How Does Duolingo Make Money? Inside Its Freemium Model

Duolingo makes money mainly from subscriptions. Its full curriculum is free with ads and usage limits, and a small share of users pay for Super Duolingo or the AI-powered Duolingo Max to remove ads, get unlimited mistakes and unlock extra features. Subscriptions were roughly 80% or more of revenue in 2025, with advertising a high-single-digit share and the Duolingo English Test a mid-single-digit share (approximate, from filings).

The model works at a scale few consumer learning companies have reached. Duolingo reported revenue of $1,037.6 million in 2025, up 39%, and about 58.7 million daily active users in the second quarter of 2026. It raised only about $183 million before going public in July 2021, and it has not needed to raise since. The reason is the number that underpins its entire model: it spends very little to acquire each user.

What does Duolingo sell?

Duolingo has four revenue lines, one of which dominates.

US list price for Super has been about $12.99 a month or about $84–$90 a year, with heavy discounting through trials and promotions. Max launched at roughly double Super, about $168 a year (reported). Annual plans dominate, so bookings run ahead of recognized revenue. In 2025 bookings were $1,158.4 million against revenue of $1,037.6 million.

How did Duolingo find a business model?

It took six years. Luis von Ahn, the Carnegie Mellon professor who co-invented reCAPTCHA, and his student Severin Hacker founded Duolingo in Pittsburgh in 2011. The original plan was to let people learn for free by translating real web content and to sell the translations. CNN and BuzzFeed signed early deals, but buyers were few and machine translation was improving fast. By 2014–2015 the translation model was quietly retired.

Duolingo then tried three things: a credentialing product (the English test, 2014), advertising (2016) and subscriptions (Duolingo Plus, 2017). The subscription won. What survived from the founding vision was a promise that learning on Duolingo would always be free.

That constraint shaped the paywall. Duolingo sells convenience, not content. The book describes the choice as the one that "preserved both the mission and the free funnel." Competitors such as Babbel and Rosetta Stone give a few lessons and then require payment. A hard paywall converts more of the people who reach it, but far fewer people reach it.

Why does Duolingo's growth model work?

Duolingo's defining economic fact is a very low customer acquisition cost. It has historically spent about 10% or less of revenue on paid marketing and grows mostly through word of mouth, app-store ranking and social media.

That matters because freemium conversion is low by design. A stylized example in the book shows what happens to a learning app that buys its users: installs at $2 each, 2% of installs converting to paid, means $100 to acquire each subscriber. At $9 a month of net revenue and 6% monthly churn, each subscriber is worth about $127 over a lifetime, barely more than its cost. A few percent of conversion is enough only when acquisition is nearly free. Duolingo's is.

The retention engine

The second part of the engine is retention. Around 2018 Duolingo's growth team, led by Jorge Mazal, built a model of user states and found that the retention rate of current users had by far the largest effect on daily active users. The company made daily active users its north-star metric and built leagues, streak protections and a smarter notification system.

Daily users rose roughly 4.5 times over about four years of that program (reported by Mazal). The ratio of daily to monthly users rose from about 20% to above 35% by 2025 (reported). The paid share of monthly users rose from about 3% in 2020 to about 8–9% by 2025. Each of those numbers sits at the "strong" end of the practitioner heuristics in our edtech business model guide.

The brand as a marketing budget

Duolingo's owl mascot, Duo, became a meme. From 2021 a very small in-house team ran its TikTok account, and stunts such as a 2024 Super Bowl ad and the February 2025 "death of Duo" campaign were credited by management with spikes in daily users. In a consumer business with low conversion, the free users are the marketing budget, and brand is how that budget compounds.

What are Duolingo's financials?

Gross margin ran in the low-to-mid 70s through 2023 and has drifted toward about 70% as AI inference costs for Max and Video Call grew. Duolingo cited AI costs as one reason its 2025 gross margin dipped to 72.2% from 72.8%. Adjusted EBITDA margin rose from negative in 2020 to about 29.5% in 2025.

How is AI changing Duolingo's business?

AI cut Duolingo's costs and raised its risks at the same time.

On the cost side, generative AI made exercises, translations, audio and whole courses far cheaper to produce. Duolingo cut about 10% of its contractors in January 2024 as AI took over content tasks, and on 30 April 2025 it announced 148 new AI-built courses.

On the risk side, three things went wrong in 2025. A chief executive memo on 28 April declaring the company "AI-first" drew a public backlash, including a wave of "I deleted Duolingo" posts. Free general-purpose AI assistants and voice tutors such as Speak began to compete directly for speaking practice, Duolingo's long-standing weak spot. And inference became a real cost of goods: a heavy user of a flat-priced Max plan can cost more than they pay.

The shares reflected both sides. DUOL rose roughly eightfold from its 2022 low to a May 2025 peak above about $500 a share, with a market value of about $23–25 billion (reported). It then fell by roughly two-thirds by mid-2026 (reported).

Why did Duolingo choose users over monetization in 2026?

By late 2025 daily-user growth had slowed from about 50% or more in 2024 to about 30%. Management concluded that years of monetization experiments, brand damage from the AI memo and the rise of free chatbots risked hollowing out the free base that feeds the business.

On 26 February 2026 Duolingo guided to bookings growth of 10–12% for 2026 and said a monetization-focused plan could have produced about 20%. Shares fell about 18% on the guidance (reported). The early results were mixed. In the second quarter of 2026 daily users grew 23% to about 58.7 million and paid subscribers grew about 17% to about 12.7 million, but bookings grew only 8%, to $289.1 million. The company's stated goal is 100 million daily users by 2028. Founder control through dual-class shares made the unpopular reset possible.

The book's reading is that the experimentation machine that built the habit "can borrow from future growth when pointed at monetisation." Duolingo's 2026 bet is that giving back some of that borrowing now protects the funnel later.

Which factor of the learning equation is Duolingo's weak point?

Duolingo is the strongest case in the sector for the third factor of the learning equation: an affordable route to the payer. Near-zero acquisition cost is what makes its low conversion rate profitable. The second factor, a willing payer, is proven by subscriptions that keep renewing.

The open question is the first and fourth factors: learning and evidence. Duolingo's in-house research reports that learners who complete its early Spanish and French units reach reading and listening levels comparable to several semesters of US university study, but these studies are company-run, mostly non-randomized and focused on completers. Educators tend to see the app as a useful practice tool and a poor substitute for instruction. The book argues that the strongest long-run defense against free chatbots is evidence that a structured path produces measurable proficiency. That is the factor Duolingo's 2026 messaging, which stresses "teaching better," is now trying to earn.

The takeaway

Duolingo makes money by keeping learning free and charging a small share of highly engaged users for convenience and AI features. The model works because its acquisition cost is close to zero and its retention engine keeps users coming back daily; without both, a 9% conversion rate would not cover the cost of the free base. Generative AI has cut its content costs but added inference costs and free competitors, and its 2026 reset trades near-term bookings for a larger funnel. For the full comparison with what happened to Chegg and other consumer models, read Chapter 9 of The Learning Equation.