Moat

Moat — What Protects Duolingo, If Anything

A moat is a durable economic advantage that lets a company protect returns, margins, share, or customer relationships against competition. The verdict here is Narrow moat. The defensible part of Duolingo is behavioural — a daily-habit engagement loop, a culturally embedded brand, and a 2-billion-exercise-a-day data flywheel that no listed peer can match. The undefended part is structural — Apple and Google take an estimated 15–30% of every in-app subscription, curriculum is increasingly commoditised by free large-language-model chat, and Duolingo competes for share of attention with infinite-scroll video, not just with other language apps. The strongest single piece of moat evidence is third-party retention data: Duolingo had a 28% Western-market churn rate in late 2023, the lowest of any EdTech app tracked by Sensor Tower, versus 58% at Babbel and 64% at Simply Piano. The strongest single piece of evidence against a wide moat is the Chegg base-rate — a thin-moat consumer EdTech subscription that lost roughly two-thirds of its paying users to free LLMs in three years. Duolingo's defences are materially better; the structural exposure is the same.

1. Moat in One Page

The conclusion is Narrow moat for three reasons.

Why it qualifies as a moat at all. Duolingo is the only listed consumer-subscription business combining +39% revenue growth, 72% gross margin, positive GAAP operating income, and 35% FCF margin. That financial profile cannot be produced by execution alone — it requires a structural advantage. The advantage that shows up in the data is engagement: a DAU/MAU ratio of roughly 40%, which is a level only the best gaming and social platforms reach, and a Sensor Tower-measured churn rate that is roughly half that of the closest direct competitor (Babbel). The third-party data and the company's own KPIs agree on the same point: paying users stick.

Why it is narrow rather than wide. Two structural exposures cap the moat. First, the platform tax: ~28% of FY2025 cost of revenue is third-party payment processing paid to Apple and Google — a toll that does not exist at Stride (B2G), Coursera (B2B/web), or in any of the institutional incumbents. Duolingo has no leverage over this line. Second, AI substitution: the exact tasks Duolingo monetises (vocabulary drilling, grammar correction, conversational practice) are the tasks where a free LLM chat session is now competitive on quality and superior on flexibility. The defence is behavioural (streaks, brand, social) — not curriculum or technology — and behavioural moats erode invisibly until the day they don't.

Why it isn't "no moat." The Chegg base rate exists because Chegg never had real engagement. Its DAU/MAU was low, its product was task-based not habit-based, and its brand was a utility. Duolingo's product is consumed daily and emotionally; the green owl is an internet meme; the company runs the largest mobile A/B testing platform outside of Big Tech. None of that is replicable in 18 months by a well-funded entrant. The moat is real — it is just not infinitely deep.

2. Sources of Advantage

Every claimed source below has to clear two tests: does it show up in numbers, and is it specific to Duolingo or merely a feature of the industry?

The two sources that pass both tests are brand + cultural footprint and behavioural switching costs. Data flywheel, scale economies, and app-store ranking are real but second-order; they protect against subscale challengers and not against the substitution risk that actually matters. Network effects are an external-rating cliché that does not hold up against the company's own disclosures — Duolingo is not a marketplace and not a UGC platform.

3. Evidence the Moat Works

A claimed moat is only useful if it shows up in actual business outcomes. The table below pulls together the strongest external and filing-based evidence — for and against.

The five "supports" items establish that an advantage exists; the three "refutes" items establish that it is narrower than the bull case implies. Investors who weight only the first five end up at "wide moat" — which the April-2025 brand event and the FY26 bookings cut have already shown to be optimistic.

The churn chart is the single best evidence point in this tab. It is the only direct, peer-comparable measurement of whether the alleged moat actually does what a moat is supposed to do: keep customers from leaving. Duolingo's churn is roughly half that of the closest direct substitute (Babbel) and a third of the closest gamified-EdTech substitute (Simply Piano).

4. Where the Moat Is Weak or Unproven

Three areas where the moat is meaningfully thinner than the consensus narrative.

