History

The Story of Duolingo

Duolingo went public in July 2021 as an organic-growth, freemium darling — DAU growth above 50% for nearly four straight years, EBITDA margin expanding from negative to ~29%, and a co-founder CEO (Luis von Ahn) who built the story brick by brick. Then in April 2025 a CEO memo on AI-first operations triggered consumer backlash, DAU growth halved over three quarters, the company guided FY2026 bookings growth to 10–12% (from 33% in FY2025) and the stock fell roughly 70% peak-to-trough by May 2026. The current chapter is a deliberate, self-confessed reset: prioritize user growth and "teaching better" over near-term monetization. Management's strategic credibility — they did transform the P&L and they did call the AI inflection — is intact. Their communication credibility took a real hit: they spent every quarter from Q2 FY22 through Q3 FY25 raising guidance, then walked into a steep FY26 cut after publicly insisting "things are going great."

1. The Narrative Arc

Three transitions matter, and they all happened on Luis von Ahn's watch:

1. The profitability pivot (FY2022 → FY2024). The IPO story was growth-at-all-costs. By Q4 FY22 management started guiding EBITDA margin to 10–12%, then re-rated to 21–23% (FY24) and 27–29% (FY25). Actuals beat guidance every year — this is the most credible thing on the page.
2. The AI declaration (Q4 FY22 letter, Feb 2023). Two months after ChatGPT shipped, von Ahn put AI at the center of the strategy. Duolingo Max launched as the second monetization tier in March 2023. Through FY24, AI was a tailwind narrative — Video Call (Q3 FY24) was the marquee feature.
3. The AI-first reset (April 2025 → today). A CEO LinkedIn memo telling employees AI would replace contractors and become a performance-review input went viral the wrong way. By the time Q3 FY25 results landed in November, DAU growth had decelerated from +49% (Q1) → +40% → +36%. Q4 FY25 results crashed FY26 guidance and the stock with it.

2. What Management Emphasized — and Then Stopped Emphasizing

The shareholder letters from Q2 FY22 to Q1 FY26 are formulaic enough that the words management chose to stop using are as revealing as the ones they added. Below is the rough frequency of headline themes in the quarterly letters and press releases (0 = absent, 3 = dominant).

Four patterns stand out:

• "Take the long view" was one of nine operating principles in 2021 and never appeared at the top of a letter for thirteen quarters. It became a defining phrase only in Q3 FY25 — exactly when management needed cover for slowing DAU growth and a coming guidance cut. The phrase carried through Q4 FY25 and Q1 FY26.
• Word-of-mouth/brand marketing was a recurring brag in FY22–FY24 (Cannes Gold Lion, Super Bowl "Superb Owl," Dead Duo at 1.7B impressions in Q1 FY25). By Q2 FY25 management noted they had toned down "unhinged" social content "as we listened to community feedback" — and the brand bragging never came back to its prior intensity. The same channel that powered organic growth had become a liability after the AI-first memo.
• Profitability / margin expansion dominated letters from Q4 FY22 through Q2 FY25 — every quarter they raised the FY EBITDA-margin guide. Starting Q3 FY25 the margin narrative recedes; by Q1 FY26 the company is guiding FY26 gross margin down (to ~69% from 73%) and EBITDA margin down (to ~25.7% from ~29%).
• Multi-subject (Math, Music, Chess) is the most volatile theme. Math launched as a separate iOS app in 2022, got folded into the main app in 2023, then disappeared from headline coverage. Music followed the same arc. Only Chess (launched Q1 FY25) sustained attention — "fastest-growing subject we've ever launched" — but by Q1 FY26 it was grouped generically with Math and Music as a "next user growth engine."

3. Risk Evolution

The structural top-3 risks (engagement, competition, brand/product) are unchanged across all five 10-Ks. The interesting moves are at the edges — what got added, what got softened, and what the company finally admitted in plain English.

Reading the risk-factor changes year by year:

• AI risk went from zero to dominant in three years. FY2021 had no AI disclosure. FY2022 added a single regulatory paragraph. FY2023 added the full operational framework after Max launched, naming OpenAI/GPT-4 dependency and content-training-data litigation. FY2024 added third-party AI vendor risk. FY2025 made it personal — the April 2025 memo is named as a contributing cause of user-growth deceleration.
• Apple concentration is the slow-burn risk. App Store revenue share rose from 50% (FY21) to 62% (FY25). The "generally 15–30%" fee disclosure has been identical for five years despite the entire epic-vs-apple regulatory backdrop. Management has started testing direct web purchase flows for US iOS users — material to FY26 gross margin if it works.
• DET shrunk from 6.1% to 4.0% of revenue between FY24 and FY25 and the risk language now explicitly cites "a country's decision to cap or reduce immigration." This is the first quantitative decline in DET disclosure.
• The pre-IPO risks faded on schedule — COVID-19, Brexit, "limited operating history," "operating losses each year since inception" all disappeared as the company matured. By FY25, "operating losses" became "maintain or increase profitability."
• The 10-K reads tougher in FY25. Adding GenAI competitors (chatbots, wearable translators), AWS us-east-1 outage example (October 2025, 12+ hours of downtime), Texas AI Act, and Trump's December 2025 AI EO all add real surface area. The brand risk language is also the most candid it has ever been.

