DuolingoDuolingo

The EU wants to turn off your streak: how to keep learning a language without the counter

Team at AiToolsForAppsWritten with AI assistance16 min read

In summary

  • The European Commission adopted the 11th of November 2026 its proposal for the Digital Fairness Act (DFA), the largest overhaul of EU digital consumer law in over twenty years (European Commission).
  • Among the "addictive design" practices that users should be able to turn off are autoplay, notifications unrelated to your own activity, and daily streak engagement incentives — and, according to the proposal, turning them off should not cost you your accumulated progress.
  • It is not a prohibition: it is an opt-out. The Greens/EFA criticize it precisely for that, and the regulation still has to go through Parliament and Council: mandatory enforcement would not arrive before 2029.
  • The streak functions as a retention lever, not as a learning measure: Duolingo reports 64.9 million daily users in the third quarter of 2026 and streak experiments that move retention to 7 days by up to +14%.
  • What supports language when the counter is turned off is measured: spaced repetition (medium-high effect in a 2022 meta-analysis), grammar worked with self-correction, and graded reading at the CEFR level where you understand 98% of the words.

If you study a language with your mobile, something happened this week that affects you more than it seems: Brussels has put in writing that your daily streak is an engagement mechanism and that you should be able to turn it off. At AiToolsForApps, we have been building aiDuolingo, a set of tools that generate flashcards for Anki, grammar exercises, and graded readings by level for those who want their own material and not a counter watching them. So this article is not about the app: it’s about what the DFA changes, what the evidence says about why the streak engages without teaching, and how language study is sustained when daily pressure disappears.

What exactly did the EU propose on November 11, 2026?

The European Commission adopted on November 11, 2026, the proposal for the Digital Fairness Act, along with a simplification directive and a review of the Regulation on consumer protection cooperation. The package outright bans dark patterns, imposes a principle of "fairness by design" that requires addictive design functions to be able to be turned off, demands that subscriptions be canceled with the same ease with which they are contracted, and restricts personalized advertising aimed at minors. The Commission itself presents it as the largest update to EU consumer rules in over 20 years.

The official figures accompanying the proposal:

DataValueSource
Date of proposal adoptionNovember 11, 2026European Commission
Estimated annual savings for consumers (full package)up to 13.7 billion eurospress release from the Commission
Consumers covered in the single market450 millionstatement from Executive Vice President Henna Virkkunen, 12-11-2026
Popular websites and apps with at least one dark pattern97% (behavioral study by the Commission, 16-05-2022)EUobserver
Penalties foreseen for cases with European reachup to 6%IAPP analysis

The Commissioner for Justice and Consumer Protection, Michael McGrath, summarized the logic of the package in one sentence: convenience should never come at the cost of fairness.

Why does the Commission consider the daily streak a dark pattern?

Because the streak does not reward what you learn, but rather that you return. The examples of addictive design included in the proposal encompass autoplay, notifications that do not respond to your own activity, engagement incentives like streak rewards, and systems that simulate a personal relationship — and the Commission wants users to be able to opt out of the daily learning or exercise streak without losing accumulated progress, a detail that is the heart of the measure.

That last condition is the technical key. Today, in most apps, turning off the streak equates to losing it: the design turns opting out into a punishment. The DFA separates the two things: the history of what you have studied is yours; the counter that pressures you is a function you can close. According to the review of the proposal published by gHacks, "aggressive notifications" and emotional pressure would also be included in the package — messages that make you feel guilty for skipping a lesson.

The proposal has received criticism from the opposite side of what one might expect. MEP Kim van Sparrentak (Greens/EFA), negotiator of the file, argued on November 11, 2026, that the Commission fell short by choosing opt-out instead of prohibition: in her view, fair and safe services should be the standard and not the exception that users have to seek out in the settings (Greens/EFA).

When would you really have the option to turn off the streak?

