ByteDance's product Gauth has become popular overseas. What are the secrets for AI applications to go global?

As the AI ​​​​wind blows, ByteDance's educational product Gauth is taking off overseas.

According to the Qimai list, as of April 15, 2024, Gauth's iOS mobile version ranked fourth in the US education app category. On April 8, the app was ranked second, second only to the language learning app Dolingo.

According to Data.ai, Gauth has been downloaded 147,400 times and has 449,000 active users. The data site gave the app a Mobile Performance Score (MPS) of 98 out of 100.

As a math teaching product launched in 2020-2021, Gauth was not very prominent in the education app rankings until the beginning of this year, when it achieved impressive results. According to Qimai data, in three months, Gauth's global downloads increased 14 times, and its daily active users increased 1.4 times.

It is reported that Gauth is an artificial intelligence-driven learning tool that can provide step-by-step solutions to problems in education subjects such as mathematics, chemistry, biology, and physics. According to the evaluation, Gauth's math answering level has caught up with that of American ninth-grade students, and its comprehensive Chinese level is also very high, comparable to that of excellent candidates for the American SAT level test.

Sitting on the cusp of the trend, Gauth has taken off, but AI is just one of the “mouths” needed to get a piece of the overseas market pie.

Photographing to solve problems AI can help attract more educational resources

Gauth's explosive growth occurred at the beginning of this year. According to Qimai data, starting from January 2024, Gauth has emerged as a dark horse in the US and even the global education application market, with a total number of users exceeding 200 million, following the popular language learning application Duolingo, and once ranked second in the US education application (free) rankings.

Gauth ranking performance in the past year as recorded by Qimai data

According to Data.ai, Gauth's daily active users (DAU) in the US market soared from 505,600 in January to a peak of 1.39 million. Globally, the DAU peak reached 2.065 million, an increase of more than 1.4 times compared to the beginning of the year. At the same time, the number of daily downloads soared to 14 times the original in early April.

It can be said that Gauth not only performs strongly in the US market, but also makes its mark in the international market. Why has it become so popular?

In fact, Gauth was officially launched in December 2020. At that time, it was still called Gauthmath. The operating entity was GAUTHTECH PTE. LTD., which was registered in Singapore. The shareholder was LEMON INC., a subsidiary of ByteDance. Until now, you can still see the advertisement of ByteDance recruiting AI project interns for Gauthmath on the Internet (the position has been removed).

This learning tool had two main points when it was first launched: overseas and AI.

When it was first launched, Gauth was not very popular. At that time, educational applications with the function of taking photos to search questions were not uncommon in the domestic market, and many technology companies in the education industry were also promoting similar products. This may be the reason why Gauth did not expand domestically but chose to go overseas. However, in the overseas market, it was not as popular as it is now, and it was unknown in the ranking of educational applications.

In 2023, generative AI became popular under the impetus of ChatGPT. Gauth, which already combines AI functions, found its "fulcrum" and started its counterattack.

Users can take a picture of the question with their mobile phone, and Gauth will analyze the question and provide detailed answer steps and problem-solving ideas in a few seconds with the help of AI. According to the official website, Gauth covers multiple disciplines such as mathematics, statistics, physics, chemistry, history, etc., and even reaches the level of the US SAT exam in terms of language and comprehensive ability.

AI is not only the basic technology of Gauth, but also a trend that it can take advantage of. But Gauth is not only about AI for users. This tool also connects a professional tutoring team, including teachers with many years of teaching experience, as well as college students and graduate students, and builds a global community for students to discuss issues.

If you test it yourself, you will find that AI is a technology that can answer most academic questions, but there are always some deep and tricky questions that even AI cannot answer. Some testers uploaded domestic college entrance examination questions and found that Gauth can recognize and answer them step by step in Chinese, but the answers are wrong.

Mathematical problems have always been a difficult hurdle for various large natural language models to overcome. As an application, it is inevitable that Gauth will have such bugs, but it has made up for this shortcoming with "people" such as expert groups and communities, and also laid the groundwork for its own commercialization.

