You have a streak of a few hundred days. You open the app every evening and finish the lesson. Then someone asks whether it is doing anything, and you are not sure what to say.
The short answer: yes, Duolingo teaches beginners something you can measure, but how much is less settled than the company's own summaries suggest. In the studies, people who stick with it improve on standardized reading and listening tests, and some reach a basic level of speaking. Almost all of those studies were run by Duolingo staff or paid for by Duolingo. The few independent ones are small: one followed nine adults, and one found that children learned about as much with the app as in a class. No study I could find tests whether Duolingo makes you fluent, or whether it gets you to the point of reading a novel.
I did over a year and a half of Duolingo while learning Dutch, along with evening classes and meetups, and it got me to a certain point before I stopped making progress. So I have a stake in this question twice over. Below is what each study measured, who ran it and who paid for it, what outside researchers make of them, and how to tell whether the app is working for you.
What do Duolingo's own studies show?
Duolingo publishes its research on a page that lists its efficacy studies. These are the ones with the most detail behind them.
Beginner course, reading and listening (2021). In a paper in Foreign Language Annals, a peer-reviewed journal, Jiang and colleagues, mostly Duolingo staff, tested 225 adults in the US who had little or no knowledge of Spanish or French, had used only Duolingo, and had finished its beginner course. On the American Council on the Teaching of Foreign Languages (ACTFL) tests, they reached Intermediate Low in reading and Novice High in listening, which the authors report as comparable to US university students. In their words, "no other skills were assessed."
ACTFL's scale goes Novice Low, Mid and High, then Intermediate Low, Mid and High, and up from there, so Intermediate Low is the fourth step from the bottom.
Further into the course (2021). A Duolingo blog post reports 340 learners who had finished Unit 7 of French or Spanish without other classes, apps or materials. On the same ACTFL reading and listening tests, they scored about as well as US university students in their fifth semester. The post does not say how the learners were recruited.
Speaking (2021). Another post reports 258 learners who took the Pearson Versant speaking test after Unit 5: 66% of the Spanish learners and 53% of the French learners scored A2 or higher, the second of six levels on the CEFR scale. The Spanish learners' pronunciation was close to B1, while for the French learners pronunciation was the lowest score. There was no comparison group, and the full report is a Duolingo white paper, not a journal article.
University students over 12 weeks (2024). The study with the most careful design is by Smith of Arizona State University, with Jiang and Peters of Duolingo, in Language Learning & Technology. Students learned Spanish on Duolingo for 12 weeks, about 27 hours in all, and took a third-party test (STAMP 4S) before and after. They improved by more than one sublevel in reading, writing and speaking, and by about two thirds of a sublevel in listening. The paper states: "This study was supported financially by Duolingo." Of 69 students who started, 48 finished, and there was no group that learned another way, so the study shows they improved, not that they improved more than they would have in a class.
Duolingo's page also lists surveys: in 2024, 97% of teachers asked rated it effective, and in a 2025 survey over 90% of the learners asked said they felt prepared. Those tell you how people feel about the app. They are not tests.
Who ran and paid for each study?
| Study | Who ran it | Who paid | What was tested | Comparison group |
|---|---|---|---|---|
| Vesselinov and Grego, 2012 | Vesselinov and Grego | Duolingo | A multiple-choice test (WebCAPE) | None |
| Jiang and colleagues, 2021 | Mostly Duolingo staff | Duolingo, as employer | Reading and listening | University students from another study |
| Unit 7 study, 2021 | Duolingo | Duolingo | Reading and listening | University students from another study |
| Speaking study, 2021 | Duolingo | Duolingo | Speaking | None |
| Smith, Jiang and Peters, 2024 | Arizona State University and Duolingo | Duolingo | All four skills | None |
| Rachels and Rockinson-Szapkiw, 2018 | Independent | Not stated | Vocabulary and grammar | Children in a regular class |
| Loewen and colleagues, 2019 | Independent | Not stated | Language measures | None |
A study paid for by a company can still be careful, and Duolingo deserves credit for using outside tests and publishing its methods. But when nearly all the evidence comes from one side, it is worth reading what people outside the company found.
What do researchers outside Duolingo say?
Two independent studies, small and mixed. Shawn Loewen and colleagues at Michigan State University and other US universities followed nine people learning Turkish on Duolingo for a semester in 2019. They improved, and those who spent more time tended to improve more, but participants also reported frustration with the materials and motivation that varied. In 2018, Jason Rachels and Amanda Rockinson-Szapkiw compared third and fourth graders who learned Spanish with Duolingo for 12 weeks with children who followed their regular class. On a 50-question vocabulary and grammar test, they found no significant difference between the groups, and none in how confident the children felt, and they concluded that Duolingo is a useful tool for teaching Spanish to children of that age.
A review of Duolingo's own reports. In 2022, Taylor, writing from a senior high school in Taiwan, reviewed three of Duolingo's research reports in the journal CALL-EJ. He credits the company for being open about its methods and for using outside tests, and he finds problems that matter for anyone reading the headline numbers:
- Who took the test. In one study, 8,367 people had finished the beginner Spanish course. 813 answered the invitation, 499 qualified, and 195 started the test. Participants were offered a certificate. Taylor calls this selection bias: the people who volunteer for a test may not be typical of everyone who finished.
- Who they were. In one report, 90% had a university degree and 51% a graduate degree.
- Who they were compared with. Over 80% of the university students in the comparison data came from three universities with a language requirement.
- What else they did. Learners were screened for other classes and apps, but not for other contact with the language. In one report, about 30% said they were learning it for social reasons and about 20% for work.
His conclusion is careful: it is "very likely … though not certain" that some of the progress came from Duolingo, the claims that it equals university semesters are not supported, and learners should not rely on it alone.
