Start learning a language and someone will tell you to quit Duolingo. It comes up under every study video and in every thread about which app to pick. Almost nobody explains why.
That is because it is not one complaint. It is four.
The app rewards the wrong thing. The lessons train the wrong skill. The company did something people did not like. And it stops helping sooner than you expect.
Some of those are fair. Not all of them.
The streak stops being about the language
The app is working, not breaking. Streaks, points, leagues, a green owl with your phone number. It works so well that people protect a streak they stopped caring about, tapping through a lesson at 11:58pm in a language they gave up on in March.
Duolingo is very good at one thing: getting you to open Duolingo.
In April 2026 the hearts started turning into energy. You get 25 units. Every question costs one, right or wrong. Careful learners run out as fast as careless ones, and a free session ends after two or three lessons. More than 11,000 people answered a community poll about the change, and close to half said they disliked it. Someone started a petition to bring the hearts back.
None of that came from teaching. It came from games. What the app really counts is dwell-time, the minutes you spend inside it, and the way past the wall is a loot-box or a subscription. That is what people mean by enshittification: a free thing you got used to, slowly getting worse until you pay.
Tapping the right tile is not speaking
The second complaint is about the exercises.
Most of them show you the answer. Four choices, a word bank, a sentence you build from tiles the app hands you. You get very good at picking the right piece. That is a real skill. It is not the one you need in a bakery in Lisbon, where nobody gives you the words first.
The help got thinner too. The sentence forums closed in 2022. The Tips notes disappeared from most of the app. Explain My Answer is free now, but it explains one answer at a time. So you learn the pattern and never the rule. You can write the right sentence and still not know why the last one was wrong.
What the research actually found
Here is the part the critics skip.
In 2021 the journal Foreign Language Annals published a study of 225 American adults. They used Duolingo and nothing else, for beginner Spanish or French. At the end they read and listened about as well as university students after two years of classes. That is a real result and it should be said out loud.
Now look at what was tested. Reading and listening. The two things you can do lying down. Nobody tested speaking, and speaking is most of what people mean by fluency when they say the app failed them. Duolingo's own researchers helped write the study, which does not make it false. It makes it something to check, not something to quote.
So there is a limit. It sits higher than the critics say and lower than the points suggest. Duolingo will take a total beginner to real reading. Then it keeps handing out points long after you stopped learning much.
The AI memo made people angry
The third complaint has nothing to do with learning.
In April 2025 the CEO, Luis von Ahn, told staff the company would be AI-first. It would stop paying contractors for work AI could do. The memo went public within days. People said they were quitting, and posted their broken streaks under posts about other things.
It showed in the numbers. On the August 2025 earnings call, daily users were up 40% on the year before. The year before that, they were up 60%. Von Ahn said he had not given enough context. The company softened the message, said it was still hiring, and stopped the jokey posting.
Keep this one separate. Someone who deleted the app over that memo is not saying the Spanish course got worse. They are saying they do not want to pay for it. That is a different question with a different answer.
The words it never teaches you
The fourth complaint is the quiet one. The course picks the words. Your life picks other ones.
A Spanish lesson will teach tio as uncle. That is correct. In Spain it is also what you call your friends, all day, every day. Learn the first meaning and you can read a family tree. Learn the second and you can join the conversation.
Some words have nowhere to sit in a lesson. Japanese speakers say よろしく all the time: when they meet you, when they ask for something, when they hand you work. English has no word for it. There is nothing to put on the other side of a translation exercise, so a translation app tends to skip it.
Knowing a word by sight can also go wrong. The German Gift looks like the English present. It means poison. The present is Geschenk.
What to do with the owl
None of this means delete the app. It means giving it a smaller job.
Use it as a warm-up, not as the whole plan. Cap it at ten minutes. Let the streak break once on purpose, because a habit you are afraid to lose is running you. Put the spare time into the two things the app skips: speaking out loud, and the words you actually ran into this week.
The part after the lesson
That last one is where Linguin comes in. A lesson gives you the word it planned to give you. A dictionary gives you the word you just walked into.
Look one up and you get the meaning in your own language, across 120+ languages. You also get what a lesson leaves out: how formal it is, who says it, and where it would sound wrong. Slang and memes get real entries, so the word your Mexican friend dropped in a message is something you look up instead of guess.
Exercises then practise your own saved words instead of a fixed course. Chat covers the speaking part, out loud, with something that answers back and tells you how it went.
The owl is not the problem
Keep it or delete it. Just decide for the right reason. As a way to build a daily habit, nothing beats it. As a way to learn a language, it was only ever one piece.
The next word you hear and do not understand will not be in any lesson. That part nobody can plan for you. Keep your easy ten minutes if you enjoy them. Just stop calling them the plan.