Original · Unique Research · 2026-04-29
Editor's note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the opening essay and full panel, including all named speaking turns and their continuation paragraphs. Revenue, conversion, user, efficiency, customer and market figures are source or speaker claims, not independently audited findings. The prediction that ChatGPT, Gemini and Doubao will disappear is the speaker's forward-looking judgment, not an established fact. The "牛马" (workhorse / beast of burden) term is colloquial Chinese slang; its English rendering preserves the speaker's tone. "Design for not human" retains the source's English phrasing. Company, personal and work titles are transliterated where official English forms remain unverified. The source is dated April 29, 2026.
Unique Awards
The Content Inflation Era: Stop Using AI to Manufacture "E-Waste"
If you're still doing marketing with traditional traffic thinking, you might not even know how you got eliminated.
"
With the same tools, why do some people see first-week conversion rates soar to 70%,
even selling over a hundred thousand US dollars in a month;
while others can only frantically compete on price in a homogeneous red ocean?
Last year, over half of global content was already AI-produced.
When everyone can generate hundreds of product-short videos at home at zero cost, I've heard a very strange disparity:
Some people use simple AI tools to run accounts, with first-week conversion rates soaring to 70%, even selling over a hundred thousand US dollars in a month; while others, clutching dozens of digital-human matrices, can only frantically compete on price with peers in a homogeneous red ocean.
The same tools—why is the gap so big?
At the Hong Kong Global Unique Awards, a roundtable moderated by Wei Fangdan, founder of Baijing (White Whale) Overseas, several veteran overseas marketers pierced through this window paper.
In this "inflation" era where AI-generated content is everywhere, if you're still doing marketing with traditional traffic thinking, you might not even know how you got eliminated.
Beneath Content Inflation, Are We Manufacturing "E-Waste"?
Actually, AI's impact on the marketing circle is nothing new.
Chen Erhang, Senior Vice President of Yixuan Technology, recalled that around 2017, automated marketing began gradually replacing the media-buyer positions in the Google ecosystem. Now that large models have arrived, the automation of content production has finally been补齐 as well.
These two wheels together are what everyone calls "dual-wheel drive."
But is running fast necessarily good?
Tony Sun Qian, co-founder of FOSHO, made a vivid analogy: "In the AI era, content has started to inflate. When everyone can make content at home with AI, what kind of content is actually effective?"
This is also a headache that several guests on stage are dealing with.
Many people have an intuitive feeling when scrolling short videos: why is the same workplace套路 everywhere, just performed by different people?
Chen Lingshan, Director of Investment and Financing at VidAU, pointed out sharply that the root of homogenization is that everyone feeds the same kind of corpus to large models.
If you only treat AI as a cost-reduction and efficiency-enhancement assembly-line tool, regardless of what the market's real pain points are, what you end up generating is nothing but a pile of unwatched "e-waste."
The Key to Breaking Through—Nobody Tells You
Since traffic is already so competitive, how can overseas brands fight their way out?
The solutions the guests offered completely defy convention.
I think the hardest-core one is the barrier of data and resources.
Everyone uses large models—why should your results be better? Tony's answer was straightforward: rely on non-public commercial data.
Their Agent for overseas major-media partnerships can finish two to three people's weeks of market research and media scheduling in minutes. This isn't because the large model is that much smarter than others, but because they hold exclusive resources and historical投放 data from major overseas media.
This is also Chen Erhang's view: enterprises must rely on their own accumulated historical data and fine-tuned small models to avoid "thousand people, one face."
Besides data, Chen Lingshan offered a more business-practical perspective: closed-loop validation.
She shared VidAU's approach: after using AI to generate hundreds of pieces of material in a day, it's never just a matter of sending them out and being done.
They throw the materials into an intelligent delivery system to run. Once costs exceed the threshold, it automatically cuts off; when three or four viral pieces emerge, they pour all resources into them.
This is essentially a survival-of-the-fittest funnel. Without this validation mechanism, no matter how much content you produce, it's just busywork.
The Most Cutting Truth—Products Will Soon Be Built for "Non-Humans"
Throughout the entire discussion, what made me gasp the most was the发言 of Wang Ming (Winter), founder of K2 Lab.
He directly threw out an explosive judgment: "ChatGPT will disappear, Gemini will disappear, Doubao will disappear. All the To C traffic entrances we see now will disappear."
His logic is that the future world will be fully automated, and centralized traffic distribution will be completely dismantled.
"In the future, products will definitely be 'Design for not human.'"
This means that future consumers will no longer browse web pages or watch ads themselves—they'll let their own Agent (digital persona) scour the entire internet for information.
In this decentralized future, there is no such thing as a "homogeneous viral hit." Your content must be based on the user's current Context, generated in real time by their Persona through encrypted exchange.
