons, they could continue playing until they lost five hearts from their own lack of skill in the language.

All right, we've deeply analyzed Duolingo's engagement design. Overall, it's fantastic! There's a reason the app has over 500 million registered users. But fantastic doesn't mean ethical, and it doesn't mean there aren't places for improvement.

Speaking of improvement, let's get real. The engagement design of Duolingo as stated by the CEO is the number one priority. But it wouldn't be a true analysis of Duolingo without diving into the learning design. Unfortunately, this is where I'll have to be a lot more critical.

Duolingo's Learning Design

As I stated at the beginning: Duolingo is not a free learning app, it's a paid language learning game. No where does that become more clear than when analyzing the learning design.

To put it simply: It's not very good.

And it doesn't have to be. Luis von Ahn has explicitly stated Duolingo is striving for engagement. We need some application that helps people get into language learning which means pushing engagement over learning. The problem is, once Duolingo gets people into language learning, it does virtually nothing to fulfill the promise the app poses: that it will teach you a language.

The truth is, Duolingo is not going to teach you a language. At least, it won't teach you a language if you're using it the way it pushes you to use it, and especially if you're on a free plan. We've seen enough people with 1,000+ streaks who can barely hold a conversation about the weather. But with a few tweaks to the learning design, we could not only have the dream of an engaging app, but an effective one as well.

As I shout out what's great about the learning design and what could be worked on, I'll try to be mindful that Duolingo is engagement first, learning second. So I'll leave many suggestions on the chopping block simply because it doesn't align with Duolingo's primary mission.

To do this, I'll dive into the learning design across the first four Arcs of my theory of learning, teaching, and knowing, Spiral Knowing. Spiral Knowing integrates many of the most insightful learning, teaching, knowing, and motivational theories to create a radical new meta-map for understanding learning design.

Let's start with Arc 1 remember, looking at what Duolingo does well and bad.

Arc 1: Remember

Before you can speak fluidly, before you can think in a language, you have to remember it. Words, sounds, structures, patterns—language is a memory-heavy skill, and there’s no way around that. But remembering isn’t about passive exposure or seeing the same sentence for the 37th time; it’s about effortful retrieval, spacing, and challenge calibrated just right.

If memory work is too easy, it creates the illusion of progress. If it’s too hard, learners disengage. Great language learning design treats memory not as a checklist, but as a living system—one that strengthens through struggle, fades through neglect, and must be continually reactivated in new contexts to truly stick.

Good Remembering Design In Duolingo

To start off, Duolingo does a decent job pushing spaced repetition.

Spaced repetition is the practice of dynamically spacing out retrieval practice over longer and longer durations to encode information into memory. Duolingo does this through its naturally paced language journey. Not only does past vocab show up, but every six lessons you practice past concepts.

Another valuable remembering feature is picture association. Often, sentences are mixed with visuals on the screen. Visuals are much easier to remember than words, so having pictures associated with words supports memory. This is the foundation of techniques like the memory palace.

Finally, one last great remembering feature is chunking. This involves breaking information down into digestible bits so learners don't get overwhelmed and remember better. Duolingo does this through its language journey and lesson game structure. Pretty nice.

But, there's a lot, that could be improved...

Bad Remembering Design In Duolingo

The biggest learning design issue of the whole Duolingo app is this: It's not hard enough.

Learning requires effort. Our brain evolved to value using less energy when we can get away with it. You can train yourself to find effort fun. But it's not our natural state. Duolingo is so fun for so many because it trades learning effort, for engaging design. The whole thesis of my learning design improvements are aimed at making the learning design progressively more difficult as users build the habit of using Duolingo.

One of the best ways we can do this is through aligning the app more with users Goldilocks Zone—not too hard, not too easy, deliciously challenging.

But Duolingo’s spacing and difficulty curve hit this zone for about two weeks… and then quitely drift into a warm bath of repetition. I’d often see the exact same sentence three times in a single session. Then again the next day. Then again two days later. By that point I’m not “remembering”—I’m recognizing, drifting through the exercises like an academic ghost: Ah yes, the cat drinks milk again. I, too, drink milk. Next.

