NBD Hot Comment: AI store-exploration videos generate 8 million yuan in monthly revenue, and AI-synthesized content should not undermine the foundation of trust in traffic

Article body

Du Hengfeng, commentator for National Business Daily

"AI store visits" have unexpectedly become a highly profitable new business. Media reports recently noted that some companies specializing in the mass production of AI store-visit videos can generate monthly revenue of up to 8 million yuan; tutorials cost as little as 0.1 yuan each, and a 12-second video costs only 9.6 yuan. With no need for a real person to visit the store, you upload a few photos of the storefront, products and models, and a seemingly professional store-visit video is produced, then immediately released in bulk across various short-video platforms.

These videos are highly similar in content. Almost identical lines: an opening "I finally discovered this shop," a few lines in the middle like "absolutely amazing" and "just go for it," and a closing line "only ×× yuan per person"; almost identical camera work: pushing the door open, taking a seat, close-ups of dishes, raising a glass, paying the bill; almost identical content arrangement: building suspense, showing the process, and recommending the product.

These highly repetitive contents are the product of a single industrialized assembly line. Professional agencies handle production, from processing materials to exporting the finished video in one go; model faces are licensed in bulk packages and become digital assets that can be called upon repeatedly; platform algorithms handle distribution, pushing low-cost content in front of massive numbers of users.

Publishing AI store-visit videos in large quantities is the result of merchants doing the cost math. A real-person store visit requires paying influencer commissions and covering travel and meal expenses, often thousands of yuan per video, or at least several hundred; an AI store-visit video costs only a few yuan or a few dozen yuan, and with the help of platform user scale and algorithmic leverage, maximum exposure can be bought at the lowest cost. Even if the exposure performs poorly, the cost of trial and error is so low it can be ignored—by any calculation, it is worth it.

Merchants save on costs, but someone always has to bear the price, and the first casualty is the trustworthiness of AI-generated content. The entire value of store-visit content is built on the real experience of "personally visiting the store, personally tasting the food, personally feeling the place." Users trust a store-visit review because they believe the person in it really went, really ate, and really has their own sense of taste and judgment. But AI has no tongue and no taste buds; all its generated "delicious," "stunning" and "just go for it" are merely the result of data stitching.

Having an AI with no capacity for experience state experiences that can only be had firsthand is a misuse of AI—AI output should be truthful and verifiable, and must not create misleading false impressions. More worrying still: when AI repeatedly receives instructions to fabricate experience narratives, its goal of serving people is led astray, and over time people's trust in AI will drop sharply.

For consumers, the purchasing decisions they make based on the default trust in "real-person experience" are in fact built on a fabricated experience; the distrust after being burned will spill over into the entire service industry and may even have a negative impact on the whole consumption environment.

For creators, a real-person store visit requires going to the store, tasting the food, distinguishing good from bad, and building up reputation. When such work is drowned out by cheap assembly-line products, people who do serious work may instead be squeezed out of the traffic pool, creating a vicious cycle in which bad money drives out good.

The Provisions on the Administration of Deep Synthesis of Internet Information Services require that deep synthesis content be conspicuously labeled, that facts not be fabricated or tampered with, and that others' portrait rights not be infringed; the Measures for Labeling AI-Generated Synthetic Content further stipulate that publishers should proactively declare and label such content. The rules have long existed; the problem lies in enforcement. Closing the door on deep synthesis content involving fabricated experiences such as AI store-visit videos requires all parties to play their proper roles.

English text is being prepared.

And merchants, as the most direct stakeholders, should be even more responsible for the authenticity of content. This is not about banning merchants from using AI to improve efficiency, but the generated content must be based on facts—especially the experiential parts, which must be truthful and objective statements—while AI can take on auxiliary creative work such as editing and subtitles.

Without a tongue to taste, how can there be a store-visit review? AI can lower the threshold for creation and assist genuine expression, but it should not become an assembly line for mass-producing "store-visit experiences." Merchants may appear to have saved money, and platforms may have produced more content, but the price is that the trust on which the entire content ecosystem depends has been overdrawn.

Short-term gain or long-term prospects—this is not a hard calculation to get right.

(Source: National Business Daily) Body middle Embedded advertisement