Duolingo, Inc. (DUOL) - Deep Dive Research Report

Research Date: April 30, 2026 Concalls Covered: Q1 2025 (May 1, 2025) | Q2 2025 (Aug 6, 2025) | Q3 2025 (Nov 5, 2025) | Q4 2025 (Feb 26, 2026)

Section 1: What the Company Does

Duolingo is a mobile-first platform that teaches languages - and increasingly, other subjects like math, music, and chess - through short, gamified daily lessons delivered on a smartphone. The experience is free for anyone with a phone and an internet connection. The company makes money by selling premium subscription tiers that remove friction and add AI-powered coaching features, and to a lesser extent, from advertising shown to free users and from the Duolingo English Test, a university-recognized online language proficiency certification.

The founding of Duolingo is inseparable from understanding why it works. Luis von Ahn grew up in Guatemala during the civil war and watched firsthand how English fluency was a passport to opportunity - but learning English cost money that most people did not have. Von Ahn had already sold reCAPTCHA to Google and was a computer science professor at Carnegie Mellon when he and his graduate student Severin Hacker decided in 2011 to build a language learning product that would be "free, forever." Not free as a loss leader, but structurally free - where the business model would fund the free experience rather than the other way around.

The original business model was clever. Learners practiced translating sentences, and the translations they produced were sold to media companies like CNN and BuzzFeed, who needed content translated cheaply. Every lesson a learner completed simultaneously translated a piece of real-world text. Learning was the by-product; translation was the product being sold. This lasted until roughly 2016, when crowdsourced translation became economically uncompetitive and Duolingo pivoted fully to freemium - free with ads, with a paid tier for those who wanted the distraction-free, unlimited experience.

What makes Duolingo hard to build is not the language content itself. It is the combination of behavioral science, gamification, and content delivery that causes 50 million people to open the app every single day. The app uses streaks (consecutive daily sessions), experience points, a leaderboard system where learners compete with friends and strangers weekly, a cast of animated characters with distinct personalities anchored by Duo the Owl, and a carefully calibrated difficulty ramp that keeps lessons short enough to feel achievable but long enough to feel meaningful. The science behind this is not accidental - Duolingo has a dedicated team of learning scientists and runs hundreds of A/B experiments simultaneously to optimize every tap, every reward, every error message. The proprietary experimentation infrastructure is one of the company's least visible but most durable competitive assets.

In 2025, Duolingo crossed 50 million daily active users - more than five times the level when the company went public in July 2021. That number is not the number of people who downloaded the app. It is the number who open and use it every single day. For context, daily engagement at this scale in education is essentially unprecedented. The closest comparison is a social media feed, which is precisely how CEO Luis von Ahn describes the product ambition: as engaging as a mobile game, as effective as a one-on-one tutor.

The AI transformation underway since 2023 is not incremental. Von Ahn has publicly stated his belief that AI is now capable of replicating the effectiveness of a private language tutor - historically accessible only to the wealthy - and delivering it at zero marginal cost to every learner. That conviction is the engine behind the company's 2026 strategic pivot: deliberately slowing near-term monetization to reinvest in product quality, free user experience, and new subject expansion, betting that superior learning outcomes and broader reach will compound into a much larger monetizable audience within three to five years.

Section 2: Business Segments

Duolingo operates as a single integrated consumer platform but generates revenue through three meaningfully distinct streams: subscriptions, advertising, and the Duolingo English Test. Understanding each as its own economic unit explains both the company's current strategy and its financial trajectory.

Subscriptions - The Core Engine

Subscriptions account for approximately 76-80% of total revenue and are growing faster than the overall business. The subscription product exists in three layers:

Free (ad-supported) - The foundation of everything. Every user starts here. Free users see banner and interstitial ads and are subject to the energy system (formerly hearts), which limits how many wrong answers they can give before being slowed down. Free users drive the viral growth engine: the streak system, the leaderboards, and the social sharing mechanics all work for free users. Duolingo's willingness to keep the free product genuinely useful - rather than a degraded bait version - is a deliberate philosophy rooted in the founders' mission. It also creates the largest possible funnel for conversion.

Super Duolingo - The base paid tier. Removes all ads. Eliminates the energy constraint entirely. Adds offline access, unlimited mistake capacity, and a small set of personalization features. Available monthly or annually. The Family Plan, which bundles Super for up to six users under one subscription, reached 29% of total subscribers by Q3 2025, demonstrating that household adoption is a meaningful growth vector. Super is where most of the subscriber growth has happened over the past two years.

Duolingo Max - The premium AI tier launched in early 2023 with GPT-4 integration. Max includes all Super features plus three AI-powered capabilities: Roleplay (practicing conversations with AI characters in simulated real-world scenarios like ordering at a restaurant or navigating a job interview), Explain My Mistakes (GPT-4 explaining the grammatical or contextual reason behind every error in plain language), and Video Call with Lily (open-ended, real-time conversational practice with Duolingo's AI character Lily, who maintains continuity of context between sessions). Max reached 9% of total subscribers by Q3 2025, up from near-zero at launch. Bookings from Max doubled year-over-year in that period, though management has characterized performance as "below our lofty expectations," particularly for beginner users who struggled with the initial monolingual video call format. The strategic response - developing a bilingual version for beginners and testing migration of Video Call from Max to the Super tier - reflects ongoing calibration of where AI features sit in the monetization stack.