Not in 2026, and probably not in 2027 either. What was adopted on November 11 is a proposal from the Commission: it is now entering negotiations with the European Parliament and the Council, and the text can change a lot before it becomes law. While that happens, apps can maintain their current design without changing anything.

MilestoneEstimated moment
Pre-public consultation17-07-2025 to 24-10-2025
Proposal adopted by the Commission11-11-2026
Parliament-Council negotiation2026–2027
Expected final adoptionlate 2027 / early 2028
Entry into force and phased implementation2028–2030
Realistic mandatory applicationnot before 2029

Estimates based on the legislative calendar itself and the legal analyses published after the proposal; the draft provides for entry into force 20 days after its publication and application 12 months after adoption. Practical translation: if you are interested in studying without relying on a counter, you have between two and three years in which that remains your method decision, not an enforceable right.

What the streak really measures (and what it doesn’t)

The streak is an excellent retention tool and a mediocre learning metric. The numbers published by the apps themselves say it bluntly: they measure return, not mastery.

Comparison between a paper calendar with daily marks and empty spaces and a box of cards with dividers at increasing intervals
Comparison between a paper calendar with daily marks and empty spaces and a box of cards with dividers at increasing intervals

In its third quarter 2026 shareholder letter (30-09-2026), Duolingo reported 64.9 million daily active users (+29% year-on-year), 153.1 million monthly users (+13%), 13.1 million paying subscribers, and $309.0 million in revenue, and described having conducted "hundreds of experiments" in parts of the product such as notifications and streaks, aimed at increasing retention and daily users.

And the results of those experiments, recounted by the company itself in its blog about streaks, are eloquent:

Streak experimentReported effect
Streak Wagerstatistically significant increases in retention at day 1, 7, and 14; the largest effect at day 7, +14%
Weekend Amulet+4% chance of returning a week later; −5% chance of losing the streak
Daily user drop on weekendsbetween 5% and 10% compared to the midweek peak

Note what is being optimized: returning, not forgetting less. None of those figures are a measure of retained vocabulary, grammatical accuracy, or reading comprehension. And the literature on gamification points out that the motivational effect of external elements declines with exposure, due to novelty wear-off, while the student’s attention shifts from the value of learning itself to the reward associated with it — the mechanism that self-determination theory has described for decades.

Spaced review is the piece with the most evidence behind it

If you are only going to change one thing in your method, change when you review, not how much. The meta-analysis by Kim and Webb published in Language Learning in 2022 grouped dozens of experimental studies with thousands of second language learners and found that spaced practice produces a medium-high effect on learning compared to massed practice, with a decisive particularity: short intervals perform equally as long ones in an immediate test, but clearly lose in delayed tests (Kim and Webb, 2022).

That has an uncomfortable consequence for the daily streak model: if the goal is to remember in a month, studying the same thing every day is worse than studying at increasing intervals. The streak pushes for the former. A spaced review system — a deck of cards with expanding intervals — pushes for the latter.

Practical implementation doesn’t need any specific app: it needs a deck of cards with two sides and a calendar of intervals. Anki is the de facto standard because it implements the algorithm and because it imports delimited text files, so the material can come from anywhere.

Paper grammar has a concrete advantage: deferred correction

The value of a printed grammar sheet is not nostalgic, it is methodological: it separates the attempt from the answer. In an app, feedback is instant and the error is corrected before you even notice it; on a sheet with the solutions on a separate page, you solve the ten exercises, then check, and retrieve the information from memory in the interval. That effort of retrieval is precisely what consolidates.

Two printed sheets overlapping: the top one with fill-in-the-blank exercises and underneath a separate solutions sheet, with pencil and eraser
Two printed sheets overlapping: the top one with fill-in-the-blank exercises and underneath a separate solutions sheet, with pencil and eraser

The three exercise formats that yield the most performance per minute on a specific grammatical point are also the easiest to set up on paper:

Exercise TypeWhat it trainsWhy it works without a screen
Fill in the blankproduction of the correct form in contextforces retrieval, not recognition
Choose optiondiscrimination between competing formsexposes the exact contrast you confuse
Rewrite the sentencecontrolled transformation (tense, voice, person)forces you to apply the rule, not to remember the sentence

The operational key is to keep the solution sheet out of sight while you solve and to be able to repeat the same point with different exercises: a mastered grammatical point is not the one you got right once, it is the one you got right three times with material you had not seen before.

