Do Language Apps Actually Work? An Honest Answer

Somebody has a 900-day streak and cannot order lunch. Somebody else insists the app is a slot machine dressed as a classroom. Both camps are loud, and both are partly wrong.

The research is actually fairly clear, and it is more interesting than either caricature.

What the apps genuinely do, and the evidence is real

Start with the part critics get wrong. Language apps are not snake oil. Peer-reviewed studies consistently find real gains in vocabulary and grammar, and reasonable engagement over time. One study of independent learners found significant improvement across receptive and productive measures after roughly twenty-seven hours of study. A Cambridge study of beginner French found app learners' communication skills broadly comparable to those of classroom learners after a semester.

Read that last sentence twice, though, because it cuts both ways: comparable to classroom learners. And we have already discussed how well classroom French works.

One necessary disclosure, which you will rarely see mentioned: a substantial share of the favourable research on Duolingo is funded and conducted by Duolingo. That does not make it false. It does mean the independent findings deserve more weight, and the independent findings are more sobering.

What the independent research found

A study published in the Modern Language Journal examined learners who completed an entire Duolingo Spanish course. Their proficiency landed around A2 on the European framework — officially a basic user. That is the level at which you can handle simple, routine exchanges on familiar topics. It is real. It is also not conversation.

More pointedly: a study of learners tracked through a semester of app-based study found that they moved from essentially no functional ability to being able to communicate minimally, in isolated words and memorised phrases. The researchers were explicit that these gains did not equip learners to hold a conversation.

The structural problem, and it is not fixable by a better app

Here is the crux. Language ability splits into two categories, and they are not the same asset:

• Receptive. What you can understand when you meet it. Reading. Listening. Recognising.
• Productive. What you can generate yourself, unprompted, in real time. Speaking. Writing.

Apps are extremely good at building receptive skill. Multiple-choice recognition, matching, and translation drills train exactly that. The research repeatedly finds users landing further along receptively than productively.

But an app cannot give you the thing that makes speech: negotiation of meaning. Being misunderstood. Noticing you have been misunderstood. Trying again, differently. Being interrupted. Being asked something you did not prepare for. Failing, in front of someone, and repairing it.

That exchange is where speaking is built, and it structurally requires a second person who does not know what you are about to say. An app always knows what you are about to say. It wrote the sentence.

What about the AI conversation features?

They are a genuine improvement, and the honest thing is to say so. Studies of learners using AI roleplay features found higher self-efficacy and greater speaking gains than those using the free version alone. That is a real finding and it is worth something.

But notice what is being measured: learners felt more prepared. Confidence is a real precursor to speaking, and it is not the same thing as speaking. The AI is patient, predictable, and never bored. A French person in a cafe is none of those things, and the gap between practising with the first and surviving the second is exactly the gap that keeps catching people out.

The honest recommendation

• Keep the app. For vocabulary, for daily contact with the language, for keeping a habit alive between lessons, it is genuinely useful and it is nearly free.
• Stop expecting it to make you a speaker. It will not, no matter how long the streak. That is not a flaw in your discipline. It is a limit of the format.
• Add one hour a week of real conversation. One hour with an actual person who responds unpredictably will do more for your speaking than twenty hours of tapping.
• Use the app for what it is good at, and stop feeling guilty about the rest. The streak was never going to be enough. Nobody told you that.

Apps solve the problem of not knowing enough French. Most adults do not have that problem. They have the other one.