The subscription business has two structural advantages that compound over time. First, annual subscribers (a significant and growing share) reduce churn by extending the commitment horizon - a user who pays for a year has 12 months of daily habits to form before the cancellation question arises. Second, the family plan expands the economic unit from individual to household: a single conversion decision captures multiple potential learners, and the per-user economics are often better than individual subscriptions because household retention is higher.

Advertising - The Hidden Subsidy to Growth

Advertising contributes roughly 6-7% of total revenue. Free users, who represent the vast majority of the 130+ million monthly active users, see ads between lessons. The advertising business has historically been sold programmatically, but the Q4 2025 strategic pivot included a stated intention to expand direct advertising sales - a move that would improve economics (programmatic ad networks take significant cuts) and allow brand partnerships like the Luckin Coffee campaign in China to be systematized at scale.

The advertising segment matters not for its current size but for its strategic role: it is the economic justification for keeping the free product excellent. Without ad revenue from free users, the cost of serving those users (infrastructure, content) would fall entirely on subscribers. Because free users generate some ad revenue, Duolingo can afford to give them a high-quality product, which is what drives the word-of-mouth growth that keeps user acquisition costs low.

Duolingo English Test (DET) - The Certification Business

The DET is a fully online, AI-proctored English language proficiency exam that costs approximately $65 per test, takes about one hour, and produces results within 48 hours. It competes directly with TOEFL and IELTS, which cost $200-300, require in-person testing at licensed centers, and take weeks to return results. Over 6,000 programs at institutions across the world now accept the DET, including 95% of U.S. News Top 100 universities, all eight Ivy League schools, and a large share of UK, Canadian, and Australian universities.

The DET is a genuinely differentiated product. It is cheaper, faster, and more accessible than alternatives - particularly valuable for learners in markets without dense TOEFL/IELTS testing infrastructure. The AI proctoring (camera monitoring, browser lockdown, response authenticity scoring) addresses integrity concerns that initially made universities skeptical. The 94% effectiveness rating from admissions officers signals institutional confidence has largely been established.

In Q2 2025, the DET underperformed expectations due to macroeconomic headwinds: declining international student visa approvals to the US and UK reduced the pool of test-takers. This is an exogenous demand dependency that management has acknowledged explicitly. The DET's long-term value is tied partly to immigration and education policy environments outside Duolingo's control.

The DET's revenue contribution is roughly 15-18% of total revenue. It operates as a distinct product sold through a separate website (englishtest.duolingo.com) with its own marketing, institutional sales team, and accreditation infrastructure. It sits in the company's financials as "other" revenue alongside in-app purchases.

Section 3: Products and Business Detail

Language Learning - The Core Product

Duolingo offers over 40 languages for English speakers, and courses in dozens of native language pairings (e.g., French for Spanish speakers). Total course count exceeds 100 pairs. Languages range from Spanish (the largest user base in the US), French, German, Japanese, Korean, Mandarin, and Portuguese to smaller-demand languages including Irish, Navajo, Klingon, and High Valyrian - the latter two serving brand-building and community purposes more than learning utility.

The curriculum progresses through a "path" metaphor - a linear series of nodes representing skills, grammar concepts, and vocabulary clusters. Each node contains multiple short lessons (3-7 minutes each) of increasing difficulty. The lesson types include fill-in-the-blank, translation in both directions, listening comprehension, speaking exercises (assessed via voice recognition), and image-matching. The progression system uses spaced repetition principles - skills fade over time and require review to stay active, which drives daily re-engagement.

By Q1 2025, Duolingo had added 148 new language courses in approximately one year, using generative AI for content creation. Before AI automation, producing a single new course required 12+ years of development time. The acceleration in content availability is not merely cosmetic - it means learners in less-served language markets (smaller language pairs, regional languages) now have access to structured learning that simply did not exist before.

Advanced content, a long-standing gap in Duolingo's curriculum, is being addressed systematically. As of Q4 2025, courses in the nine largest languages were rolling out content up to Duolingo Score 129 - roughly equivalent to CEFR B2 (upper-intermediate), which is the threshold for employment qualification in most international professional contexts. This is significant: Duolingo's historical criticism was that it worked for beginners but abandoned intermediate learners. The advanced content initiative directly addresses the ceiling problem.

Duolingo Max - The AI Tier

As described in the segments section, Max adds Roleplay, Explain My Mistakes, and Video Call with Lily. The technical infrastructure behind these features is a direct integration with OpenAI's GPT-4. Lily, the AI character used in video calls, maintains a memory of prior conversations within a session and across sessions, allowing the conversation to feel progressive rather than reset. The 3D update to Lily (tested in A/B experiments during Q1 2025) made the character more expressive and lifelike, with the intended effect of increasing emotional engagement and conversion.

Video Call with Lily is the most technically complex and expensive Max feature. It uses real-time voice recognition, natural language understanding, response generation, and text-to-speech synthesis in sequence, with sub-second latency requirements to feel conversational. The initial version required learners to speak only in their target language (monolingual), which was a barrier for beginners. The bilingual format - where Lily occasionally speaks the learner's native language to provide scaffolding - is in development and intended to address the beginner dropout problem that management has cited across multiple calls.

By January 2026, "Explain My Mistakes" was made free for all users - a deliberate move to reduce friction at the top of the funnel and expose more learners to the Max feature set without requiring an upgrade.

