Duolingo Stock After Q3 2026 Earnings: It Stopped Selling Its AI Tier

Duolingo has essentially stopped selling Max, its AI tier, and I think that says more about the moat than the 29% jump in daily users that sent the stock up 26% on November 4. Max was small, single digits of subscribers and bookings, so the money barely moves. What matters is why it went: the AI calls that justified a higher price now cost under a cent. My read is that AI no longer earns Duolingo a premium, the daily habit is the moat, and the habit hasn't yet shown it can charge more. One headline said the user surge starts paying off. I don't think Q3 shows that yet.

Duolingo has a free tier, a paid plan called Super, and Max, a pricier plan built around AI conversation. In August CEO Luis von Ahn said Duolingo might sunset Max. On the November 3 call CFO Gillian Munson said: "We have essentially stopped selling Max to new users." Individual Max subscribers move to Super Family, the group plan, at a similar price.

The cost drop is what matters for the moat. Von Ahn said it came mainly from moving to open models, and any competitor can use those too. Google launched free practice inside Translate back in August 2025. So a speaking partner can't hold a higher price for long.

AI still does a job here. Von Ahn said moving Video Call into Super tested as good for revenue, and that people who try tailored calls during a free trial are more likely to subscribe. So it helps sell the plan at Super's price. Nothing on this call shows it earning a price above that. The savings are real too, but the same price cut reaches every rival.

Bookings, the money customers committed in the quarter, did beat. The August guide for Q3 was about 9% and the result was 12%, with subscription bookings up 16%.

The monthly rows are the ones I'd look at. Monthly users grew 13% and subscribers 14%, so the share of monthly users who pay barely moved. What jumped is how often people open the app, and von Ahn says the vast majority of the improvement comes from retention. That's a strong habit. It also means much of the 29% is existing users coming back more often, rather than a new crowd waiting to convert.

The 29% itself deserves some care. Yahoo reports that the $11.4 million bonus pays only if Q4 daily user growth reaches 25%, so the company is paid to push this number, and that's part of why margin fell.

Duolingo spent 2026 making the free product easier, and in February von Ahn said "we are prioritizing growing the size of the pie". He also said that whenever Duolingo removed friction, meaning the limits that push a free user to pay, daily users grew. My read is that the reverse holds too, and he seems to agree. On this call he described "a strict guardrail inside the company to not degrade the free user experience anymore."

So the 2027 push has to come from elsewhere: longer free trials, avatar purchases, family plans, a direct ad sales team and Chess, which has 10 million daily users and is, in his words, "completely under-monetized." Ads were flattish in Q3, though Munson said they exited the quarter growing. The cheaper ad-supported tier hasn't worked yet: "our A/B tests for Lite have just not been home runs." No 2027 numbers come until February.

Duolingo still holds things back from free users, including Video Call, so it does have something to sell. The guardrail only stops it making the free product worse. Spotify is the contrast I'd draw: its free tier comes with ads and limits, and that's a lever Duolingo has just said it won't pull further. Roblox is the honest counter-example. Its network is free to join and it earns from optional spending, which is close to where avatars and Chess could take Duolingo. But Roblox players spend in front of other players, and learning a language is mostly done alone.

The bull would say I'm grading a harvest in the planting season. Von Ahn says users monetize over time, sometimes over years. Subscription bookings already outgrew total bookings in Q3, Q4 is guided higher, and Munson says subscriber growth probably bottoms in a couple of quarters. More engaged users should be easier to sell to, whatever the share looks like today.

The bull also has an answer on AI. Von Ahn says nobody has as much data on people learning, and words spoken per user have roughly doubled in about a year. He expects to match a one-on-one human tutor over the next few years. Live inference on free users should cost only single-digit millions of dollars next year. If that data makes Duolingo's AI clearly better than a rival's at the same cost, AI could become part of the moat, and I'd be wrong to treat it as just a cost.

The opposite worry is that I'm too kind about the habit. Daily user growth had help: the Q2 letter names a one-time event to revive lost streaks, and von Ahn says game partnerships such as Brawl Stars did bring users, though he says they retain well and most of the growth comes from the app. He also says the product is "not as good as a one-on-one human tutor" today. If a free AI tutor on every phone gets close first, the streak might be all that holds people. I'd want to see the 29% hold after the bonus is paid in early 2027.

This isn't a price call. The scorecard below carries the live valuation, and the scenarios on the Duolingo page should be re-struck when 2027 guidance arrives in February. Management's goal is 100 million daily users in 2028, so I expect growth to slow from here without that meaning much.

So I've set the lines apart. If total bookings growth reaches 20% for two straight quarters by Q3 2027, the habit can charge and the rally was right about more than users. If daily user growth falls below 15% for two straight quarters, the habit is weaker than I think. Anything in between means the habit is holding and pricing is still unproven.