Art Editor 丨 Li Chengxi

'AI serves as a tool to enhance top teachers, not as a substitute.' When Gaotu's founder, Chen Xiangdong, underscored this notion with expansive hand gestures on camera, the once-dominant online education behemoth was already ensnared in an AI transformation morass, desperately seeking a path to redemption.

The third-quarter 2025 financial results disclosed a 30.7% year-over-year (YoY) revenue surge for Gaotu. However, net losses still amounted to 147 million yuan. Under its 'All-in-AI' strategy, products like 'Smart Azu' and 'AI Score Booster' faded into obscurity, absent from rankings dominated by competitors such as KuaiDui AI and Xiaoyuan Mental Arithmetic.

While New Oriental carved out a second growth trajectory through livestreaming e-commerce and TAL Education Group reaped substantial profits with 'AI Learning Devices + Apps,' Gaotu's status as an 'AI-era laggard' stems not from a fundamental incompatibility between AI and education but from triple missteps in strategic vision, resource allocation, and organizational inertia—transforming golden opportunities into survival crises.

Strategic Dilemma

Gaotu's cognitive bias toward AI fundamentally condemned its transformation to inefficiency.

Chen Xiangdong's 'Three-Teacher Model' (Master Teacher + Tutor + AI Companion) positions AI as a mere 'supporting tool' rather than a core engine for redefining educational experiences. Within this framework, AI handles grading and basic Q&A, while human teachers dominate core instruction—a seemingly safe approach that entirely overlooks the transformative potential of AI-driven education.

The industry has long demonstrated that AI education's breakthrough lies in harmonizing 'personalized efficiency gains' with 'scenario-based solutions.'

Zuoyebang's KuaiDui AI secured seventh place on Q3's AI App Value Rankings with its 'AI Problem-Solving + Writing + Translation' all-in-one toolkit.

TAL's AI Learning Device achieved a 120% YoY hardware sales growth through a closed loop of 'academic diagnosis - intelligent test paper generation - precision score improvement.'

These successes share a common trait: AI ceased being a 'sidekick' and directly addressed core pain points for students struggling to improve scores and parents exhausted by tutoring.

In contrast, Gaotu's AI strategy fell into a 'scattered without focus' trap: 'Gaotu AI English' and 'Maodou Aixue' for primary/junior students, 'AI College Application Guidance' and 'AI Score Booster' for high schoolers, and 'Oral English Practice Partner' for adults. While seemingly covering all scenarios, each product remained superficial.

Take its English course collaboration with Daniel Wu as an example: 'Smart Azu' merely handled basic interactions, while the core remained traditional course sales. The model of attracting customers with a 398-yuan low-cost entry point before pushing 3,980-yuan packages exposed its essence of 'using AI for marketing,' ultimately drawing 'scamming' accusations due to poor user experience.

This approach of repackaging traditional businesses with AI kept Gaotu perpetually outside AI education's core value proposition.

Resource Misallocation

Financial data exposed the hypocrisy of Gaotu's AI strategy.

In Q3 2025, sales expenses skyrocketed to 873 million yuan (55.29% of revenue), while R&D spending plummeted 14.21% YoY to just 163 million yuan—less than one-fifth of sales expenses.

This 'marketing-heavy, R&D-light' structure directly contradicts AI technology's development patterns.

The core competitiveness of AI education stems from large model training, data accumulation, and algorithm iteration—all requiring sustained, heavy R&D investment rather than short-term scaling through celebrity endorsements and channel spending.

Contrasts with competitors are stark. TAL allocated 22% of Q3 revenue to R&D, developing its proprietary 'MathGPT' large model while building a K12-wide question bank and academic performance database.

Zuoyebang invested 18% of revenue in R&D, achieving 98% accuracy in KuaiDui AI's problem-solving through millions of authentic test questions and continuous algorithm optimization. These investments translated into competitive moats.

TAL's AI Learning Device boasts a user renewal rate exceeding 60%, while Zuoyebang's KuaiDui AI surpassed 10 million monthly active users. In contrast, Gaotu's AI products garnered mere dozens of app store reviews, with near-invisible engagement.

More critically, Gaotu remains trapped in traditional education's 'labor dependency.'

Q3's primary business costs jumped 24.6% YoY, primarily due to expanding teams of lead instructors and tutors, with labor costs remaining stubbornly high.

In an era where AI enables 'thousand-person, thousand-face' personalized instruction, Gaotu's increased reliance on human teachers not only raised costs but also ran counter to trends in customized learning, preventing the formation of a viable business model.

Organizational Constraints

Gaotu's AI predicament fundamentally stems from a 'path dependency' crisis typical of traditional online education giants.

From its inception, Gaotu rapidly scaled through a 'master teacher-led + low-cost acquisition' model, surpassing New Oriental's market cap post-IPO in 2019. This success bred organizational inertia around 'relying on star teachers, prioritizing marketing, and neglecting product'—a formula that became its greatest obstacle in the AI era.

New Oriental's transformation through Dongfang Zhenxuan succeeded precisely by breaking its 'master teacher delivery' path dependency, merging educational DNA with content e-commerce to create new growth engines.

In contrast, Gaotu's transformation remained trapped within traditional education frameworks: after policy shifts, it pivoted to quality-oriented education and adult learning, then hastily followed New Oriental into livestreaming with 'Gaotu Jiapin' before shutting it down due to lack of core competitiveness.

Its current AI bet merely reverts to 'empowering master teachers with AI'—essentially refusing to abandon existing faculty and curricula, lacking the resolve for radical transformation.

Organizational weaknesses further amplified strategic flaws.

Gaotu's leadership, predominantly from traditional education backgrounds, understands AI only at a 'tool level,' lacking capability for AI product strategy design. Its sales-driven evaluation system continuously diverts resources to marketing, starving R&D teams of budget and decision-making power to build technological barriers.