Even more cutting was a test he shared next.
They build Agent OS for overseas creators, and found that foreigners simply aren't willing to sit in front of a screen for months grinding to learn sales and editing like we do.
"They think life should be about going to parties, going on vacation… they're not 'workhorses' like us," Wang Ming joked.
The result? After handing all this dirty, tiring work to AI for end-to-end automated processing, the top creator lay down and sold US$130,000 in a month.
Human Warmth Has Become the Most Expensive Luxury
After listening to this roundtable, my biggest takeaway is: no matter how fast the dual-wheel drive goes, the steering wheel is ultimately in human hands.
As Wei Fangdan summarized, everyone has the same production tools—if you don't用心 understand users, involution is inevitable. Even very subtle differences are extremely worth pursuing.
When future machines can handle 99% of copywriting, video editing, and ad placement, the remaining 1%—your resonance with user emotions, your unique experience, your original gameplay—will become the decisive factor between life and death.
The colder and more efficient machines become, the more that bit of "warmth" and "heart" belonging to humans becomes the most expensive luxury.
So, when future consumers all come shopping with their Agents, is your product ready to communicate with "non-humans"?
More Conversation Details
Hong Kong · Global Unique Awards Trends Roundtable Panel
Theme: Traffic Rules: Dual-Wheel Drive of AI Content Industrialization and Marketing Automation
Guests: VidAU Director of Investment and Financing — Chen Lingshan; K2 Lab Founder — Wang Ming; Yixuan Technology Senior Vice President — Chen Erhang; FOSHO Co-founder — Tony
Moderator: Baijing Overseas Founder & CEO — Wei Fangdan
Wei Fangdan: The topic we're discussing today is about the AI part, and the segment assigned to me is called "Traffic Rules: Dual-Wheel Drive of AI Content Industrialization and Marketing Automation." Our company was founded in 2014, and we've long focused on Chinese companies going overseas. We also run our own conference called the Global Traffic Conference, so I feel I have some authority on today's topic. First, let me ask our four guests to briefly introduce their respective businesses and what specific things they're responsible for at their companies. Let's start with Ms. Chen.
Chen Lingshan: I'm Chen Lingshan from VidAU. VidAU has actually gone through the entire marketing stage—from 1.0, 2.0 to the current 3.0 stage. Because we've been around a long time, we're considered veterans of the marketing industry. At the beginning, we mainly empowered overseas enterprises, helping them with overall overseas digital marketing solutions. In the 2.0 stage, we introduced some AI elements, especially on the content side—we used AI marketing empowerment to improve the efficiency of our own marketing content. In the 3.0 stage, we integrated the entire automated delivery process to feed back into the intelligent marketing system. So our company overall does intelligent marketing solutions.
Wang Ming: My name is Winter, from K2 Lab. Our company is in Hangzhou, and we mainly do Agent OS for overseas influencers to directly help them—it can be considered somewhat related to marketing. We believe that in the future everyone can make money through online content platforms, so we build an Agent OS for everyone to enable them to make money. I'm the company founder and basically don't manage things at the company.
Chen Erhang: I'm Chen Erhang from Yixuan Technology. Yixuan Technology provides export marketing services for Chinese foreign-trade enterprises. We've always focused on building the foreign-trade independent-site ecosystem, providing enterprises with comprehensive solutions for foreign-trade independent-site construction and marketing. Our company has about 30 service centers and branches in key coastal cities nationwide, serving over 10,000 Chinese foreign-trade enterprises. Many of the companies exhibiting today are actually our target customer group.
Tony Sun Qian: I'm Tony Sun Qian, co-founder of FOSHO. FOSHO is also an AI marketing overseas service provider. We use AI agents to help Chinese overseas brands achieve affiliate marketing sales growth overseas, as well as major-media partnerships overseas. "Major media" doesn't just mean traffic placement like Facebook, but traditional overseas media with significant influence. For example, in the tech 3C industry, CNET and Wired; in the lifestyle field, major media like Vogue and GQ. We've developed a set of AI agents based on proprietary data to help enterprises achieve marketing and sales growth overseas.
Wei Fangdan: Today our four guests are very authoritative on traffic, with rich experience, and everyone's fields differ somewhat. I've designed several questions to discuss with our guests. The first is:结合 your company's specific business, how do you understand today's theme—"Dual-Wheel Drive of AI Content Industrialization and Marketing Automation"? Which of these two factors do you think is more important, or which has taken more of your time? Let's start with Ms. Chen.