To fix this, spaced repetition should become dynamic. If you nail an item repeatedly, push it way out into the future. If you struggle, bring it closer.

There could be a difficulty setting for the app. It starts on easy but moves to medium or hard if users want more of a challenge.

And for the love of language learning, interleave more.

Real memory doesn’t grow from doing ten straight problems about fruit vocabulary—it grows when fruit intersects animals intersects verbs intersects past tense. Life is interleaved. Language is interleaved.

Duolingo should be too.

Here’s the another core sin of Duolingo’s remembering design: Almost everything is recognition, not recall.

Boxes, word banks, multiple choice—the whole app is a buffet of options, which is fantastic for engagement and terrible for memory.

Recognition is the easiest form of retrieval. It tricks you into feeling competent because the right answer is sitting right there, waving its little linguistic hand at you. But language learning requires recall—pulling a word out of your mind with nothing but raw mental strength.

Duolingo occasionally uses cued recall (“type the missing word”), but almost never free recall (“produce the meaning entirely from memory”). The only place where Duolingo consistently forces you into true recall is Call Lily, where you actually have to generate Spanish from scratch. And surprise: that’s where all the real growth happens.

Duolingo should progressively shift users from recognition → cued recall → free recall as their streak grows and their habit solidifies. Imagine if hitting a 60-day streak didn’t just give you a badge, but unlocked a whole new difficulty tier. Imagine if your practice evolved with you.

That’s remembering by design—not remembering by accident.

A third issue is Duolingo's vocabulary.

Unfortunately… It teaches you a lot of vocabulary you will never, ever use.

I don’t need to order a unicycle at a café. I don’t need to tell someone my cousin has three purple shoes.

One of the superpowers of personal flashcard systems (Anki, Notion, paper cards taped to your bathroom mirror) is personalization—you learn what you need. Duolingo cannot do that with its current static units.

To personalize vocabulary, Duolingo would need:

• User-defined vocab spreadsheets
• AI-generated lessons around your life (work, friends, dating, hobbies)
• Adaptable units that appear or disappear based on your goals

And yes, that means the learning journey would have to be AI-generated with human touch. Honestly, this is one of the harder features I'm suggesting. But it's more an inevitability than speculation.

If Duolingo truly wants to help people remember, it needs a system that aligns the language with the learner’s life, not the other way around.

Finally, Duolingo needs to improve its interleaving.

Interleaving is one of the most powerful evidence-backed learning tools we have. It’s the practice of mixing old content with new content so your brain forms deeper connections instead of siloed vocab islands.

Without interleaving:

• Short-term learning becomes too easy → passive recognition
• Long-term learning becomes too hard → catastrophic forgetting

And Duolingo falls right into that trap.

Units often focus almost exclusively on the unit’s new vocabulary. Then, two weeks later, Duolingo suddenly resurrects an old verb conjugation and expects you to remember it. Spoiler: you won’t.

More interleaving would fix this. As users progress deeper into the language journey, Duolingo should become more interleaved, not less—blending past units into new ones until your memory network looks like a cosmic web rather than nine isolated planets.

In short, Duolingo adheres to some good remembering principles, but it falls short of personalizing and making them hard enough to really help learners learn.

With that being said, let's dive into Duolingo's learning design in Arc 2, Understand.

Arc 2: Understand

Remembering words is useless if you don’t understand how and why they work together. Understanding is what turns vocabulary into meaning and grammar into choice. It’s the difference between parroting phrases and actually thinking in a language. For adult learners especially, comprehension accelerates when patterns are made explicit—when tense shifts, conjugations, and sentence structures stop feeling like arbitrary rules and start revealing their logic. Without understanding, learners rely on memorization and guesswork; with it, they can generalize, adapt, and self-correct. In language learning, understanding isn’t a luxury—it’s the bridge between memory and real-world use.