The brand has been measurably damaged in the last twelve months. The April 2025 AI-first CEO memo is named in the FY25 10-K risk factors as a realised event that "may have contributed to unfavorable publicity, adverse impacts on the Company's brand and social media presence, and a deceleration in user growth." DAU growth went from +49% in Q1 FY25 to +21% by Q1 FY26. Companies rarely name their own communications as the cause of a realised risk. A brand moat that fades on a single LinkedIn post is not the kind of brand moat investors should be paying a premium for; it is closer to a consumer-sentiment moat, which is real but mood-sensitive.

The data flywheel is being commoditised. Frontier LLMs ship with high-quality multilingual capability out of the box, with no Duolingo-style A/B testing required. The Birdsong experimentation platform and the 2-billion-exercise-per-day data flywheel are advantages over a subscale rival like Babbel — but they are not advantages over OpenAI or Google. Management has accepted this implicitly by guiding gross margin down ~300 bps to fund AI features that ride on third-party model APIs; the cost structure now includes a variable inference line whose unit economics are set by Anthropic, OpenAI, and Google, not by Duolingo.

App-store distribution is the moat that nobody wants to talk about. Roughly 28% of FY25 cost of revenue is payment processing fees to Apple and Google. The advantage of being #1 in the App Store Education category is enormous — but it sits inside a duopoly that owns 100% of mobile distribution and that has historically not been a value-accretive partner for any subscription business. Spotify won a multi-year regulatory campaign to weaken Apple's hold; Duolingo has not. The moat against Babbel is real; the moat against the platform on which Duolingo and Babbel both run is zero.

The unproven moat sources — patents, true network effects, and curriculum quality — should be excluded from any underwriting case. Patents are a defensive line item in this industry; "network effects" as cited by external rating services confuses social features with two-sided markets; curriculum is the one place where a free LLM is structurally competitive.

5. Moat vs Competitors

Duolingo's moat has to be sized relative to the alternatives, not in absolute terms. The comparison below uses the listed peer set from Competition tab, plus the most relevant private substitutes and the AI-tutor reference point that does not appear on any public peer table.

The peer table tells two different stories about the same name. Among listed consumer-EdTech, Duolingo has the strongest moat — by a wide margin. Among scaled freemium consumer-engagement platforms, Duolingo's moat is mid-pack: weaker than Roblox's true two-sided network, weaker than Spotify's regulatory progress on app-store billing, but with better unit economics than either. Where Duolingo is genuinely alone is in its financial profile — the upper-right of growth versus margin — not in its moat depth. That gap between "unique financial profile" and "narrow moat" is the source of investor confusion when valuing the business.

6. Durability Under Stress

A moat only matters if it holds up under pressure. The table below tests Duolingo's defences against the specific stresses likely to test them over the next 24 months.

The stress map is unbalanced by design. AI substitution sits above every other risk because the base-rate (Chegg) is severe, recent, and directionally applicable. Every other stress is manageable inside an engagement-led moat; the AI scenario is the one stress that could compress the moat itself rather than just rotating around it.

7. Where Duolingo Fits

The moat does not apply uniformly across the business. The strongest defences are concentrated in the consumer language-app core. The weakest defences are at the edges.

The honest reading is that the moat is overwhelmingly concentrated in the consumer language-app Super tier, which carries roughly four-fifths of revenue. Max, advertising, DET, and the multi-subject expansion are all reasonable optionality but should not be weighted heavily in any moat underwriting. The moat is one product, not a portfolio.

8. What to Watch

Five signals would tell an investor whether the moat is strengthening, holding, or fading — and would do so before the revenue print and before sell-side price-target cuts.

The first moat signal to watch is paid % of MAU. The Chegg pattern started exactly there — engagement held while paid conversion flattened, then fell. If Duolingo's paid % of MAU drifts down even one quarter while DAU growth holds, the worst-case AI-substitution scenario is already starting to play out, and every other moat metric will follow within two quarters.