The Gauth App can be downloaded for free, but some features, especially those with “real expert team” such as teachers solving problems, require a paid subscription. These features are included in the Gauth Plus membership version, which, in addition to unlocking all answers, also provides 24/7 teacher Q&A service.

In the United States, Gauth Plus subscription fees are divided into three tiers: monthly, quarterly, and annual, which are US$11.99 (approximately RMB 86.78), US$21.99 (approximately RMB 159.15), and US$99.99 (approximately RMB 723.66).

It seems that in the commercialization of Gauth, AI is not only a technical capability but also a signboard for attracting traffic. The combination of AI technology with real-person teaching resources and rich user experience also explains the value of people beyond "AI-driven tools."

When AI goes overseas, will it definitely be a blue ocean?

As AI becomes increasingly popular around the world, domestic application-oriented enterprises have begun to seek to expand overseas at the stage of technology conception, hoping to establish a presence in overseas markets as soon as possible. This seems to be the experience that Chinese enterprises that are good at making products have gained in the entire wave of mobile Internet development.

Starting from 2010, under the wave of mobile Internet, Chinese enterprises tended to export their applications overseas. Games were the pioneers in this field, represented by Mihoyo. Then, cross-border e-commerce emerged, and applications went overseas in the form of platforms, providing thousands of products for export. Tiktok became a leader. Later, brands also had a place overseas. When Chinese people saw Mixue Bingcheng in Southeast Asia, they could not hide their affection.

Today, this trend of going overseas has also reached the AI ​​track. Many AI products are thriving in overseas markets. Domestic AI applications such as Heygen, Opus Clip, Meshy, and PixVerse are constantly reporting their “overseas popularity”.

Baidu has launched the AI ​​camera Meira, AI chat SynClub and AI social WiseAI for overseas markets; one of the co-founders of HeyGen, which became popular overseas last year, is Liang Wang, a former product designer at ByteDance; and the products brought by many individual entrepreneurs and teams have also performed well in the overseas market. TheB.AI, FlowGPT and other popular AI native platforms were all independently developed by the founders in the early stages of the project.

There are three main reasons why domestic AI players are rushing to go overseas: first, foreign users are more willing to pay; second, the English processing level and open source nature of domestic large language models need to be improved. Overseas products mainly access the APIs of large models such as GPT, Claude, and Bard, starting from vertically segmented scenarios and then encapsulating scenario-based functions, which can quickly realize the launch of small products.

However, rejecting internal competition and going directly overseas also means that opportunities and challenges coexist.

The popularity of AI going overseas is heating up and the competition is fierce. Not all products can be successfully commercialized. Even for large companies, it is difficult to quickly achieve good results without rich user resources accumulated overseas.

Although Baidu launched Meira, SynClub and WiseA, the market response was not optimistic. SynClub's application ranking was outside the top 100, and the number of downloads was only in the thousands. The daily active users of the currently outstanding Gauth surged, and the daily active users of the language learning software Dolingo were already in the tens of millions.

Take Dolingo as an example. Its Max version has added AI-based role-playing and intelligent error correction, which is also after the rise of AI. In the fourth quarter of 2023, Dolingo's revenue was US$151 million, a year-on-year increase of 45%; net profit was US$12.11 million, a year-on-year turnaround, and net profit increased by 187%. This is the third consecutive quarter that Dolingo has achieved profitability. The fact that the application can stand firm in the ranking of educational applications does not rely on AI to a large extent, but more on years of market accumulation.

Dolingo CEO Luis von Ahn once said that AI education should focus more on the development of the application layer and the C-end user experience. Let AI become the "door-knocker" of education, rather than letting education companies get lost in the halo of technology and ignore the product itself because of the volume model.

AI technology has indeed brought unprecedented opportunities for products to go global, but the focus of product development is still the product, not simply "+AI". In a homogeneous large-scale market, how to use AI technology to make good products is the key to success in going global. The rise of Gauth has sent a positive signal to the market and also demonstrated what AI products should look like: falling into application scenarios and returning to user experience.