Explain the Drama: Duolingo’s AI-First Backlash

Imagine opening your language-learning app and discovering that its CEO wants to stop using contractors for work AI can handle. Then you read the next line: employees’ use of AI could count in their performance reviews. That was Duolingo’s pitch in April 2025. For some learners, the immediate question was: wait—who, or what, is teaching me?

Duolingo is the app with the persistent green owl. Luis von Ahn is its CEO, and this wasn’t a rumor about a secret plan: Duolingo posted his company-wide “AI-first” email publicly. Commenters worried about work moving away from contractors and questioned whether they could trust an AI-first language-learning product.

There was history behind that reaction. In January 2024, Duolingo cut ten percent of its contractors, in part because AI could perform some of their tasks, according to Engadget. Then came the April 2025 email. Von Ahn said creating enough lessons manually would take too long. His proposed changes included gradually stopping the use of contractors for work AI could handle, considering AI use in hiring and performance reviews, and granting teams more headcount only if they couldn’t automate more of their work. He also wrote that he’d rather move urgently and take occasional small hits to quality than miss the moment.

That last idea is a striking thing to hear from an education app. In comments beneath Duolingo’s post, one person questioned how the company could say it cared about employees while planning to move work away from contractors. Another said they couldn’t fully trust a language-teaching product that called itself AI-first. Those are reactions, not proof that Duolingo’s lessons became inaccurate. But they explain why this hit a nerve: learners want the app to get a language right, while commenters also worried about the proposed reduction in contractor work.

Von Ahn’s defense was that access matters too. He argued that AI could help Duolingo make far more content for far more learners, and free its employees to focus on creative problems instead of repetitive work. He said the plan was not about replacing Duolingo employees. There’s an important distinction in his wording, though: the email separately proposed reducing contractor work. Duolingo’s December 2025 roundup said it had launched 172 new language courses that year. That number shows the scale it was pursuing; it does not, by itself, tell us how every lesson was made or reviewed.

And then came a change. In comments reported in April 2026, von Ahn said Duolingo had backtracked on measuring employees by how much they used AI. What matters most, he said, is doing the job well: AI might help, but if it doesn’t, he wouldn’t force its use. That is a retreat from one part of the original memo, not an announcement that Duolingo has abandoned AI—or that it has reversed its contractor plans. The company had reported its course expansion; the question of how to build lessons learners trust remained.

So this drama isn’t simply “AI versus no AI.” It’s about who gets the benefit of moving faster, who carries the risk, and whether a change to workplace rules answers learners’ doubts. Here’s the question for a follow-up: if Duolingo no longer wants to judge employees by their AI use, will it also rethink how much contractor work it asks AI to handle?