What helps Duolingo is its ability to quickly generate course content based on input instructions. Duolingo uses AI to generate scripts and audio for DuoRadio and Adventure. According to founder LuisvonAhn, DuoRadio courses previously took decades to prepare. With AI, it takes just a few months.Duolingo has been using AI technology in content production since 2016. The Duolingo English Testing Team (DET) has been using AI technology since 2016. According to the DET team, the company's English test saves more than 50 percent in time and cost compared to traditional paper-based tests. In 2021, the company partnered with OpenAI to generate a large number of different example sentences and test questions using a generative AI-GPT-3 model. In early 2024, Duolingo will further increase its use of AI by reducing the number of contract workers responsible for translation and content creation, increasing the percentage of generated AI in these areas, and shifting the focus of its remaining contract workers to overseeing and controlling the quality of the AI output. In 2023, Duolingo Max, a more advanced subscription service based on generative AI, will be launched. On top of all the features of Super Duolingo, Duolingo will add two features, Explain My Answer and Role Play, bringing Duolingo closer to the role of a human mentor. More interesting, more attractive. Explain My answers are AI-generated in-depth parsing to help users understand mistakes made during the exercise. Roleplaying offers users the opportunity to interact with Duolingo characters, helping users improve their conversational abilities, such as using AI to guide users in learning how to order coffee or ask for directions, and giving feedback on their progress. Video calls added to Duolingo Max 24. Users can practice their conversation skills via video calls to a purple-haired teenage girl character named Lily in a stress-free environment. Lily remembers past exchanges, which made the learning experience more personal and brought the characters to life.

As a mature AI education product model, Duolingo has the following key elements for its success: first, it uses the massive learning data created by massive users to conduct timely and in-depth data analysis and machine learning, trains its own vertical AI model, and realizes personalization externally, improves service quality and temperature, and improves efficiency internally. Second, AI products have added value, users are willing to pay for better services, and have a stable profit model, such as subscription/membership, courses, etc. The third is to encourage users to maintain the motivation of continuous learning, such as solving immediate needs, stimulating interest, providing emotional value, etc.

Education technology companies in China can learn from Duolingo's example of AI applications that fully embrace AI. First, generative AI is deeply integrated into the content production system to realize the agile development and large-scale supply of educational resources by automatically generating course scripts, dynamic question banks and interactive scenarios. The second is to build a closed loop of "algorithm optimization-user feedback", such as Duolingo's birdbrain, which continuously trains models based on learning behavior data to achieve accurate matching of personalized learning paths. Third, use A/B testing mechanism to drive product evolution and upgrade products around retention rate, conversion rate, update rate.