Good Understanding Design In Duolingo

Duolingo actually shines in a few places when it comes to helping learners understand the language.

One of the strongest features is Explain My Answer.

Corrective feedback is one of the most powerful tools in learning theory, and Duolingo nails it—except for the fact that it’s paywalled behind Max. This shouldn’t be a luxury add-on; it should be a baseline feature for anyone trying to understand why they got something wrong.

Secondly, another great feature is the Fix Mistakes flow.

I love that the app takes you back to your errors, but imagine how much deeper learners could go if Duolingo actually analyzed the themes of your mistakes—maybe you consistently confuse gender agreement, or maybe verb tenses evaporate from your brain like steam. A real “weakness map” would let you target the patterns that slow your learning, instead of simply redoing old mistakes you’ve already solved.

Third, Duolingo deserves real credit for using stories, roleplays, and audio slowdown.

Humans have learned through stories since we sat around fires trying not to be eaten by wolves. Stories give context, roleplays simulate real situations, and audio slowdown helps you actually hear the language instead of pretending you did. These are features that genuinely support comprehension.

And fourth, it’s worth acknowledging how Duolingo prioritizes all four modalities—reading, writing, listening, and speaking.

A balanced input-output ecosystem like this is essential for true understanding, and many language apps don’t even attempt it. But every unit has at least one lesson for each and ever lesson has a myriad of games which target each of the modalities.

Bad Understanding Design In Duolingo

But after these highlights, the cracks start to show—and they’re big.

To start, Duolingo doesn’t teach explicit grammar, conjugations, stem changes, or tenses in any real way, which is almost criminal from a learning design perspective.

Children learn through pure immersion because they have years to bumble around inside a language; adults learn faster when we scaffold our understanding with clear, digestible rules.

If Duolingo doesn’t want to drop engagement-killing videos into lessons, that's okay—then give us compact cheat sheets for stem changes, tense charts, common conjugation patterns, and essential grammar for whatever lesson we're in. Even better, offer an option to toggle grammar explanations on or off depending on the learner. Or put the deeper explanations in a central “school” section of the app. Anything is better than pretending these crucial things don’t exist.

Another major limitation is that Duolingo rarely lets you translate in your own way.

You’re forced to match the app’s preferred phrasing, even though there are often five equally correct ways to express the same thought. This creates the illusion that language is a single narrow path rather than a wide, expressive landscape. With more AI integrated directly into lesson mechanics, Duolingo could accept the full range of natural translations and evaluate your response based on meaning—not memorization. And that shift alone would dramatically deepen user understanding because it would train your brain to actually think in the language.

In short, Duolingo has some great understanding design in the defined features of the app, but it struggles by prioritizing passive learning, and rejecting answers that aren't exactly what they want.

With that being said, let's dive into Duolingo's learning design in Arc 3, apply.

Arc 3: Apply

Application is where language stops being a puzzle and starts becoming a tool. It’s the moment when knowledge leaves the flashcard and enters the body. This is also where Duolingo—somewhat surprisingly—shows flashes of brilliance.

Good Application Design In Duolingo

First and foremost: Call Lily.

I’ve already sung this song, but it deserves another verse.

Call Lily is the single best application mechanic in the entire app. Full stop.

Application works best when learning happens in the same context you want to use it in. And nothing gets closer to that than actually speaking in a conversation that feels alive. You aren’t choosing from boxes. You aren’t filling in blanks. You are producing language under real-time cognitive pressure—searching for words, correcting yourself mid-sentence, and navigating meaning as it unfolds.

Even better, Lily can become a micro feedback loop. If you ask her to give feedback on your grammar, pronunciation, or tense usage while you’re talking, application and correction happen simultaneously. That’s gold-standard learning design. You speak, you err, you adapt—over and over again.

This is what language learning is supposed to feel like.

Second: Roleplays.

Roleplays are powerful for the same reason Call Lily is powerful—because they simulate use. But they add one more crucial layer: intentionality.