Math, Music, and Chess

These three subjects represent Duolingo's bet on becoming a broader learning platform rather than purely a language app. Each launched in different years (math in 2023, music in 2023, chess in June 2025) and all use the same gamification infrastructure as language learning.

Chess has been the standout. Launched on iOS in June 2025, it reached 7 million daily active users within one year - a faster ramp than either math or music at equivalent maturity. At Duocon 2025, Duolingo announced Android availability and an upcoming player-versus-player mode, signaling the company's commitment to building a full chess ecosystem, not just introductory lessons. Management described chess as already the second-largest chess application globally by DAUs, a remarkable feat given the entrenched position of Chess.com and Lichess.

Math is positioned as the longest-term opportunity. Von Ahn has cited approximately 1 billion math learners worldwide, primarily students following K-12 curricula. The full K-12 math curriculum is in development, and AI dramatically accelerates content creation - Von Ahn estimated in Q1 2025 that math content would "quadruple within approximately one quarter" due to AI-generated lesson production. Math follows the same path-based progressive curriculum as language learning.

Music teaches music theory and instrument basics (piano-focused) using the gamified lesson format. Uptake has been slower than chess, and management has committed to a full music course revamp.

The Energy System

In May 2025, Duolingo replaced the "hearts" system for free users with an "energy" system. Hearts penalized mistakes by removing a life - run out of hearts, and free users had to wait for regeneration or watch ads to refill. Energy works differently: users start with 25 units per session, correct answers earn small energy bonuses, and wrong answers cost 1-2 units. The net effect is similar - free users have a session cap - but the framing is positive (earning energy through success) rather than punitive (losing hearts through mistakes). The practical business impact: better engagement and revenue metrics (management confirmed increased DAUs, time spent, and revenue per user in Q2 2025) while reducing the psychological harshness that drove criticism and churn.

Geographic Footprint

Duolingo operates globally with meaningful user concentrations in the United States (largest absolute user base), Latin America (especially Mexico and Brazil), Europe, India, and increasingly China. China has been the fastest-growing country across both Q2 and Q3 2025, driven by high organic demand for English learning and amplified by the Luckin Coffee brand partnership campaign in July 2025. China represents 5-6% of total business and is the second-largest country by DAUs. Max remains unavailable in China due to Chinese AI regulations requiring government approval of locally-hosted large language models - approval that management has said is entirely outside their timeline control.

The LinkedIn integration announced at Duocon 2025 allows users to display their Duolingo Score - a standardized measure of language proficiency generated by the DET - directly on their LinkedIn profiles. This is a significant distribution move: it ties Duolingo's proficiency data to the world's largest professional network and creates a credential value proposition that motivates learning completion.

Section 4: Customers

Who Uses Duolingo

Duolingo's free product is used by an extraordinarily broad demographic - from high school students in the United States learning Spanish for a college application to office workers in Shanghai learning English for career advancement to retirees in Germany learning Italian before a vacation. The breadth is intentional and structural: the free tier imposes no demographic barrier.

The paying subscriber base skews toward adults (18-45) with disposable income and a clear motivation for learning - career advancement, travel, cultural connection, or academic requirements. English learners currently represent approximately 50% of Duolingo's user base, which is striking given that roughly 80% of all language learners globally are learning English. This gap represents an identified opportunity: non-English learners (who are majority-native-English-speakers learning other languages) are a large segment that is, for now, underrepresented on the platform.

Why They Buy

The decision to subscribe to Super Duolingo is rarely a high-consideration purchase. The trigger is typically reaching the frustration threshold of the free experience - hitting the energy/heart limit mid-lesson, seeing an ad at an inopportune moment, or losing a streak without offline access. The subscription pitch is ambient: it is presented every time the free experience creates friction. This low-consideration, friction-triggered conversion model means marketing spend is largely unnecessary to drive subscription - the product sells itself by temporarily worsening.

Max subscribers are a different buying decision. The motivation is usually explicit: a learner who has been using Super for a while, wants conversation practice, and is willing to pay a premium for the AI interaction. English learners show higher Max adoption than non-English learners (cited by management in Q1 2025), likely because English proficiency has higher instrumental value (jobs, academic credentials) that justifies premium spending.

DET test-takers buy for a specific purpose: to satisfy a university or visa requirement with something cheaper and faster than TOEFL or IELTS. The buying decision is often not between DET and nothing - it is between DET and TOEFL, with DET winning on price, convenience, and speed.

Switching Costs

For language learners, Duolingo's switching costs are primarily behavioral. A streak of 300 days represents genuine psychological investment that a user would lose by switching to a competitor. The XP history, the completed skill tree, the league placement - these are all sunk-cost anchors that make departure painful. They are not technical lock-in (there is no account export that a competitor recognizes), but behavioral lock-in is often more durable because it does not require contractual enforcement.

For the DET, switching costs are low at the individual level (a test-taker can simply take TOEFL instead) but high at the institutional level. Universities that have integrated DET into their admissions processes, verified its validity with their accreditors, and trained admissions staff on interpreting Duolingo Scores have invested meaningful process capital. The 6,000+ accepting institutions represent years of institutional sales work that competitors would need to replicate.

Concentration

No individual user or corporate account represents a material concentration of revenue. The business is structurally diversified across millions of individual consumers. The closest thing to customer concentration risk is geographic: the US market represents a disproportionate share of paid subscription revenue because American consumers pay at higher absolute dollar amounts than users in lower-GDP markets.

