00:00 Subscribe to listen to this audio When US President Donald Trump waged a trade war on the world earlier this year, it sent shockwaves through global industries—none more so than among Chinese exporters, whose biggest market, despite ongoing diversification efforts, remains the United States. As companies scrambled to navigate retaliatory tariffs and an uncertain future, Xu Kaijie, a Gen Z from the small city of Tongxiang in Zhejiang province, kept his cool. His Etsy store has continued to pull in around 30,000 US dollars a month thanks to a product that’s proven tariff-proof: AI-generated, watercolor-style pet portraits, sold as digital files to customers around the world, including in the US. “AI has already revolutionized traditional industries like cross-border e-commerce. In the past, you couldn’t even enter the game without hundreds of thousands in capital,” Xu tells TWOC. “You had to spend money on inventory, hire designers, find a factory to produce the goods, and then sell them. By the time everything was ready, the opportunity had already passed—you’d missed the wave.” Read more about AI in China: Smart Earth: How AI Is Rewriting Rural China What Booming AI Revival Services Mean for China’s Relationship to Death Will ChatGPT Really Kill the Chinese Translation Industry? Xu is hardly the first—and certainly won’t be the last—to figure out ways to profit from AI in the overseas market. When the free-to-use, China-developed large language model (LLM) DeepSeek launched in early 2025, it further democratized the country’s e-commerce landscape—enabling sellers to quickly generate sales pitches and promotional videos in multiple languages, with a raft of success stories making domestic headlines . Many AI tools created by Chinese developers—like Manus, Flowith, and HeyGen—are also eyeing the global market. Some are already widely used and even majority-funded by overseas investors. According to marketing intelligence platform Sensor Tower, two of the 10 most downloaded AI apps worldwide in 2024 were from China—ByteDance’s Doubao chatbot and the AI companion app Talkie developed by Shanghai-based company MiniMax. However, for your average Chinese, tapping into overseas markets remains an uphill climb—filled with risks and pitfalls that AI alone can’t navigate. And as AI tools become increasingly powerful and accessible, questions regarding the ethical nature of their use are only growing more urgent. When AI sells Just over a year ago, Xu’s life looked nothing like it does today. Having just quit his sales job due to health issues, he was scraping by as a mobile game booster —someone paid by other users to improve their gaming performance—unsure of what came next. Then, in January 2024, he met his mentor, who opened the door to an unexpected frontier: generative AI. Despite having barely any English language skills and never having worked in tech—or even used a computer seriously, having grown up in China’s mobile-first Gen Z culture—Xu plunged into the world of AI headfirst. From his small rental room, he navigated a steep learning curve through 15-hour work days in front of the computer, even during Chinese New Year. Feeling the domestic market was already saturated with content creators, he first tried posting AI-generated videos on overseas platforms like Instagram and TikTok. From aggregated news reels to cute pet videos, Xu experimented with it all. While he managed to attract around 80,000 followers across these platforms, as well as decent traffic, there was no clear path to monetization—aside from about 1,000 yuan in ad revenue overall. An example of an AI-generated pet portrait sold by Xu Kaijie on Etsy. His clients now mainly come from the US, Canada, the UK, and Australia. (Courtesy of Xu Kaijie) “The early stage was like a black hole of despair. From January to July [of 2024], it was all expenses with no income. It really felt like there was no light at the end of the tunnel,” Xu tells TWOC. Beijing native and recent Singapore transplant Max Deng (pseudonym), who posts AI-generated videos of anime characters when he’s not busy with his office day job, echoes Xu’s experience. The 28-year-old also only receives ad requests, which pay around 300 yuan per video. “Some of the inside jokes in the videos also only resonate with the Chinese audience and don’t translate well on international platforms,” Deng adds. He has, for now, paused posting to figure out his next steps. Xu, on the other hand, didn’t give up, instead switching lanes to focus on e-commerce. After doing some market research, he found that artists who provide hand-drawn artworks usually have a long waiting list, as they can take a long time to produce. A single painting can also be sold for thousands of dollars to buyers in Australia or North America. Given the significant traffic his AI-generated animal videos had experienced, he decided to try offering similar, watercolor-style pet portraits on the craft-leaning e-commerce site