Key Takeaways
- Preply now counts more than 150,000 tutors and reached a $1.2 billion valuation in January 2026, nearly tripling its prior valuation, by positioning AI as a copilot to human tutors rather than a replacement.
- OpenAI published a case study on June 12, 2026, framing Preply as proof that AI deployed through human experts outperforms AI deployed instead of them.
- Preply’s first annual Progress Report (October 1, 2026) found nearly half of learners struggle to apply what they learn in real life, undercutting the case for AI-only instruction.
- Duolingo’s Q3 2026 results (reported November 3–4, 2026) showed DAU growth accelerating to 29% while AI video-call costs fell below $0.01 per call, evidence that AI features are becoming deflationary table stakes.
- The hybrid model faces an unresolved cost question: human tutors are expensive and non-scaling, and Preply has not disclosed what share of sessions use AI-generated prep.
The Lede
Preply, the Barcelona- and Kyiv-founded language-learning marketplace, has surpassed 150,000 tutors and a $1.2 billion valuation by selling a deliberate counter-thesis to Duolingo: that AI works best as a copilot to human teachers, not a substitute for them. The position gained institutional weight on June 12, 2026, when OpenAI published a case study casting Preply as the model for “human + AI” learning.
Key Facts
- June 12, 2026: OpenAI publishes “How Preply combines AI and human tutors to personalize learning,” describing an architecture in which AI handles lesson summaries, homework generation, and personalization scaffolding while human tutors deliver speaking practice, correction, and accountability.
- January 21, 2026: Preply raises $150 million (€127 million) at a $1.2 billion valuation, nearly tripling its previous valuation and reaching unicorn status. The round is earmarked for global expansion, including Latin America, and scaling AI.
- October 1, 2026: Preply releases its first annual Progress Report, finding nearly half of learners struggle to apply lessons in real life. Separate research on “second-language anxiety” follows on September 30, 2026.
- October 2, 2026: Preply announces it has surpassed 150,000 tutors.
- October 14, 2026: Preply hires Bilal Zia as VP of Data Science and Analytics and Lillian Reaume as Chief People Officer.
- November 3–4, 2026: Duolingo reports Q3 2026 results: DAU growth accelerates to 29%, AI video-call costs drop below $0.01 per call, and the company sunsets its Max tier, folding AI features into the cheaper Super tier.
- Context: Preply’s AI pivot traces to a $70 million round in July 2023 aimed at “supercharging” tutors with AI; the company opened its first U.S. office in October 2023.
The Architecture: AI as Copilot, Not Instructor
The core of Preply’s model is a deliberate division of labor. AI performs the invisible scaffolding of language learning: summarizing lessons, generating homework, tracking vocabulary gaps, and personalizing the next session’s content. Humans handle what machines still cannot credibly replicate, live speaking practice, real-time correction, and the accountability that comes from a learner showing up for another person.
The clearest technical window into this wiring came from Dmytro Voloshyn, Preply’s Chief Technology Officer, in a July 29, 2026 interview with Pulse 2.0, which described how AI is embedded directly into the tutor workflow rather than offered as a standalone product. The strategic logic was formalized on August 5, 2025, when Preply announced new AI features under a positioning it summarized as building “the world’s best AI for language learning, powered by real tutors.”
That framing is a direct answer to Duolingo’s May 2025 “AI-first” memo, which declared that the company would prioritize AI over human contractors and triggered a measurable user backlash. Preply responded in July 2025 with its “The Better Duo” campaign, positioning its human tutors as the antidote to automated instruction.
The pattern is reusable well beyond education. In healthcare, legal advice, and financial planning, the same architecture is emerging: AI absorbs research, documentation, and triage, while licensed humans retain judgment and liability. OpenAI’s interest in the case study is not incidental, it functions as a deployment blueprint for how the model maker wants enterprises to adopt its technology, as an augmentation layer with human accountability rather than a headcount-reduction tool.
The Economics: Why Hybrid Is Now the Defensible Position
Duolingo’s Q3 2026 numbers supply the counter-evidence that makes Preply’s positioning legible. The company reported daily active user growth accelerating to 29%, and disclosed that AI video-call costs had fallen below $0.01 per call. It also sunset its premium Max tier, folding AI features into the cheaper Super tier, an admission that AI conversation practice is becoming a commodity rather than a premium product. Duolingo’s stock, meanwhile, trades roughly 80% below its peak.
That combination, cheap, deflationary AI features and a punishing public-market reaction, is precisely the environment in which a human-in-the-loop model becomes defensible. Preply’s $1.2 billion valuation arrived despite Duolingo’s decline, a signal that investors are paying for differentiation that cannot be replicated by dropping inference costs.
The empirical case arrived on October 1, 2026, with Preply’s first annual Progress Report. Its central finding, that nearly half of learners struggle to apply lessons in real life, was amplified by Fortune’s coverage the same day, which reported that 57% of Gen Z feel “stuck” despite heavy AI use. The implication is blunt: access to AI instruction does not equal mastery.
The unresolved tension is cost. Human tutors are expensive and do not scale the way software does. Preply’s $150 million raise must fund enough AI leverage to raise tutor productivity, more students per tutor hour, better retention, higher session value, to protect margins against competitors whose marginal AI cost approaches zero. Preply has not disclosed what share of sessions use AI-generated prep, nor how its take rate affects tutor earnings, leaving the model’s unit economics unverified from the outside.
The reputational backdrop adds pressure. A class action filed October 27, 2026, alleges that Duolingo’s AI efforts and heavy advertising hurt user growth, a reminder that the “AI-first” brand carries legal and commercial risk that Preply’s human-centric framing is designed to avoid.
Industry Reaction and Expert Perspectives
Kirill Bigai, Preply’s Co-Founder and CEO, has become the most prominent voice for the human-led, AI-enhanced thesis, arguing in an August 23, 2026 Fortune profile that technology should elevate tutors rather than displace them.
Luis von Ahn, Duolingo’s CEO, defined the opposing camp, and then softened it. “I do not see AI as replacing what our employees do,” he said in the May 2025 walkback that framed the entire debate.
OpenAI’s role is more than chronicler. By publishing the Preply case study, the company is acting as a strategic author, signaling to enterprises that the sanctioned deployment pattern is augmentation with human accountability, an approach that reduces regulatory exposure and public-relations risk compared with wholesale automation.
Preply’s organizational moves reinforce the dual bet. The October 14, 2026 appointments of Zia and Reaume invest in both the data science that powers personalization and the people function that manages a 150,000-tutor marketplace. The fifth annual Online Teaching Conference on October 21, 2026, meanwhile, kept the human side of the model in public view.
The Bottom Line
The “human + AI” framing has moved from marketing slogan to investable thesis, validated by Preply’s tripled valuation and amplified by OpenAI’s case study. But the model’s durability depends on a number Preply has not published: how much AI raises tutor productivity relative to tutor cost. Watch for the company’s next Progress Report to shift from measuring the problem to proving the solution’s efficacy, and for whether “AI-first” competitors follow Duolingo in folding AI into cheaper tiers, ceding human differentiation to the hybrid players.
