Inside Duolingo's Quiet Elimination of Its Contractor Workforce
In April 2025, Duolingo told investors something worth pausing on. The language-learning company, which had employed roughly 1,000 contractors to translate content, illustrate characters, and write curriculum, confirmed it had cut a substantial portion of that workforce. The reason, stated plainly in an internal memo that later circulated publicly: the work those contractors did was now being done by generative AI. CEO Luis von Ahn described the move not as a cost-cutting measure but as an evolution of the company's production model. Duolingo, he said, was becoming an "AI-first" company.
The specifics, reconstructed from the memo, subsequent earnings commentary, and reporting by The Verge and Bloomberg, tell a more structured story than a single announcement implies. The contractor reductions had been underway since late 2024. By the time von Ahn's memo surfaced, the company had already shifted the bulk of its content localization and illustration pipeline to AI-generated outputs, with human review limited to quality assurance passes rather than original creation. The translators and illustrators who had built Duolingo's course library across more than 40 languages were not laid off en masse on a single day. They were, instead, not renewed.
Non-renewal is how this particular displacement tends to work. Because the affected workers were contractors — engaged through platforms and short-term agreements rather than payroll — they carried no severance entitlement and generated no WARN Act filing. The transition left little formal paper trail. What it left instead was a before-and-after: a workforce that had numbered in the hundreds for content-related roles, and a company that now routes the same production volume through a substantially smaller team of employees managing AI pipelines.
What the Numbers Show
Duolingo reported 2024 annual revenue of approximately $748 million, up 41 percent year-over-year, according to its February 2025 earnings release. Headcount growth did not keep pace. The company ended 2024 with around 800 full-time employees — a figure that had grown modestly — while simultaneously acknowledging the contractor reductions. The divergence between revenue growth and workforce growth is precisely the dynamic that economists at the Goldman Sachs Global Investment Research division flagged in their 2023 report on generative AI's labor effects, which estimated that roughly 300 million full-time-equivalent jobs globally could be exposed to automation. Content localization and creative production were among the categories that report identified as particularly susceptible.
The Oxford Martin School's work on task-level automation, building on the foundational Frey and Osborne research, had placed translation occupations at elevated displacement probability even before large language models became commercially viable. What Duolingo's 2025 transition demonstrated was how quickly that theoretical probability became operational reality once a company with the resources and incentive to automate actually ran the experiment.
Duolingo is not unique in having run it. But it is unusually well-documented. The internal memo, published in part by The Verge in April 2025, gave the company's own framing in writing — which most companies undergoing similar transitions avoid producing. Von Ahn's language was precise: managers would be required to demonstrate that a given role could not be performed by AI before receiving approval to hire. The policy applied to contractors first, employees second. The sequencing matters.
Contractors represent an exposure class that sits largely outside the labor protections and income supports designed for traditional employees. The Bureau of Labor Statistics does not maintain a continuous, comprehensive count of independent contractors in the United States, a data gap that the Brookings Institution has criticized repeatedly. What that gap means in practice is that displacement events that route through contractor non-renewal are difficult to aggregate or measure. Duolingo's case became visible because the memo leaked. How many equivalent decisions were made in the same quarter, at companies without a leaked document, remains genuinely unknown.
The Workers Who Were Not Renewed
In the months following Duolingo's announcement, several affected contractors spoke to journalists and posted publicly on LinkedIn and Reddit. The accounts describe similar patterns: project volume declined through late 2024 without explicit explanation, contracts came up for renewal and were quietly not extended, and inquiries to project managers produced vague responses about "changing production workflows." A translator who had worked with Duolingo on Spanish-language curriculum for three years told Rest of World in May 2025 that she received no direct communication attributing the end of her contract to AI. She inferred it from the memo.
That inference problem — workers knowing something structural has changed, but lacking formal confirmation — is one of the features that makes this displacement category difficult to document from the ground up. It also makes income planning difficult. A contractor who might have anticipated six months of continued project work while searching for alternatives instead found the work absent without a defined transition point.
Aria Chen, who built the income-protection service clankers.today after her own displacement from a software engineering role in 2024, has described the contractor population as the group her product was most urgently designed for — workers who lack access to traditional unemployment insurance and often have less runway to absorb an abrupt income loss. The Duolingo situation, she has noted in interviews, is a concrete illustration of the gap that product attempts to address: technically skilled workers, engaged in legitimate professional work, whose relationship to their employer offered none of the buffers that standard employment provides.
The McKinsey Global Institute's 2023 report on generative AI estimated that occupations involving language and translation tasks faced some of the sharpest near-term exposure, with the caveat that exposure to automation does not automatically produce displacement — it requires a company to act on the capability. Duolingo acted. The Stanford AI Index 2025 report noted that enterprise adoption of generative AI for content production accelerated sharply through 2024, with cost reduction in creative and localization workflows among the primary cited motivations in corporate surveys.
The World Economic Forum's Future of Jobs Report 2025 offered the most widely cited forward estimate: approximately 92 million roles displaced globally by 2030, offset in their projection by 170 million roles created. The net figure, positive in aggregate, does not address the distribution problem — that the roles eliminated and the roles created tend to accrue to different workers with different skill profiles and different geographies. A translator in Mexico City who built a professional practice around Duolingo contract work does not automatically transition into the prompt-engineering or AI-quality-assurance roles that the same company now staffs.
Duolingo has not disclosed precise numbers on contractor reductions, and the company did not respond to a request for comment for this article. Von Ahn's memo remains the primary corporate document on record. What it established, in language that was unambiguous if brief, was a firm-level policy: AI capability creates a presumption against human hiring. For the contractors who built the library that trained users of that same AI, the policy arrived after the work was already done.
PROTECT YOUR INCOME NOW
If AI automation threatens your job, preparation is the only honest answer: know your risk, build a runway, and reskill before you have to. The Clankers 90-day programme structures exactly that — free, which is exactly why today is the right time to start.
Start the 90-Day Programme — Free →GET THE CLANKERS BRIEF — FREE
One email every Monday: the week's AI displacement moves that matter, and what to do about them. No spam. Unsubscribe anytime.