4. How They Handled Bad News

Three episodes test the framing.

The DAU deceleration (Q4 FY23 → Q3 FY25)

Management telegraphed the deceleration honestly. The Q4 FY23 letter (Feb 2024) warned: "so far in Q1 2024, our YoY DAU growth is closer to the mid-50s than the 60%+ it was throughout 2023." That softened the landing for the next eighteen months. The break came in Q3 FY25 — the deceleration accelerated, and management connected it to the AI memo for the first time.

The AI-first memo backlash (April–May 2025)

This is the messiest episode. The Q2 FY25 letter (August 2025) — published four months after the memo went viral on LinkedIn and triggered consumer-press coverage from PCMAG, Inc, and others — contains zero direct reference to the memo. The only acknowledgment is buried in the Q3 FY25 letter (November 2025): "Growth was slightly slower than Q2 in part because we posted less 'unhinged' content on our English-speaking social media accounts as we listened to community feedback and prioritized building long-term brand sentiment." By the time the FY2025 10-K was filed in February 2026, the company finally said it plainly in the risk factors.

CEO von Ahn then walked back the memo's most controversial pieces twice — first publicly retreating from contractor replacement in May 2025, then scrapping the AI-usage performance-review rule in April–May 2026. Both walk-backs were covered externally; neither got top-line treatment in a shareholder letter.

The FY26 guidance cut (February 2026)

This was handled the way it should have been: pre-announced as a deliberate, strategic decision. The Q3 FY25 letter set the stage ("we're prioritizing user growth over monetization in the A/B tests that get launched"). The Q4 FY25 press release (Feb 26, 2026) bundled the FY26 cut (10–12% bookings vs 33% the year prior) with a $400M buyback authorization and a new CFO (Gillian Munson, replacing Matt Skaruppa, who never appeared by name in shareholder letters despite being CFO since the IPO). The stock fell ~14% on the day, ~22% intraday — investors did not buy the "deliberate" framing.

5. Guidance Track Record

The promises that mattered, plain reading:

• Financial guidance: kept and exceeded, every year. Revenue, bookings, and EBITDA margin beat the initial annual guide for FY22, FY23, FY24, and FY25 — no exceptions. The EBITDA margin path beat initial guide by 200–700 bps each year. Long-term margin guide of 30–35% (set Q4 FY22) is still notionally outstanding but is no longer the headline.
• The DAU promise is where the credibility damage sits. Management spent 2023–early 2025 implying that organic, word-of-mouth growth was structurally durable. They warned in Q4 FY23 that 60%+ was unsustainable, but Q1 FY25 still printed +49% and the Q1 letter projected continued strength. Three quarters later DAU growth was +21% and FY26 was guided to a 10–12% bookings deceleration.
• The AI promise was double-edged. AI as a tailwind (Max, Video Call, content scaling) is real — Max contributes meaningful bookings, content production accelerated ~10× by Q1 FY26. AI as a cost lever has been managed. But AI as a communications topic has been mismanaged: the April 2025 memo was a self-inflicted growth headwind that the company now formally cites in its risk factors.

Credibility: 6 of 10 — financial guidance pristine, communications damaged. Financial guidance has been a clean beat for four consecutive years — that is genuinely above average. Strategic foresight was real on AI (early Max launch, content automation, Video Call) and on multi-subject extension (Chess is the most interesting new product since the IPO). The downgrade is entirely about communication: management let an internal-facing AI memo go viral, then waited until the 10-K to acknowledge it had cost the company users; they kept raising guidance into Q2 FY25 and then cut FY26 hard; and the unannounced CFO transition (Matt Skaruppa to Gillian Munson) arrived in the worst possible market window. A 6 reflects "competent operators we believe on numbers, less on narrative."

6. What the Story Is Now

The current story is a previously-loved compounder mid-reset. The reset is part voluntary (lean into AI; tone down brand voice; sacrifice near-term monetization for long-term DAU) and part involuntary (the AI memo permanently changed the consumer brand-perception risk profile; immigration policy permanently lowered the DET ceiling; the AI-cost path means gross margin trades 200–400 bps lower for now).

What has been de-risked: the P&L. Duolingo is profitable, generates real free cash flow, has $1B+ of cash, just announced a $400M buyback, and has now disclosed honestly that the AI memo cost it users. The financial guidance machine — the most robust thing about the company — has not given investors a reason to disbelieve it for four years.

What still looks stretched: the long-term 30–35% EBITDA margin target, the assumption that DAU growth re-accelerates without further heroics, and the implicit narrative that this was a "deliberate" reset rather than a forced one. Reasonable people can read Q1 FY26's +21% DAU as either the cycle bottom or the new normal — and that uncertainty is what the ~70% drawdown is pricing.

The next two prints (Q2 and Q3 FY26) decide which.