Etsy, selling them for as little as 10 to 13 dollars each. Lacking any substantial investment, he decided to sell only the digital file—or, if requested, send the design to a third-party printing platform like Printify that handles production and shipping—to minimize the financial burden of operating an online store. While relatively simple and with minimal overhead, such a business model is certainly not a guaranteed path to success. Clip art like this is already a “multimillion-dollar market,” making competition fierce and hard for individual sellers to stand out, according to The Atlantic. From product selection to operations, Xu still had to experiment at scale, review sales and promotional results daily, and constantly refine his approach before finding a sustainable workflow. To accurately replicate the pet photos provided by clients, Xu estimates that he created approximately 5,000 samples with the AI image generation program Midjourney, fine-tuning his prompts and formulas. “It has basically become a style I developed myself. I can now make one with AI in about two minutes,” he says. “Plus some final tweaking in Photoshop.” He has since expanded his product line to include AI-generated custom figurine designs, following the same digital-file-only model. International trade companies are also using AI tools for client outreach and communication (VCG) To this day, Xu still doesn’t know much English. He uses Google Translate to read messages from customers, and when the results seem confusing, he takes screenshots of them and asks LLMs, such as ChatGPT or Claude, to explain them for him. “Replying would take a few extra steps. I’d have to train ChatGPT to be an English translator,” Xu explains. “I’ll input things like ‘You’re an American’ or ‘You’re from Sydney, Australia,’ and ask the AI to respond the way customer service from that region would. Once the reply sounds good enough, I just copy and send it to the customer.” Xu has since turned his experience into a series of course packages aimed at others looking to break into the industry. Prices range from a basic 299-yuan text-only instruction package to a premium one-on-one business-launch guide at 4,980 yuan. With around 200 subscribers, teaching has now become a significant part of his income. In AI, an inhuman touch Not everyone is as lucky as Xu. As the world’s largest manufacturer, China still relies on selling and shipping physical products overseas. And among its sellers, AI is still primarily used to generate product images for their online store and ads. T-shirts and mugs with AI-generated illustrations are among the easiest products to create. A quick search on Chinese social media reveals countless posts sharing experiences of selling such items overseas through platforms like SHEIN, Temu, and Shopify—but many of these ventures fall short. There are still numerous challenges that AI can’t solve, from determining what illustrations might sell to managing inventory and logistics to addressing negative reviews. Even on Chinese e-commerce platforms, many complain about receiving products that don’t match the AI-generated pictures showcased in their description pages. An AI program showcases apparel modeled on generated figures at a Shanghai expo (VCG) For companies offering higher-end products with unique designs, even AI-generated photos have limited value. Shen Ying (a pseudonym), the social media manager of an accessory company primarily targeting the North American and Australian markets, first proposed the idea of using AI to generate model images featuring the jewelry two years ago, but the results at the time were underwhelming. “Our products need to clearly showcase the fine details—and those details vary from one item to another. For example, every ring needs to be photographed from different angles, depending on its design. There are just too many variables, and right now, AI isn’t capable of providing truly customized services at that level of precision,” explains Shen. For now, the only way Shen’s team uses AI is to check the grammar of messages engraved on their jewelry and sentences printed on accompanying gift cards. She has made a point not to use AI to generate social media content after finding that it lacks originality. “If you really want to write something that resonates with people—something that makes a gift feel thoughtful—then, of course, it’s better when it’s written by a real person,” says Shen. “It’s like when you receive a package from a brand: a handwritten, original note always feels more sincere than a printed one.” Drawing the line While Shen is sticking with original content, Dragos Cacio, founder of the Beijing-based marketing consultancy Crelong Media, finds that his clients now demand such high volumes of content that only AI can keep up. “Unfortunately, AI [has enabled] a lot of people to create a lot of content very fast. So, if a company cannot generate the same amount of content in the same amount o