Roleplays let you practice specific scenarios you actually care about—ordering food, making plans, navigating conflict, flirting terribly in a second language. You’re not learning abstract sentences floating in a vacuum. You’re learning for something.

This ensures transfer. And transfer is the holy grail of learning.

Third: Writing sentences from scratch.

These are the hardest lesson types in the app—and unsurprisingly, the best for learning.

Being forced to generate a sentence from nothing requires recall, synthesis, and decision-making. You must choose the words, the structure, the tense, the agreement. That’s real application.

The only downside? Duolingo only accepts one correct answer.

Language is expressive, not deterministic. There are often multiple valid ways to say the same thing, and rejecting those alternatives subtly trains users to memorize rather than communicate. Still, despite this flaw, these exercises are among the few that genuinely push learners into productive struggle.

And struggle, when well-scaffolded, is where learning accelerates.

Bad Application Design In Duolingo

Unfortunately, for every moment Duolingo asks you to apply, there are ten moments where it lets you coast.

The biggest issue: Almost everything is already filled in for you.

You are not learning to speak or write a language. You are learning to recognize language fragments and slot them into pre-approved gaps.

When was the last time you spoke by dragging words from a word bank in your mind? When you talk to someone, you don’t start with a sentence template and swap out one word. You generate meaning from scratch.

This is why we see users with 1,000+ streaks who still freeze when asked a basic conversational question. They’ve trained a different skill entirely: gap-filling fluency.

This could be fixed—beautifully—by increasing generative demands as users progress. Early units can scaffold heavily. Later units should pull those scaffolds away. Let the learner wobble. Let them fall a little. That’s how competence is born.

Right now, Duolingo gives users floaties and never takes them off.

Second: There is nowhere near enough context.

Almost every exercise exists as a single, isolated sentence.

But language doesn’t work that way. The real difficulty of language isn’t translating one sentence—it’s holding context across time. Remembering what was said two turns ago. Adjusting your response based on tone. Tracking who did what to whom and when.

Duolingo almost never trains this.

Stories and radio lessons could do this—but they’re far too easy, and the questions rarely demand meaningful integration of information. Most of the time, you can answer correctly without understanding much at all.

Application should increase contextual load as learners advance: longer conversations, multi-turn responses, evolving scenarios. Instead, Duolingo keeps learners in the kiddie pool, congratulating them for splashing.

In short:
Duolingo can do application incredibly well—when it lets learners actually use language in context. But most of the time, it defaults to safety, scaffolds too much, and asks too little. The result is engagement without embodiment.

And language, at its core, must be embodied.

Next up: Arc 4 – Analyze, where we’ll talk about why recognizing errors isn’t the same thing as understanding why they happen.

Arc 4: Analyze

By the time we reach analysis, we’re no longer talking about getting exposed to a language. We’re talking about understanding its internal logic well enough to compare, contrast, and reason within it. And unfortunately, this is where Duolingo’s learning design almost entirely collapses.

Bad Analysis Design In Duolingo

The first major issue is that analysis in Duolingo is almost entirely implicit.

Yes, there is some analysis baked into recognition, discrimination, and error-identification tasks. You notice that one answer is wrong and another is right. You feel, vaguely, that this tense works here but not there. But this is slow, inefficient learning—especially for adults.

We do not need to learn languages the way we learned our first one.

Do you know how long it took you to become fluent as a child? Ten years. Probably more. And that was with full immersion, zero distractions, and a developing brain optimized for language absorption. Most adult learners do not want—or have—the luxury of a decade.

Explicit analysis is what lets us compress that timeline.

Understanding why ser is used instead of estar, or why the imperfect is used instead of the preterite, allows those distinctions to become automatic later. Without explicit analysis, learners are forced to keep re-deriving rules subconsciously, again and again, wasting cognitive resources that could have been used for fluency and expression. Duolingo largely avoids this kind of analysis, opting instead for exposure-heavy pattern recognition—which feels friendly, but plateaus fast.

The second issue is that Duolingo never helps learners see the bigger picture of what they’re learning.

You’re given a path. A very pretty path. A clean, linear path.