Section 5: Competitive Landscape

Direct Competitors in Language Learning Apps

Duolingo holds approximately 85% of language learning app daily active users globally. That number requires context: the entire language learning app category is not the entire language learning market. But within the app-centric, gamified mobile experience, Duolingo's share is dominant by a wide margin.

Babbel is the nearest direct competitor in the European market. Babbel charges upfront (no meaningful free tier) and positions on structured curriculum quality, with content developed entirely by human linguists. It targets adult learners willing to pay for a more academic experience. Babbel's model is the polar opposite of Duolingo's: it wins on perceived quality but loses on scale and discoverability. Babbel is private and does not disclose user metrics, but its scale is believed to be substantially smaller.

Busuu sits in a similar position: paid-first, community-enhanced (native speaker feedback on written and spoken exercises). Busuu was acquired by Chegg and has been losing relevance as Duolingo's AI features extend into the human-feedback territory Busuu once owned exclusively.

Rosetta Stone represents the legacy incumbent - brand recognition from its physical CD-ROM era, now fully digital, primarily sold to enterprises and educational institutions rather than consumers. Its consumer app is a fraction of what it once was. In the direct consumer segment, Rosetta Stone is no longer a competitive threat.

Lingoda and Preply and iTalki operate in the live tutoring space - human instructors for hire, structured classes, scheduled sessions. These are genuinely different products: higher quality at higher price points, requiring scheduling and human availability. They are not competitors in the same usage moment as Duolingo (a 5-minute commute lesson vs. a scheduled 50-minute class). However, as Duolingo's Video Call with Lily improves, it increasingly competes with the value proposition of a live tutor at dramatically lower cost.

ChatGPT and AI-native competitors are the structural threat that analysts and management engage with most directly. Any large language model can be prompted to have a conversation in any language, correct grammar, explain mistakes, and simulate roleplay scenarios. What AI general-purpose tools cannot replicate, currently, is: the gamification and habit-formation infrastructure (streaks, leagues, characters, adaptive curriculum), the structured progression from beginner to advanced, and the behavioral science layer that causes millions of people to open the app habitually. Von Ahn has addressed this multiple times across concalls, arguing that the engagement data (50M+ daily users) demonstrates that the motivation architecture is the real product, not the language content, and that this architecture is what AI chatbots do not have. The risk is not zero, but it is more nuanced than "ChatGPT will replace Duolingo."

In the English Test Market

The DET competes with TOEFL (administered by ETS) and IELTS (administered by British Council and IDP Education). TOEFL and IELTS have 50+ years of institutional trust, global testing center infrastructure, and are required by far more programs globally. DET has disrupted the market for convenience-sensitive test-takers and programs willing to trade some institutional prestige for accessibility. The macro risk is that international student visa policy (US and UK most acutely) has been tightening, which reduces the test-taking population independent of which test they prefer.

Barriers to Entry

The barriers to building a Duolingo competitor are real but not primarily technical. Content is now cheap to produce (AI). The mobile distribution channel (App Store, Play Store) is open. The barriers are behavioral: it takes years to build a habit-forming product with meaningful engagement data across millions of users, to tune the gamification to drive daily retention, and to establish the brand recognition that makes a new learner choose you over an app they've already heard of. Duolingo's streak database - representing hundreds of millions of learning sessions - is also a training asset for improving adaptive curriculum algorithms that no new entrant can replicate without time.

Section 6: Industry

What Drives Demand

Demand for language learning is driven by four overlapping forces: globalization and cross-border labor mobility, immigration and academic credential requirements, consumer curiosity (travel, cultural connection, personal enrichment), and AI-era reskilling anxiety. The last driver is newer and still developing: as automation displaces knowledge work, workers are reassessing which human skills are worth developing, and language is near the top of that list.

English is the dominant learning language globally by volume. Asia-Pacific - led by China (300M+ English learners), India (250M+ English learners), Japan, South Korea, and Southeast Asia - is the largest and fastest-growing demand region. Latin America (Spanish-English bidirectional learning) is the second major region. Europe exhibits strong demand for secondary European languages (German, Spanish, French, Italian) alongside English.

Market Size

The online language learning market was valued at approximately $21 billion in 2025 and is projected to grow at roughly 15-16% CAGR through 2031, reaching approximately $51 billion. The broader digital language learning market (including enterprise training, K-12 institutional, and professional certification) is estimated at $26-85 billion depending on scope and methodology, with projections to $116+ billion by 2033 across the widest definition. Duolingo operates in the consumer-facing segment of this market, which is meaningfully smaller than the institutional segment but growing faster due to mobile distribution.

Language Testing

The English language testing market is dominated by IELTS and TOEFL, which together serve millions of test-takers annually. The DET has disrupted the market at the margins - taking share from candidates who previously would have taken TOEFL primarily because of convenience advantages. The testing market is fundamentally tied to international student enrollment, which is a function of immigration policy, foreign exchange rates, and geopolitical relationships between student-sending countries (China, India, Korea, Nigeria) and destination countries (US, UK, Canada, Australia).

Regulatory Environment

Language learning apps themselves face minimal direct regulation. The DET is subject to institutional accreditation dynamics - universities choose to accept or reject it based on their own policies, influenced by accreditor guidance. In China, Duolingo's Max tier faces specific regulatory friction: the requirement to use locally-approved large language models for AI features (a consequence of China's AI regulatory framework) has blocked Max availability in a market that is Duolingo's second-largest by DAUs.