Chen Lingshan: As our company just mentioned, from stage 1.0 to 3.0, we actually do both content generation and backend intelligent marketing. For us, when we first started doing content, why did we need to do backend automated delivery? Because although doing content could improve efficiency (originally it might take three to four days to produce only a few pieces of material, later we could produce hundreds a day, and would also produce based on traffic hotspots), we always had a pain point—we didn't know what the delivery效果 of these produced materials actually was. So to more accurately get terminal feedback, we built an intelligent delivery system. We divide the 500 generated pieces of content into different material packs to see their delivery效果. When a certain piece of material's delivery cost exceeds a certain value, it automatically stops delivery of that material and reallocates resources to other materials. Ultimately, three to four viral pieces of material may emerge, and we can pour all resources into them, forming a good positive cycle. So this is a process where content industrialization drives first, then marketing automation is done later.
Wang Ming: The moderator is famous, your organization is impressive. Here's how I understand this. I think the first thing is that the world is about to become automated—if you don't do automation, you can't even connect to this world. So never mind whether it improves efficiency; my feeling is that without automation, it means you might not appear in the next era. The second point is from the content industrialization perspective. I don't understand marketing; my feeling is that when we do content, we're "borrowing momentum" or "creating momentum," building势能 based on understanding users. Today a very powerful system has emerged in the world, based on the large-model ecosystem, making it possible for us to achieve language understanding, multimodal content understanding, and even batch creation. So many things we previously considered creative will today be deconstructed and industrialized. In the future, products will definitely be "Design for not human." This means we need a much larger volume of content, and don't need to单纯 pursue some absolute super-top viral hit—in the future, super-viral hits won't be able to influence public opinion and trends either. Because after traffic decentralizes, centralized traffic entrances will also disappear. Like I just communicated with the audience: ChatGPT will disappear, Gemini will disappear, Doubao will disappear—all the To C traffic entrances we see now will disappear. The result of disappearance is that we need a larger content scale to truly understand users and give them what they need, rather than simply fooling them. So industrialization is, on one hand, that technical capability is already available, and on the other hand, it's driven by demand.
Chen Erhang: Actually, the "dual-wheel drive" metaphor is very apt. In the process of helping enterprises do overseas marketing, we mainly do two things: one is producing content, the other is doing marketing delivery. Before AI came out, these two things were basically done by people, which also constituted the core competitiveness of service providers. Historically, the development progress of these two wheels was actually not the same. Automated marketing developed slightly faster than content production. Our company started helping enterprises do automated marketing around 2017, basically among the earliest batch of service providers in China doing marketing automation for Google. At that time, there was a very common position in the industry called SEM. Later, after automation came out, this position gradually disappeared, and even dedicated media-buyer positions in enterprises with large ad spend were replaced by system automation. After AI came out, this trend became even more obvious. On content production, progress was slightly slower, but now the trend is also that a large amount of content is being produced by AI. The relationship between these two is indispensable. We often say "thousand people, thousand faces"—if one wheel is missing, it becomes "one person, thousand faces" or "thousand people, one face," and they can't match. So both must achieve automation and AI-ification for the entire marketing to have higher efficiency.
Tony Sun Qian: I very much agree with what Mr. Chen said. Let me add a bit. From the essence of marketing, it's actually how a brand or product reaches people, and how people can more conveniently find the brand and product. Material content generation is equivalent to "means of production," while automated delivery and marketing belong to "productive forces." Our company actually doesn't do content generation. As an observer, my understanding is: in the AI era, content will inflate. Top institutions like VidAU are doing content generation, while at the same time every person and every brand is also at home trying to make content with AI. With so much content generated, in a time of inflation, what kind of content will be effective in the future? I don't have the answer to this question, but it may be a challenge. From the automation perspective, what our company does is marketing automation, including automated operations of affiliate marketing, and using agents to automatically generate and下发 plans for major-media cooperation and placement. To borrow Nongfu Spring's words: we don't produce content, but we are the porters of content.
Wei Fangdan: I particularly agreed with what Tony just said—in the AI era, everyone can generate a large amount of content, and much of it is very high quality. A lot of content you see on platforms like Hongguo Short Drama—some things ordinary people shoot can rival big-name directors. This leads to my second question: many enterprises find when落地 that the content we generate is severely homogenized. For example, when scrolling Douyin, you find many workplace套路 are the same, just shot by different people. And automated delivery is extremely prone to falling into low-price involution—after all, everyone can do automation, and Google, Meta and other companies are all doing it too. How do we prevent our dual-wheel drive from becoming "dual-wheel involution"? Where does the real differentiation lie? Is it in technology, business model, or something else? Hope everyone can share their views.