But you’re never told why it looks the way it does.

There is no meta-explanation for why certain tenses appear when they do, why vocabulary clusters are grouped together, or how earlier units scaffold later ones. You see the map, but you are never taught how to read it. As a result, learners often feel like they are moving forward without knowing what “forward” actually means.

Analysis thrives on structure and orientation. Without a sense of where you are in the language as a whole—what you’ve mastered, what you’re approximating, and what’s coming next—progress becomes shallow and easily forgotten. Duolingo optimizes for momentum, not meaning.

The final—and perhaps most painful—gap is that Duolingo almost never asks you to analyze using the language itself.

It teaches you the language.

But it rarely asks you to think in it. There are no exercises where you compare two passages and explain how tone differs. No prompts where you analyze why one phrasing feels more formal than another. No opportunities to critique a short text, summarize an argument, or reflect on meaning in the target language itself.

This is tragic, because analysis is one of the fastest ways to turn language from an object into a medium.

Stop just teaching me Spanish. Start helping me think in Spanish.

At the level of analysis, language stops being something you’re learning and starts being something you’re using to think. And Duolingo, almost entirely, opts out of that opportunity.

Taken as a whole, Duolingo’s learning design tells a clear story.

It excels at getting people started, struggles to help them deepen, and largely abandons them at the higher levels of learning that lead to real fluency. Memory is made too easy, understanding is kept implicit, application is selectively brilliant, and analysis is almost entirely absent.

This isn’t an accident—it’s a consequence of prioritizing engagement over cognitive challenge. And while that tradeoff has made Duolingo one of the most successful gamified products ever built, it also explains why so many learners plateau.

With that in mind, it’s time to step back, synthesize what Duolingo does exceptionally well, where it fundamentally falls short, and what it would take to design a language learning experience that is not just engaging—but genuinely transformative.

The Dream Of What Duolingo Could Become

Duolingo is, without exaggeration, one of the most masterfully engineered engagement systems ever released to the public. Its streaks, leagues, quests, badges, and leaderboards don’t merely coexist—they lock together like clockwork, forming a self-reinforcing loop that pulls users back day after day. Few products achieve this level of systemic harmony, and even fewer do so at planetary scale.

More impressively, Duolingo doesn’t just motivate behavior—it reshapes identity. Through carefully crafted notifications, social pressure, and status signals, it nudges users to see themselves not as people using an app, but as people who are language learners. The app becomes a morning ritual, a quite guilt at night, a small green owl perched on the shoulder of your conscience. That identity shift is powerful. And when done ethically, it’s one of the strongest forces in human motivation.

Some real learning does happen. Duolingo excels at lowering the barrier to entry, helping people build habits, and introducing the basic sounds and shapes of a language. Its early memory design works, and flashes of brilliance—especially Call Lily—reveal what the platform could be when engagement and learning briefly align. In those moments, language stops being a puzzle and starts becoming a living thing you can touch, bend, and use.

But as learners progress, the experience begins to feel like a beautifully decorated treadmill.

Duolingo is not a free language learning app—it is a paid language learning game. The energy system, aggressive paywalls, and restriction of the most effective learning tools turn practice into a rationed resource. Fluency is dangled just out of reach, available in full only to those willing to upgrade. The owl still smiles, but the gate is locked.

More critically, the learning design never grows up with the learner. Memory remains soft and forgiving. Understanding stays implicit and murky. Application appears in flashes, then retreats. And higher-order analysis—the act of thinking with the language rather than about it—is almost entirely absent. Learners aren’t guided up the mountain toward fluency; they’re encouraged to keep circling the base, collecting badges, watching numbers rise, mistaking altitude for progress.

None of this is inevitable.

With better scaffolding across the user journey, clearer cognitive progression, and a willingness to let learning become difficult in the right ways, Duolingo could be more than the most engaging language app ever built. It could be a bridge rather than a loop. A path rather than a cage. A system that doesn’t just teach you to return every day—but teaches you to finally, confidently, step outside the app and speak.