Graded reading works for a numerical reason: 98% coverage

Graded reading is not "easy reading": it is reading calibrated so that you know about 98% of the words in the text. Below that threshold, comprehension breaks down and you can no longer infer the meaning of the unknown through context; right at it, the remaining 2% is deduced and learned without a dictionary. That threshold, documented in Paul Nation's research on extensive reading and included in the reference material of the Extensive Reading Foundation, is the reason why graded readers are published at levels aligned with the CEFR.

Open short reading book on light linen next to a notebook with two blank columns as a glossary, a bookmark, and cards
Open short reading book on light linen next to a notebook with two blank columns as a glossary, a bookmark, and cards

From there comes the recipe: a text at your CEFR level, a short glossary with the words you truly did not know, and a few comprehension questions to verify that you have understood and not just recognized. The jump from the highest graded readers to unadapted novels is, in fact, a studied problem: the vocabulary gap between the two is larger than assumed (Extensive Reading in a Foreign Language, 2016). Moving up reading levels has to be gradual, just like review.

How many hours it really takes to level up in CEFR

Between 150 and 200 hours of guided work per level, according to Cambridge English's reference. It is the data that best deactivates the anxiety of the streak: progress in a language is measured in hundreds of accumulated hours, not in consecutive days.

CEFR LevelAccumulated guided work hours (approx.)
A190–100
A2180–200
B1350–400
B2500–600
C1700–800

Source: Cambridge English, guided work hours. The figures are accumulated from zero, not from one level to the next; Cambridge estimates around 200 hours for the jump between consecutive levels.

Do the arithmetic with your own case: 20 minutes daily on an app is about 120 hours a year. That is, a whole year of perfect streak leaves you around half a level jump if all that time were effective work — and much of it is not, because it is consumed in interface, animations, and repetition of what you already know. The European context helps to situate the effort: according to Eurostat, in 2022 25.3% of adults aged 25 to 64 in the EU reported not knowing any foreign language, and only 27.6% of those who knew at least one considered themselves competent in their strongest foreign language.

A week of study without a counter: the distribution that is sustainable

The substitute for the streak is not more willpower: it is a week with three types of different sessions and a day of flexibility. The streak fails because it demands all seven days equally; a plan with your own material allows unequal sessions without penalty and, above all, without losing anything when you skip a day.

DaySessionTimeDeliverable
Mondayspaced review of the deck15 minpending cards up to date
Tuesdaygrammar sheet + separate solutions25 minone grammatical point with 10 corrected exercises
Wednesdayspaced review15 min—
Thursdaygraded reading + glossary30 min8–12 new terms that go to the deck
Fridayspaced review15 min—
Saturdaysecond grammar point or re-reading25 minconsolidation
Sundaydeliberate flexibility0 minnothing that breaks

Total: about 2 hours and 5 minutes per week, close to 100 hours a year, with a physical or downloadable deliverable each session. The difference with the streak is what continuity depends on: here it depends on the accumulated material — a deck that grows, a folder of sheets, a glossary notebook — not on a number that resets to zero.

What does all this mean for the language app market

That the retention engine of the category becomes an optional feature, and that pushes the value towards exportable material. If the streak can be turned off and progress cannot be confiscated, what differentiates a product stops being how many consecutive days it retains you and starts being what you take with you: a deck that you can import into Anki, a sheet that you can print, a text at your level with its glossary.