Chen Lingshan: I think the question you just raised is very good, and it's something our enterprise has been thinking about. The reason we fall into homogenized competition or price wars is that we've produced homogenized content. And the reason we generate homogenized content is that we may have fed the same kind of material to the model. To improve differentiation, the core is still whether our automated delivery system can capture what users truly value, convert it into marketing language, integrate it into the content, and thereby generate high-quality content that differs from others. Then we test efficiency and allocate resources through the delivery system—this kind of process can effectively avoid homogenized competition. Another point is that we need to see whether this is truly a business pain point. Many times we make homogenized things just to use them as efficiency tools, not truly treating them as an industrialized process. Things produced by industrialization must have a market, not just solve process efficiency problems.
Wang Ming: I think today's problem is actually more severe. As you may know, last year over half of global content was already produced by AI (non-human). Our company's business is doing sales content for overseas influencers. The reason we value this track is not only because overseas is a traffic depression, but also because overseas MCN is not easy to do. Foreign creators are unwilling to sit at a small desk for two months grinding to learn sales like Chinese people—even bloggers with millions of followers are unwilling to sit there livestreaming for two hours every day to make money. They think life should be about going to parties, going on vacation, let alone doing creative and editing work with various AIGC tools—they're not "workhorses" like us. So combining AI's end-to-end automation capabilities, we did a small test last month, and the top creator already sold US$130,000 in a month.
I believe this world will be flooded with AI content, which means "human value will be amplified." In the future, when everything you see is of unknown authenticity, your ability to understand user Context will determine that everyone can produce personalized sales content in the future. AI has extremely strong real-time generation capabilities, and the Context you provide can produce immediate changes in marketing content. In the future, you won't even be able to make centralized homogenized content. The current homogenization exists because platforms are still dominated by centralized traffic, but products like ChatGPT already understand users very well and can recommend highly personalized content. In the future, as decentralized traffic blooms, everyone will have something like OpenClaw or an Agent OS. This means they have fully authorized their Persona (digital persona) on Context, and when obtaining content, their Persona will engage in encrypted exchange on various Servers. The content they obtain must be customized to themselves and generated in real time. So previous marketing methodologies will become basic skills in the future, and homogenized content is unlikely to exist.
Chen Erhang: From another angle, homogenization or involution—actually it wasn't brought by AI; it has always existed before, but AI has accelerated the process. We can look at it from two aspects: content and marketing.
Content homogenization happens because everyone uses the same underlying large models, the same corpus and tools, so the output content is definitely the same. To achieve differentiation, enterprises can start from several aspects: First, on top of general large models, have their own fine-tuned small models, adjusted结合 the enterprise's product situation. Second, use the enterprise's unique historical data and user behavior characteristics for training, which can produce different content. Third, like our company which started in 2009, we have our own data structure and knowledge base—this is also a manifestation of differentiation. So under the general trend of AI industrialized production, enterprises fully have the opportunity to achieve content differentiation.
The involution of marketing automation is essentially that everyone takes identical content and competes for the same traffic on the same channels, naturally只能 competing on price. But the combination paths that affect traffic and conversion are different. If you have unique points in content, channels, etc.—for example, different platforms have different conversion rates for the same content—that's your competitive advantage. For a single enterprise, the trial-and-error cost is high, but with the help of a service provider platform with historical沉淀, you can better跳出 involution and fight your way out.
Tony Sun Qian: Mr. Chen is relatively optimistic. From our startup perspective, the core differentiation still lies in data and resources. As an application-layer company, it's hard to say technology leads the world by much. Why can we stand out? Take our overseas major-media cooperation agent "MIA AI" as an example—say a brand has a budget of US$500,000 to do a new product launch overseas, and the target audience is set. Which media and which ad slots should they buy? In the traditional model, this requires two to three people spending two to three weeks doing market research and producing a Media Plan. Now our Agent can complete 80%–90% of their work in minutes. General large models can't do this well because we hold unique overseas major-media resources and non-public commercial data. That's where we can provide unique value. Also, although involution is severe now, AI development has also greatly reduced the time and cost of brand trial and error, which is also a good thing.
Wei Fangdan: Thank you to the four guests for your answers. Regarding involution, my personal view is that this is a problem all companies will face together in the future. Everyone has the same production tools—if you lack innovation or don't用心 understand users, involution is inevitable. But if you spend more time understanding users, even if the products are similar, you can make differences in marketing methods, reach methods, and product emotional attributes, making your enterprise appear more vivid. There's no need to force absolutely huge differentiation—even very subtle differences are something very worth pursuing.
Today we discussed content industrialization and marketing automation. The guests generally agreed that the two are complementary, a dual-wheel drive relationship that spirally promotes each other. Just like a car, dual-wheel drive makes it go faster; one wheel just spins in place. Regarding involution, everyone also shared their own ways to break through. Thank you very much to the guests for your time. Our roundtable concludes here. Thank you!