Duolingo said in April 2025 that it would gradually stop using contractors for work artificial intelligence can handle. That formalized a shift already underway: the company had reported offboarding about 10% of its contractors around the end of 2023, with AI among the factors reducing demand for some work. Neither announcement establishes that Duolingo cut 10% of its employees or eliminated all human content specialists.
What Duolingo announced in April 2025
CEO Luis von Ahn described an “AI-first” operating philosophy in an all-hands message the company published on LinkedIn. He compared the shift with Duolingo’s decision to prioritize mobile technology in 2012. The message framed AI as a company-wide way of working, not simply a new feature in the language-learning app.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
Duolingo said it would “gradually stop using contractors to do work that AI can handle.” That is a conditional plan, not a claim that every contractor would be terminated immediately. The company also said it cared about employees and intended to use AI to make teams more effective. It said it would consider AI in internal areas including hiring and employee performance reviews. Duolingo’s published message does not give a number of contractor roles affected by the new policy.
How the contractor reductions unfolded
| Date | What happened | What the evidence establishes |
|---|---|---|
| Late 2023 | Contractor work was reduced | Duolingo later confirmed that about 10% of its contractors were offboarded. The figure refers to contractors, not 10% of all employees. Bloomberg reported the figure. |
| January 8–10, 2024 | The reduction became public | Reports described work including translation and lesson content. Duolingo said AI contributed to reduced demand, while also pointing to completed projects and disputing the idea that AI alone explained the change. TechCrunch’s account and the Pittsburgh Post-Gazette report describe the company’s explanations. |
| September 24, 2024 | Von Ahn discussed AI and jobs | In a Forbes interview, the CEO discussed AI’s potential to eliminate some jobs and the need for worker retraining. |
| April 28, 2025 | Duolingo announced its “AI-first” approach | The company said it would phase out contractor work AI could handle, without specifying a total number of roles affected. |
| February 26, 2026 | Duolingo reported FY2025 results | The company reported more than 50 million daily active users and more than $1 billion in bookings for 2025, alongside a strategy emphasizing user growth, teaching improvements, AI, and new subjects. Its results presentation provides the figures and business context. |
Contractors are not the same as employees
The 2023–2024 reduction was reported as affecting contractors, not as a disclosed 10% cut to Duolingo’s full-time employee headcount. The company used terms such as “offboarded” and described contracts not being renewed; news reports commonly used “layoffs” as shorthand. A contract ending can mean a worker loses recurring work, but it is not evidence of a company-wide employee layoff.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
The available reporting does not show that Duolingo eliminated its entire contractor pool, replaced all translators or writers, or broadly replaced its full-time workforce with AI. It also does not establish that total company employment rose or fell as a result.
What AI is doing in Duolingo’s products and content
“AI” can describe several different tasks, and they should not be conflated. Reporting on the earlier contractor reduction said Duolingo used GPT-4 to help generate content and translations, with human experts validating output. Some content work may therefore involve AI-assisted drafts or alternative translations rather than a system independently producing and approving a finished lesson. TechCrunch reported on the use and human validation.
- Content production: Generating or assisting with exercises and course material, potentially making it faster to create more content.
- Translation: Producing or suggesting translations that experts may need to check for accuracy, context, and teaching standards.
- Personalization: Duolingo’s “Birdbrain” system adapts practice to learners. This is machine-learning-driven personalization, distinct from a language model writing a lesson.
- Practice features: Duolingo Max offers AI-powered feedback and conversation experiences, according to company materials and reporting.
- Product expansion: Duolingo has linked AI to adding material and experiences in subjects beyond languages, including chess, math, and music. Its investor materials describe this broader product strategy.
What “AI-first” can mean for work
In practice, an AI-first approach puts the question “Can AI do this?” near the start of a work process. For repeatable tasks such as drafting or adapting exercises, that can reduce the need for contract writers or translators. People may instead focus more on editing, checking, handling exceptions, or building and evaluating the systems. Human review of AI output does not mean humans wrote the first draft, and it does not mean the workload or pay is distributed in the same way as before.
The approach is not synonymous with AI-only. Duolingo has described human experts validating content, while the 2025 policy is specifically about contractor work AI can handle. The company’s September 2024 comments, reported by Forbes, also acknowledged that AI could eliminate some jobs and raised the issue of retraining. Whether workers move into new roles or simply lose assignments depends on the work, the skills required, and the company’s hiring choices.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Why Duolingo sees a business case
More content and faster expansion
Creating and adapting course material with AI could let Duolingo add lessons or subjects without increasing content-production labor at the same pace. That is strategically important to a company seeking to expand its learning products and serve a large global user base. AI-enabled practice also supports features such as personalized feedback and conversation, not just automated writing.
Lower marginal production costs
If AI reduces the time and contractor hours required per lesson or exercise, the company may be able to produce content at lower marginal cost. That could reduce dependence on large pools of contract translators and writers. The company has not disclosed a specific savings figure attributable to the contractor policy, so lower costs are a plausible business rationale rather than a quantified result.
A platform bet, not just a labor policy
Von Ahn presented the change as comparable to prioritizing mobile in 2012. In that framing, AI is a platform shift that could reshape how Duolingo builds products and organizes teams. The FY2025 results presentation says the company was prioritizing longer-term user growth and teaching improvements, even at the expense of some near-term financial growth. Its investor materials continued to present AI and expansion into subjects such as chess, math, and music as parts of the strategy. Duolingo’s 2025 product-update materials also discuss AI-powered features.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What this could mean for learners—and what is not known
Automated drafting may help a company create more practice material, but volume and speed do not establish educational quality. Generative systems can produce plausible language that misses context, register, idiom, cultural meaning, dialect, or the teaching point a lesson is supposed to reinforce. In language learning, a convincing but incorrect answer can be particularly damaging because learners may treat the app’s feedback as authoritative.
Human experts can catch errors, but review can become a bottleneck if automation sharply increases the amount of content to check. Less human course development could also leave less room for nuance in smaller or less commercially important languages. AI conversation features may make practice more available while still producing unnatural dialogue or inconsistent feedback.
The available sources establish that human validation has been part of some content workflows. They do not provide an independent, systematic audit of post-AI course accuracy or learning outcomes. Anecdotal complaints about course quality do not establish a general decline, just as user growth and bookings do not prove that AI-made lessons teach better. The company’s 2025 results reported business scale, not a causal test of AI’s educational effect.
How to tell whether the strategy is working
For learners and workers, “AI-first” is more meaningful when judged by measurable outcomes than by the number of items generated. Useful evidence would include:
- Accuracy and quality: Error rates, correction rates, expert-review workload, and learner-reported errors.
- Learning: Course completion, retention, proficiency gains, and results on standardized assessments.
- Coverage: How many courses, language varieties, and subjects are added, and whether smaller communities receive attention.
- Speed and cost: Time from course proposal to release and content cost per lesson or learner—not just total headcount.
- Worker outcomes: Contractor roles ended, renewed, or shifted into review and evaluation work.
- Trust and accountability: Complaints, cancellations, public corrections, and whether a named human team remains responsible for educational accuracy.
What remains unclear
Duolingo has not publicly specified in the cited materials how many contractors the April 2025 policy affected, how many contractors it currently uses, or how many full-time roles changed. It has also not supplied AI error and correction rates, an independent comparison of learner outcomes, or a figure for savings reinvested in product development. Those gaps prevent a precise accounting of the policy’s labor impact or a verdict on whether it improved teaching.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