App store policies represent a form of implicit regulation. Apple's 30% commission on in-app purchases is a structural cost that Duolingo, like all app developers, has worked to partially circumvent through direct web purchase flows - an avenue that became more accessible following EU Digital Markets Act requirements and subsequent global regulatory pressure on Apple. In Q2 2025, management reported testing web purchase flows on iOS with "minimal bookings loss but significant profit improvement," reflecting the potential of this avenue.

Cyclicality

Consumer education spending is relatively non-cyclical at the individual subscription level ($10-15/month is not a meaningful household budget line). However, the DET business does show cyclical sensitivity to macroeconomic and geopolitical factors - specifically to international student flows, which are highly sensitive to visa policy, exchange rates, and bilateral relationship dynamics. In 2024-2025, tightening US and UK immigration policy created a measurable headwind for DET volume that management acknowledged in Q2 2025.

Section 7: Growth Triggers

All triggers sourced directly from earnings call statements.

Target of 100 million daily active users by 2028, up from 50M+ in 2025

"We believe that reaching 100 million daily active users by 2028 is the right goal." - Luis von Ahn, Q4 2025 concall (Feb 26, 2026) This was framed as the company's organizing north star for the 2026 investment cycle. (Q4 2025, Feb 26, 2026)

Speaking Adventures - new gamified conversational feature for all users, launching mid-2026 Management described Speaking Adventures as a feature that brings conversational practice out of the Max tier and into the core product experience for all users, designed as gameplay rather than a formal tutoring interaction. Expected mid-2026 rollout. (Q4 2025, Feb 26, 2026)

Video Call with Lily migration from Max to Super tier The company is A/B testing moving Video Call - previously a Max-exclusive feature - down to the Super tier. If the economics work (cost per call decreasing as AI model costs fall), this would meaningfully increase the value proposition of Super subscriptions and potentially accelerate Super subscriber growth. (Q4 2025, Feb 26, 2026; first flagged in Q3 2025, Nov 5, 2025)

Advanced content rollout to CEFR B2 level (Duolingo Score 129) for nine major languages

"Rolling out within weeks" - Luis von Ahn, Q4 2025 concall (Feb 26, 2026). Content at employment-qualification level addresses the longstanding ceiling criticism and retains intermediate/advanced learners who previously churned. (Q4 2025, Feb 26, 2026; flagged across Q1, Q2, Q3 2025 as in progress)

Chess reaching second-largest chess application globally, PvP mode launching Chess hit 7 million DAUs within its first year, surpassing math and music growth trajectories. Duocon 2025 announced PvP (player-versus-player) mode on iOS, with Android support to follow. PvP increases the social and competitive loop that drives retention, applying the same leaderboard mechanics that made language leagues effective. (Q4 2025, Feb 26, 2026; Q3 2025, Nov 5, 2025; Duocon 2025, Sep 16, 2025)

K-12 math curriculum expansion - targeting 1 billion math learners globally

"I believe that this year we're gonna have the best tutor app for math...there's about 1 billion people learning math in the world." - Luis von Ahn, Q4 2025 concall (Feb 26, 2026). AI-generated math content was described as quadrupling within a single quarter (Q1 2025, May 1, 2025). Full K-12 curriculum in active development. (Q4 2025, Q3 2025, Q2 2025, Q1 2025 - repeated across all four calls)

China market: Max subscription approval expected but timeline unknown Duolingo has selected a local LLM partner and submitted for regulatory approval in China. Approval would unlock Max in a market that is already Duolingo's second-largest by DAUs and fastest-growing country. Management has been explicit that the timeline is not in their control. (Q3 2025, Nov 5, 2025; Q2 2025, Aug 6, 2025)

Direct advertising sales expansion - reducing dependence on programmatic ad networks The Q4 2025 strategic pivot explicitly named direct ad sales (brand partnerships on the Luckin Coffee model) as a monetization avenue that improves economics for free-user ad revenue without requiring user upgrades. (Q4 2025, Feb 26, 2026)

Bilingual Video Call format for beginner users - expected to meaningfully improve Max adoption among beginners The current monolingual format (learner must speak only in target language) proved too intimidating for beginners, who showed lower Max adoption and video call utilization. A bilingual scaffolded version is in development. (Q3 2025, Nov 5, 2025; Q2 2025, Aug 6, 2025)

LinkedIn integration of Duolingo Score - creating credential value for completed learning Users can now display Duolingo language proficiency scores directly on LinkedIn profiles. This creates an external, professional-context incentive for language learning completion and DET adoption. Announced at Duocon 2025 (Sep 16, 2025)

$400 million share buyback program authorized Authorized February 2026, no expiration date. Intended to offset stock-based compensation dilution and return capital at what management views as depressed valuations. (Q4 2025, Feb 26, 2026)

Trigger	Timeline	Concall Source	Status
100M DAU target	2028	Q4 2025 (Feb 26, 2026)	New announcement
Speaking Adventures	Mid-2026	Q4 2025 (Feb 26, 2026)	New
Video Call to Super tier	2026 (A/B test)	Q4 & Q3 2025	Repeated
Advanced content to B2 level	Weeks from Feb 26, 2026	Q4 2025 (Feb 26, 2026)	Repeated (4 calls)
Chess PvP mode	2026	Q4 & Q3 2025, Duocon	Repeated
K-12 math curriculum	2026 ongoing	All 4 calls	Repeated
China Max approval	Unknown	Q2 & Q3 2025	Pending, repeated
Direct ad sales	2026	Q4 2025 (Feb 26, 2026)	New priority
Bilingual video calls	2026	Q2 & Q3 2025	Repeated
LinkedIn integration	Active (Sep 2025)	Duocon 2025	Delivered
$400M buyback	Active (Feb 2026)	Q4 2025 (Feb 26, 2026)	New

