Duolingo Cuts 10% of Contractors as It Uses More AI to Create App Content
Duolingo’s move to replace about 10% of its content contractors with AI-generated lessons is prompting debate over both language-learning quality and the human cost of automation. Many users report the app has become more gamified while its pedagogy stagnates or worsens, and several say they now get better, more flexible tutoring from tools like ChatGPT or comprehensible-input videos. Others argue that AI is a natural fit for translation and practice, but worry about accuracy, overreliance on machine-generated material, and the erosion of paid language and translation work.
Overall sentiment
- Mixed-to-negative reaction to Duolingo’s direction: many feel quality has declined, gamification has increased, and AI-driven changes are likely to worsen this, though some think Duolingo is a natural fit for AI.
Language learning effectiveness
- Many report Duolingo is fine for vocab, basic grammar, and light practice, especially for travel-level use.
- Several say long Duolingo streaks (months/years) produced far less progress than a few weeks of classroom instruction or immersion.
- Common view: Duolingo alone won’t lead to fluency; it’s best as a supplementary tool “instead of scrolling social media.”
- Some users describe it as “language-flavored entertainment” or a “guilt-easing game” rather than a serious course.
- Others share success using it as a foundation combined with classes, immersion, or other resources.
Gamification, UX, and product changes
- Strong complaints about aggressive gamification (streaks, leagues, chests, microtransactions) that reward point-chasing over real learning.
- Several users feel “enslaved” by streaks and deliberately plan to break them to regain control.
- Perceived “enshittification”: more ads, more friction, and slower apparent progress, especially in some courses (e.g., Japanese).
- Removal of exercise-level discussion threads and community content is widely lamented; those threads were seen as the most educational part.
AI and content quality
- Duolingo uses GPT‑4; some users say the new AI exercises are more challenging and Duolingo would be foolish not to adopt AI.
- Others fear AI-generated content makes Duolingo indistinguishable from any basic “AI phrase + TTS + cartoons” app, losing its former human-crafted edge.
- Multiple reports of wrong answers being accepted and bad translations, predating or continuing with AI.
- Debate over AI translation quality:
- Some say GPT‑4 is excellent across many languages, including Chinese and Polish, with strong nuance when prompted well.
- Others highlight weaknesses, especially in gendered languages and idiomatic use, and note that a small error rate undermines learner trust.
AI as tutor vs Duolingo
- Several commenters now prefer ChatGPT (sometimes with speech-to-text/text-to-speech loops) over Duolingo or even human tutors for:
- Infinite patience, detailed explanations, and tailored practice.
- On-demand idioms, role-play scenarios, and grammar breakdowns.
- Concerns are raised about hallucinations and subtle errors; mitigations suggested include checking with native speakers and cross-verifying in separate AI sessions.
- Some predict AI chat-based practice and comprehensible-input media will quickly outcompete static apps like Duolingo.
Labor and societal impacts
- Contractor cuts framed as part of a broader trend: AI and automation reducing language-related work (e.g., translators shifting to low-quality post-editing).
- Broader worries about loss of meaningful work relationships and crafts under automation; others counter this is similar to past technological shifts and accuse critics of neo-Luddism.