That’s where aiDuolingo fits in, the tool from AiToolsForApps for those studying languages with their own materials. It generates decks of exportable cards in CSV, PDF, TXT, and ZIP —the CSV is already formatted for Anki, with tab separator, basic note type, and two columns (front and back)—, grammar cards in PDF with the solutions sheet on a separate page, and graded readings with a glossary of 8 to 12 entries and 5 comprehension questions with their solution. It works on five CEFR levels (A1, A2, B1, B2, and C1), across ten topics, with 24 grammar points from the catalog —12 in English and 12 in Spanish— and 240 curated exercises of three types: fill-in-the-blank, choose an option, and rewrite. You can input your own list of up to 40 words, your own topic or your own grammar point, correct all the cards in the editor before exporting, and request audio of the words from the deck. And there is a clarification worth reading: it is not affiliated with Duolingo, Inc., it does not access your account nor read your progress.

Frequently Asked Questions

Has the EU banned Duolingo streaks?

No. The European Commission has proposed on November 11, 2026, within the Digital Fairness Act, that users should be able to turn off addictive design features —including streak incentives— without losing accumulated progress. It is a right to turn them off, not a ban, and it still needs to be approved.

When will I be able to turn off the streak without losing my progress?

Not immediately. The proposal must be negotiated with the European Parliament and the Council during 2026-2027, with a likely adoption by the end of 2027 or early 2028 and phased implementation between 2028 and 2030. The legal analyses published after the proposal place the realistic mandatory application in 2029 or later.

What is a dark pattern and why does the streak come into play here?

A dark pattern is an interface design that directs, deceives, or pressures the user into decisions they would not freely make. The streak falls within the perimeter because it operates on loss aversion: the psychological cost of breaking it is disproportionate compared to the satisfaction of maintaining it, and that turns daily return into an emotional obligation rather than a study decision.

Is learning worse with a streak?

The streak improves user retention, not necessarily learning. Duolingo's own experiments measure the probability of returning (+14% by day 7 with the streak bet), not retained vocabulary. And research on gamification warns that the effect of external incentives diminishes with exposure and can divert attention from the value of learning itself.

What should I replace it with to not abandon the language?

Three pieces with evidence behind them and one day of leeway per week: spaced review with your own deck (increasing intervals, not daily), grammar exercises with solutions on a separate sheet for later correction, and graded reading at your CEFR level with a glossary. The anchor stops being a counter and becomes the material you accumulate.

Why 98% coverage and not 95% in reading?

Because below 98% of known words, comprehension degrades and you can no longer infer the unknown from context, which is the mechanism by which extensive reading teaches vocabulary. Graded readers are calibrated by levels precisely to maintain that margin.

How much time do I need to level up a CEFR level?

Cambridge English estimates around 200 hours of guided work between consecutive levels, with about 350-400 accumulated hours to reach B1 and 500-600 for B2. That’s hundreds of hours, which is why consecutive days are a misleading metric of progress.

Can I use my material in Anki?

Yes, if the export comes in the format that Anki imports. A CSV with tab separator, two columns (front and back) and the note type and deck directives loads directly, without touching anything by hand.

Your language should not depend on a number

The Digital Fairness Act, proposed on November 11, 2026, will eventually give you a button to turn off the streak. But the method will not be given to you by any law: that is decided now, and it is decided by choosing material that stays with you when you close the app. aiDuolingo, from AiToolsForApps, is designed exactly for that part: to turn your word list, your weak grammar point, and your CEFR level into decks for Anki, grammar cards with solutions, and graded readings that you can download, print, and review on the day that suits you — without a counter, without guilt-inducing notifications, and without anything to start over from scratch.

Prepare your materials before the counter turns off: choose your CEFR level, your topic and your words, and download the Anki deck, the grammar sheet with solutions, and the graded reading with glossary. Try it without registering.

Prepare your materials before the counter turns off: choose your CEFR level, your topic and your words, and download the Anki deck, the grammar sheet with solutions, and the graded reading with glossary. Try it without registering.

↑ Back to top