Section 8: Key Risks

1. DAU Growth Deceleration - The Number That Defines the Narrative

The mechanism: Duolingo's valuation has historically been underwritten by extraordinary user growth - from 2022 to mid-2025, no quarter saw year-over-year DAU growth below 40%. Growth decelerated throughout 2025, ending at approximately 30% year-over-year in Q4 2025. Management has guided 20% year-over-year DAU growth for all of 2026. The risk is that 20% proves optimistic - that the combination of a very large installed base (50M+ DAUs creates tougher comps), slowing new user acquisition in saturated markets (US growth has been explicitly described as decelerating and below global averages), and reduced viral campaign activity (intentionally scaled back in Q2 2025) produces DAU growth below 20%. If DAU growth misses to the downside in a year where management has explicitly traded near-term profitability for future user growth, the credibility of the entire 2026 investment case is in question.

This is a high-probability moderate drag, not a catastrophic risk, but it is the risk with the most direct connection to market confidence.

2. AI Disruption of the Core Learning Model

The mechanism: Large language models (GPT-4o, Gemini, Claude) can today provide on-demand, conversational language practice in any language with immediate feedback, grammar explanation, and tone adjustment - for free, as part of a subscription consumers may already own. If a meaningful cohort of language learners discovers that a ChatGPT conversation is as effective as Duolingo lessons for their purposes, the free-user funnel could narrow. If AI tutors become more engaging and habit-forming (through integrations into messaging apps, voice assistants, etc.), the behavioral moat of Duolingo's streak/league system could erode.

Management's counter-argument - that engagement data demonstrates habit architecture is the real product, not language content - is credible but not permanent. A well-designed AI habit-forming product (which several startups are building) could replicate the engagement layer over time. This is a lower-probability but higher-magnitude risk that plays out over a 3-5 year horizon.

3. Monetization Gap - The 90% Who Don't Pay

The mechanism: 90% of monthly active users do not pay for any subscription tier. The Q4 2025 strategic pivot is deliberately widening this gap in the near term - improving the free experience reduces the friction triggers that convert free users to subscribers. The bet is that a better free product grows the overall MAU base faster, and that the larger base eventually produces more subscribers in absolute terms. But the path from "better free experience" to "more subscribers" is not guaranteed, and the timeline is uncertain. If improving the free product does not accelerate user growth sufficiently, or if user growth does not convert to subscriptions at an adequate rate, the company will have sacrificed near-term financial performance for no long-term gain.

4. China Regulatory Risk

The mechanism: China is Duolingo's second-largest market by DAUs and fastest-growing country. Max - the highest-margin subscription tier - is unavailable in China because the regulatory approval of the required local LLM has not been granted. If approval is delayed indefinitely, Duolingo leaves significant monetization on the table in its highest-growth geography. More acutely, any deterioration in US-China relations that creates friction for American tech companies operating in China (as has happened repeatedly with other US consumer apps) could threaten Duolingo's Chinese user base entirely.

5. Platform Dependency - Apple's 30% Tax

The mechanism: The majority of Duolingo's subscription revenue flows through Apple's App Store, which takes a 30% commission. The company has been testing direct web purchase flows (reducing the commission to approximately 2% through payment processors) but has not yet fully deployed this capability. Until web purchasing is available at scale for all subscription types, Apple's commission represents a structural cap on gross margins. Any adverse policy change by Apple (restricting web purchase flows, closing regulatory loopholes) would reverse the margin improvement Duolingo is working toward.

6. DET - International Student Flow Dependency

The mechanism: DET revenue is directly tied to the volume of international students applying to English-language universities. Tightening US and UK visa policies in 2024-2025 reduced this volume. Further tightening (as has been politically active in both countries) would reduce DET test volumes and the associated revenue. DET pricing power is limited by its competitive positioning as the "affordable" alternative to TOEFL - raising prices would undercut the primary differentiation.

7. Stock-Based Compensation Dilution

The mechanism: Duolingo awards substantial equity to employees. Management has cited net dilution of approximately 2% per year after offsetting with buybacks. The newly authorized $400M buyback is intended to manage this, but if the stock remains at depressed levels (shares were approximately 70% below their 52-week high of $544.93 at the time of the Q4 2025 earnings call), buybacks consume more cash per share retired, reducing the financial flexibility to invest in growth.

Section 9: Walk the Talk

Concalls used: Q1 2025 (May 1, 2025) | Q2 2025 (Aug 6, 2025) | Q3 2025 (Nov 5, 2025) | Q4 2025 (Feb 26, 2026) The most recent concall (Feb 26, 2026) is within 63 days of today, satisfying the freshness requirement.

Q1 2025 - Confidence and Momentum

Coming into 2025, management projected an era of sustained growth. Von Ahn described 49% year-over-year DAU growth in Q1 2025 as evidence of compounding momentum. The narrative was optimistic but grounded in real numbers: Max had reached 7% of subscribers, chess was about to launch, and 148 new language courses had been added in one year via AI. Management reaffirmed full-year 2025 guidance and described Q2 as showing no macro headwinds so far. The tone was: this machine is working, we are accelerating.

