Duolingo’s 52.7M daily users: the habit behind the hype

What happens after the owl gets your attention?

The green owl is easy to remember. The more useful question for an iOS builder is what makes someone open the app again on an ordinary Tuesday, when there is no campaign to react to.

Duolingo’s 2025 annual report, filed on 27 February 2026, offers a useful snapshot:

Average daily active users

52.7M Q4 2025 · 40.5M in Q4 2024

Paying subscriber accounts

12.2M 31 December 2025 · 9.5M a year earlier

These are global company figures, not iOS-only results. DAU averages daily unique users of the app or learning website across the quarter. Subscribers are paying accounts active at period end, excluding free trials and non-paying family members. These populations and windows differ; dividing them does not give a conversion rate.

The report attributes DAU growth largely to retaining current users. That makes the return visit a useful place to look, even though the filing cannot isolate the contribution of any single feature.

The small change: let one lesson count.

In a product write-up published on 19 November 2020, Duolingo described separating its streak from its daily learning goal. Previously, learners could show up without doing enough to preserve their streak. The revised version let a single completed lesson extend it.

The A/B test improved day-14 retention, while fewer learners reached their full daily goals. That trade-off matters: lowering the minimum made consistency easier, but changed how much some people did in a session. This older experiment illustrates a mechanism; it does not prove that one change caused the company’s 2025 scale.

Our reading: a demanding goal can accidentally turn partial progress into failure. A useful product can recognize the small win and still leave room for a longer session. The interesting design work is deciding what counts as a meaningful minimum.

Give the upgrade a job to do.

Duolingo describes a freemium model with advertising and paid subscriptions offering additional features. Its annual report also notes that some people subscribe after months or years of use.

For a small iOS studio, this suggests a sequencing hypothesis: demonstrate repeatable value, then offer a paid improvement that matters to someone already doing the activity. A prettier paywall cannot answer why the user should come back.

A hypothetical journaling app might help someone finish one useful reflection before offering ways to revisit and organize a growing history. The paid benefit becomes easier to understand once that history exists. This is an experiment to test, not a claim about Duolingo’s exact conversion journey.

Try the smaller promise in your own app.

Choose a recurring task your users already value. Keep the acquisition channel and offer consistent, then randomly assign eligible new iOS users to the existing completion goal or a smaller meaningful one. Keep the assignment stable.

• Define the win: a saved reflection, completed practice, or useful result. Opening the app alone is too easy to reward.
• Watch both sides: measure completion and returns at fixed cohort ages, alongside quality or depth. More visits can hide less useful work.
• Follow the money later: compare paid starts, renewals, refunds and service costs once both groups have had equal time to mature.

A daily streak only fits a daily job. An occasional travel utility may need a reliable return experience months later. Copy the question Duolingo tested: are we making useful progress unnecessarily hard to recognize?

Use the small event plan to instrument the experiment, and keep paywall conversion denominators separate from engagement totals.