The first study, read closely. Duolingo's first efficacy study, by Vesselinov and Grego in 2012, was funded by Duolingo. The linguist Stephen Krashen read it in 2014 and pointed out what the averages hide. The volunteers answered an internet ad, averaged 35 years old, and 69% were college graduates. They were asked to use Duolingo for 30 hours, but the average was 22 hours, a quarter did 8 hours or less, and only 90 of 156 lasted to the end. Their gains varied a lot by why they were learning: people learning for travel gained the most per hour, people learning out of personal interest or for school the least, though the travel group had only ten people in it.
So does Duolingo work?
Put together, the evidence says this:
- People who use Duolingo regularly learn something you can measure, mostly in reading and listening at a beginner level. That is consistent across Duolingo's own studies and the one small independent study of adults.
- How it compares with a class is not settled. Duolingo's comparisons use university students from another study, chosen differently. The one independent study with a class to compare with, in children, found the two about equal.
- No study tests fluency. None measures holding a conversation with a stranger, following a film, or reading a novel.
- Most of the evidence comes from Duolingo. That does not make it wrong, but it means the independent picture is thin.
What I take from it: Duolingo is good at what it is built for. It takes a total beginner to a foundation, for free. The evidence does not support the stronger claims, and it says little about the stage after that.
How can you tell if Duolingo is working for you?
The 2024 study found something useful here: the hours students logged did not predict how much they improved, but the number of sessions they completed and how accurate they were did. So counting your streak or your minutes is not the test. Check what you can do outside the app:
- Read something you did not choose for the app. A short news item for learners, a children's story, or the first page of a graded reader at your level. If you follow it without translating every line, it is working.
- Take a level test. A free reading test in Spanish, French, Portuguese, German, Italian or Dutch places you on the CEFR scale in about five minutes. Take it again in two or three months and compare.
- Listen to something slow. A podcast for learners or a video with subtitles in the language. Notice how much you follow compared with a month ago.
- Watch your accuracy, not only your streak. If you finish lessons by guessing and repeating mistakes, the study suggests you are getting less from the time.
If you are learning more in the app every week but can do no more outside it, that is the point where many people feel stuck, and I wrote about why the intermediate plateau happens.
What should you add when Duolingo stops working?
The students in the 2024 study told the researchers what they wanted more of: practical listening, speaking, pronunciation and conversation. That is a fair list of what to add.
I have written the practical side elsewhere, so I won't repeat it here. What to do after Duolingo sorts the gaps (volume, unpredictable language, speaking) and what fixes each one. The Duolingo alternatives post is about the one I chose: reading books, and listening while you read. For me that was the thing that got me past where Duolingo left me, in Simply Fluent and before it existed.
You don't have to stop Duolingo to add something. Everything you learned there counts toward the next thing you read or hear.
What to do next
Take the reading level test for your language today, and write down the result. Then add one thing outside the app for the next two months, a short story a few times a week or a podcast on your commute, and take the test again. The number will tell you more than the streak.
Sources
All checked on November 17, 2026.
- Duolingo, results of efficacy studies (2024): blog.duolingo.com/results-duolingo-efficacy-studies
- Duolingo, intermediate learning study (2021): blog.duolingo.com/effective-intermediate-learning
- Duolingo, speaking study (2021): blog.duolingo.com/how-well-does-duolingo-teach-speaking-skills
- Jiang, X., Rollinson, J., Plonsky, L., Gustafson, E. and Pajak, B. (2021). Foreign Language Annals, 54(4), 974–1002: doi.org/10.1111/flan.12600
- Smith, B., Jiang, X. and Peters, R. (2024). The effectiveness of Duolingo in developing receptive and productive language knowledge and proficiency. Language Learning & Technology, 28(1), 1–26: hdl.handle.net/10125/73595
- Taylor, C. (2022). A review of Duolingo's in-house research reports. CALL-EJ, 23(3), 156–162: callej.org
- Krashen, S. (2014). Does Duolingo "trump" university-level language learning? International Journal of Foreign Language Teaching, 9(1), 13–15: sdkrashen.com
- Loewen, S. and colleagues (2019). Mobile-assisted language learning: A Duolingo case study. ReCALL, 31(3), 293–311: doi.org/10.1017/S0958344019000065
- Rachels, J. R. and Rockinson-Szapkiw, A. J. (2018). The effects of a mobile gamification app on elementary students' Spanish achievement and self-efficacy. Computer Assisted Language Learning, 31(1-2), 72–89: digitalcommons.memphis.edu
- Duolingo status page: status.duolingo.com
Common questions
- Is Duolingo effective?
For beginners who keep using it, the studies say yes, to a point. Learners who finished Duolingo's beginner Spanish or French course reached a low intermediate level in reading on a third-party test, and university students who used it for 12 weeks improved in reading, writing and speaking. Both studies were run by or paid for by Duolingo, and neither had a group that learned another way.
- Can you become fluent with Duolingo?
No study has tested that. The published studies test beginners and early intermediate learners on reading, listening and speaking tests, and none measures fluency. What to add once the course is not enough is in what to do after Duolingo.
- Is Duolingo worth it?
As a start, I think so. The core app is free, it gets you from nothing to a foundation, and every word you learn there is one you will not have to look up later. It is worth less once lessons are the only thing you do, and the students in a 2024 study asked for more listening, speaking and conversation practice.
- Duolingo not working today: is it down?
Check status.duolingo.com, Duolingo's own status page, which shows whether the app, the Duolingo English Test or Duolingo for Schools has a problem right now. If it shows everything running and the app still fails, Duolingo's help centre is the place to ask.


