The specific promise worth tracking: management stated that gross margins would decline only 150 basis points year-over-year for full-year 2025, less than the feared 300 basis points, because AI API costs were falling faster than expected. This was delivered - Q2 2025 beat on gross margins specifically because AI infrastructure costs came in below expectations.

Q2 2025 - The First Headwind

The Q2 2025 call introduced the first notable deviation from the established growth narrative. DAU growth came in at 40% year-over-year - at the low end of the 40-45% guidance range, and representing a deceleration from Q1's 49%. Management's explanation: a deliberate reduction in viral social media campaigns in North America, combined with the "Dead Duo" campaign in Q1 having pulled forward some engagement.

The explanation was credible but raised a question that became more important in retrospect: if DAU growth is sensitive to campaign intensity and management is choosing to reduce campaigns, what does that say about the organic growth rate? Management also disclosed that Max adoption was "underperforming lofty expectations," particularly for video calls. Simultaneously, they raised full-year guidance (bookings and EBITDA) and reported better-than-expected gross margins. The message: financially excellent, user growth softer than peak but still strong, AI features need iteration.

The web purchase testing was a genuine piece of forward progress reported here - "minimal bookings loss but significant profit improvement" when bypassing the App Store - a specific, falsifiable claim that has not been retracted.

Q3 2025 - The Strategic Pivot Signal

Q3 2025 is the most consequential of the four calls for understanding management's decision-making. DAU growth continued decelerating to 36% year-over-year, below both Q1 and Q2. Management raised full-year bookings guidance to nearly $1.2 billion (33% growth) and maintained the 29% adjusted EBITDA margin, which management described as "very, very close" to the company's long-term target range. By financial measures, 2025 was on track to be a strong year.

But in the same call, management revealed a strategic shift: they had "over the last month or two" begun reallocating resources away from monetization-driving work toward user growth and teaching effectiveness. CFO Matt Skaruppa characterized the financial impact as "relatively small," and the pivot was not framed as a crisis response but as a deliberate, opportunistic move triggered by conviction in AI's transformational potential.

In retrospect, this call contained the seed of the bigger announcement three months later. The pivot was being executed in Q3 while management communicated it as a minor adjustment. Whether this represented appropriate management discretion (announcing a strategic shift before you know if it works would be destabilizing) or a gradual revelation of a larger problem (DAU deceleration requiring structural response) is the central management credibility question.

Q4 2025 - Full Revelation and Reset

The Q4 2025 call crystallized what had been signaled in Q3. The company beat on financial metrics for the year (over $1B in bookings, over $300M in adjusted EBITDA, 52.7M DAUs, 30% year-over-year growth), but simultaneously announced 2026 guidance that dramatically disappointed expectations: 10-12% bookings growth, 15-18% revenue growth, and a 25% EBITDA margin - down from 29.5% in 2025. The strategic rationale was explicit: Duolingo is deliberately choosing near-term financial performance moderation to prioritize user growth, with the target of 100 million DAUs by 2028.

A new CFO (Gillian Munson, replacing Matt Skaruppa) appeared for the first time, adding a leadership transition dynamic to an already consequential call. Munson's framing: "We're really motivated to go for the bigger prize, which is much more aligned with what Luis and Severin were thinking about when they founded the company."

The stock's reaction tells part of the story: despite beating Q4 expectations, shares were already approximately 70% below their 52-week high, having declined throughout 2025 as DAU growth decelerated. The stock rose 5% in after-hours on the Q4 beat but remains dramatically below peak valuations.

Management Credibility Assessment

On financial delivery, Duolingo's management has a strong record. Every financial metric guidance range in 2025 was either met or beaten. Gross margin, EBITDA margin, bookings, and revenue all came in at or above stated expectations. When they said AI costs would be better than feared, they were right. When they said Max subscriber count would grow, it did.

On DAU and user growth, the record is more nuanced. The deceleration from 49% in Q1 2025 to 36% in Q3 to 30% in Q4, with 20% guided for 2026, was not clearly foreshadowed. Management's Q1 2025 tone did not suggest that within nine months they would be guiding for DAU growth of one-third the current rate and explicitly trading profitability for user growth. The Q3 "small adjustment" that became the Q4 "strategic pivot" feels like a gradual disclosure rather than a sudden decision - which raises the question of whether management was managing the communication timeline rather than the business timeline.

The conclusion: this is management that delivers on financial commitments but has been less transparent about the slowing of the core engagement flywheel that drives everything downstream. They are not misleading, but they are optimistic communicators who have allowed the market to discover bad news gradually rather than directly. The $400M buyback and the explicit 100M DAU target by 2028 are bold public commitments that will now be tracked closely.

Section 10: Shareholder Friendliness Index

Dividends

Duolingo has never paid a dividend. There is no dividend history on record, and no indication from management across any of the four concalls that a dividend policy is under consideration. For a company with a growth-stage reinvestment philosophy and stock-based compensation as the primary employee compensation vehicle, the absence of dividends is not a surprise. This is standard for consumer technology companies at Duolingo's stage. There is no dividend to evaluate.

Share Buybacks

2023: No buyback program. The company was in a phase of heavy reinvestment and had not yet reached the profitability scale to justify returning capital.

2024: No buyback program. Management was preparing for the profitability inflection point.

2025: No buyback program in place during 2025, but the financial performance of the year (>$300M adjusted EBITDA, $360M+ free cash flow) created the capacity.

February 2026: The board authorized a $400 million share repurchase program with no expiration date, announced alongside Q4 2025 results. Management stated the program is intended primarily to offset stock-based compensation dilution. Trailing three-year average net dilution from stock grants has been approximately 2% per year. At 2026 share prices (approximately $113-114 per share at the time of announcement), $400M would retire approximately 3.5 million shares, which represents approximately 7.5% of the approximately 47 million shares outstanding.

The buyback is the first formal return-of-capital program in Duolingo's public company history. Its authorization at a point when shares are approximately 70% below their 52-week high and management is simultaneously guiding for lower near-term profitability suggests the buyback serves two simultaneous purposes: genuine capital return and a signal of management confidence in intrinsic value during a communication-sensitive period.

The net dilution picture over three years: Duolingo's stock-based compensation has been significant (consistent with high-talent-intensity technology companies), but management's statement that net dilution has run at approximately 2% per year indicates they have been repurchasing some shares (or retiring shares through settlement of vested equity) to partially offset gross issuance. The formal buyback authorization converts this from an informal practice to a disclosed, board-approved program.

Verdict: Duolingo is not shareholder-hostile, but it is unambiguously growth-oriented. Capital is reinvested in product development, AI infrastructure, and geographic expansion rather than returned to shareholders. The first buyback program is a meaningful milestone but does not change the fundamental capital allocation philosophy. Shareholders who bought this stock bought a growth company, and management's behavior has been consistent with that positioning.

Section 11: Scenarios

Bull Case - The Platform Expands and AI Unlocks True Tutoring

In the bull scenario, Duolingo's conviction that AI can replicate the quality of a private tutor turns out to be right, and it turns out to be right faster than the market currently prices. Video Call with Lily, upgraded to bilingual format and progressively cheaper to run as AI inference costs fall, achieves retention and satisfaction metrics that justify migration to the Super tier. Super subscriptions grow not just because the free experience worsens, but because the paid experience becomes meaningfully more valuable. Max continues to grow as advanced learners discover that AI-powered conversation practice accelerates real-world proficiency in ways that lesson-based learning alone cannot.

Math becomes a genuine second pillar. The K-12 curriculum completes, Duolingo's brand in education extends beyond languages, and the 1 billion global math learners begin discovering the product through the same word-of-mouth channels that built the language franchise. Chess, already the world's second-largest chess app by DAUs within its first year, builds a PvP community that retains users at social-game engagement levels. China's regulatory approval arrives for Max, unlocking the premium tier in the company's fastest-growing market and providing a bookings inflection point in 2027.

In this world, 100 million DAUs by 2028 is achievable, the pivot year (2026) looks prescient in retrospect, and the company exits the investment cycle with a monetizable user base more than double its 2025 size. The 10% of MAUs who currently pay becomes 12-15% as AI feature quality increases the value of paying.

Base Case - Controlled Execution on a Narrower Path

In the base scenario, management delivers roughly what it has guided. DAU growth holds at approximately 20% in 2026 - not reaccelerating meaningfully but not collapsing further. The free experience improvements increase word-of-mouth slightly in markets where Duolingo is already established, and China and Asia-Pacific continue growing faster than the global average. Chess PvP proves engaging but does not replicate the scale of language learning; math makes progress but requires another 2-3 years before representing a meaningful revenue contributor.

The Video Call migration to Super happens in second-half 2026, lifts Super subscriber growth modestly, and improves the value-per-subscriber metric. Max adoption grows slowly as AI inference costs fall and bilingual features launch, but does not achieve the dramatic penetration management had originally hoped for. DET stabilizes as the visa environment stops worsening. By 2027, DAU growth reaccelerates to 25-30% as the 2026 investments in free experience and new subjects begin bearing fruit, and bookings growth returns to 20%+ territory. The 100M DAU target becomes achievable by 2028-2029 range.

The buyback makes a modest difference to per-share metrics. Management turnover (new CFO) is non-disruptive.

Bear Case - The Flywheel Slows and the Bet Doesn't Pay

In the bear scenario, the 2026 pivot underdelivers on user growth without recovering monetization. DAU growth in 2026 comes in at 15% or below as competition from AI-native learning tools intensifies, the US market saturates for the core demographics, and the improvements to the free experience are not compelling enough to generate word-of-mouth at scale. Simultaneously, reducing monetization friction means fewer free-to-paid conversions without a compensating increase in total subscriber base. Bookings growth comes in at the low end of guidance or misses.

China's Max approval is delayed beyond 2026, removing the most identifiable bookings catalyst. The DET continues facing international student headwinds as immigration policies remain tight. The $400M buyback burns cash that could have been used for product investment or acquisition, without meaningfully supporting the stock.

In this world, the 2026 "pivot year" becomes a permanent step-down: a company that sacrificed financial performance and did not get the user growth it needed to justify the sacrifice. The market narrative shifts from "temporary growth investment" to "structural deceleration," and the multiple compression that already happened in 2025 (70% decline from peak) continues. The bear case is not bankruptcy or collapse - Duolingo generates real cash flow and has a genuine product - but it is a multi-year period where the company is valued as a deceleration story rather than a growth story, and where the 100M DAU target slips from 2028 to 2030+.

Sources: