《鄭哥點評：AI探店不到10元一條，800萬月流水的生意，賣的是什麼？》Zheng Ge's Take: AI Store-Visit Videos at Under 10 Yuan a Clip, an 8-Million-Yuan-a-Month Business — What Is It Really Selling?

[Zheng Ge's Take] Column: Zheng Ge's blog commentary | Date: 2026-10-28 | Keywords: AI store visits, visiting without visiting, 8 million yuan a month, false advertising, AI labelling rules

What the report actually says

What happened: on 26 October the Securities Times ran a front-page report. Now that AI video generation has matured, a "store visit" no longer requires going to the store.

The key numbers (measured by the reporter):

• AI store-visit tutorials on Xianyu: from 0.18 yuan, typically around 1 yuan
• Tool credits: 1 credit = 1 yuan, minimum top-up 20 yuan
• Annual membership: 299 yuan a year
• One 12-second 480P store-visit video: 9.6 credits = 9.6 yuan
• Monthly revenue of a leading bulk-production company: 8 million yuan
• Agency quotes: milk-tea and snack shops 300-500 yuan a video; hotpot, barbecue and beauty 600-800 yuan
• Cost of one traditional store visit: in the thousands of yuan

The workflow: upload the shop interior + product close-ups + model photos → add the shop's information → the script writes itself → choose model / aspect ratio / length / resolution → one click and the video is finished. No real person on camera, no copy to write, no visit to film.

The risks:

Ye Weiming, associate professor at Peking University HSBC Business School: "The core value of a store visit is the authenticity of the experience. The assessments and recommendations of an AI digital human fundamentally run against real human consumption needs."

The Hangzhou Intermediate People's Court ruled in June this year: AI ghost-writing "seed-planting notes" constitutes unfair competition; the defendant stopped the infringement and paid damages.

The Measures for Labelling AI-Generated Synthetic Content took effect on 1 September 2025 — text, images, audio and video generated by AI must carry, in a prominent position, a label the user can clearly perceive. But on the short-video platforms, "the rate of voluntary labelling is not high".

Huang Taiyu, a lawyer at Guangdong Lianyue (Dongguan) Law Firm: where the blogger never went at all and simply fed a merchant's own footage to AI to generate a complete "experience", this may constitute the crime of false advertising under Article 222 of the Criminal Law; using AI face-swapping, or cloning a well-known person's voice and likeness, also engages portrait rights and voice rights.

The red line in the rules: AI tools should serve technical steps such as copy polishing and editing. They cannot replace the authenticity of a real person's visit.

Zheng Ge's take: the ledger this report never balanced

The report itself is solid — the reporter's own testing, interviews with lawyers, academic opinion, a court ruling, the regulatory rules: all of it is there. But it turns one thing into "technology disrupting cost", whereas I think it is actually "the industrialisation of trust" — two different things, and the two ledgers have to be kept apart.

1. The cost ledger: between 9.6 yuan and 600 yuan, the gap is not computing power

The reporter measured 9.6 yuan a video; agencies quote 600-800 yuan. The 500-700 yuan in between is not a computing-power premium — the computing power is worth that 9.6 yuan and nothing more.

So why will a merchant pay 600-800?

What they are buying is distribution and trust endorsement.

• Distribution: an agency has matrix accounts, creator distribution channels and paid-traffic experience. Once the video is out it can enter the feed, be pushed by the algorithm, and drive group-buy conversions.
• Trust: it arrives in the shape of "a creator's visit and review", so what the user sees is "a real person's review", not "the merchant praising himself".

The AI video itself is worth 9.6 yuan. It is worth 600 yuan only once it is packed in the shell of a "creator review".

And that is exactly the problem — if an agency generates everything with AI, gross margin approaches 100%; and a merchant doing it himself spends almost nothing. So why pay 600? Because what he wants is not the video but the identity of "a creator account". AI made the least valuable thing (the video itself) cheap, and in doing so exposed the most valuable thing (the creator's trust premium) as a packaging fee with a price tag on it.

The report skips over that ledger.

2. The market ledger: what 8 million yuan a month actually means

8 million yuan a month, at the hotpot tier's average of 600-800 yuan, means 10,000-13,000 videos delivered in a month.

300-400 a day.

What kind of number is that? It sits several orders of magnitude above the capacity of any traditional store-visit studio in the country — a traditional creator has to travel to the shop, shoot, edit and push, and can turn out one video a day at most; two on a fast day.

8 million yuan a month is not "bulk production" scale. It is "video factory" scale.

Once factory logic walks in, the whole trade of the "store visit" is rewritten by industrial logic:

• Scarcity is gone — a real person's visit was the scarcest and most expensive asset of all; now it is mass-produced for 9.6 yuan.
• The threshold is inverted — one real store visit costs thousands; one fake AI store visit costs 9.6 yuan. A rational merchant picks the latter, obviously.
• Bad money drives out good — real store-visit videos are drowned out by fake AI ones; users can tell the difference less and less, and believe less and less.

This is what Ye Weiming calls "traffic arbitrage and false prosperity" — AI did not make store visits better; it made faking cheaper than the faking that came before it.

3. The legal ledger: the line is drawn, enforcement is still grey

The two rulings and the one regulation in the report are all real, and all in force:

• Hangzhou Intermediate People's Court (2026-06): AI ghost-writing of seed-planting notes constitutes unfair competition; infringement stopped plus damages. It is the first case of its kind in the country and the signal is clear — "generating a review without ever visiting" is itself already unlawful; nobody has to wait for a user to sue the merchant.
• The Measures for Labelling AI-Generated Synthetic Content (in force 2025-09-01): AI-generated content must carry a label in a prominent position.
• Article 222 of the Criminal Law, false advertising: serious cases can carry a criminal sentence.

But the report itself admits: "Although the platforms require publishers to label voluntarily and have launched helper features, the rate of voluntary labelling is not high."

In one sentence: the law exists, the precedents exist, the regulation exists — enforcement has not moved.

Why is enforcement hard? The report actually points at it:

• an agency may have been subcontracted several times over, so the responsible party cannot be found;
• for an ordinary person the cost of defending a right is high — "when you do the sums it may not be worth it";
• pure-text content dodges the keywords, so a platform can hardly judge whether it was AI-generated.

This is the common deadlock of every new content business — the rules run ahead of the regulator, the regulator runs ahead of enforcement, and enforcement runs ahead of what users understand.

My judgement: in the next 6-12 months a second and then a third precedent will come, and there will be a wave of takedowns at the same time. Right now is precisely the window in which "the line has been drawn but nobody has really been hurt by it yet".

4. The business ledger: where the real opening is

The last paragraph of the report actually says the key thing — "AI tools are not unusable, but the principle should be that they serve technical steps such as copy polishing and editing; they cannot replace the authenticity of a real person's visit."

That sentence separates the real opening from the fake one.

The real openings:

1. Editing and polishing for real visits — you go and shoot in person, and AI edits, grades, subtitles and cuts it into several versions. Cost falls from 3,000 yuan to 300, it is still "a real visit", and it still has value.
2. "AI-assisted store visits" for local merchants — the real visit is the core; AI supplies the efficiency (multi-camera assembly in one click, a fast script draft, distribution cut to fit each platform).
3. "AI material authenticity checks" as a service — platforms and enterprises will both need it. It is a defensive business, and demand jumps the moment the law hardens.

The fake openings:

1. Bulk production of purely AI-fabricated experiences — the Hangzhou ruling has already decided this one; it is digging your own hole.
2. Teaching people to make fake AI store visits — the 0.18 yuan tutorial is cheap to sell, but the chain that teaches people to break the law carries its risk on your own head.
3. Agencies that just re-skin AI output — 100% gross margin is a fine thing, but one expose and the whole thing collapses.

In one sentence

What 9.6 yuan buys is not "the industrialisation of the store-visit video". It is "trust as an asset, priced and labelled". On the day the video is worth nothing, the value of actually going is redefined — because it is the one thing that cannot be bought for 9.6 yuan.

Merchants in local life should note: the era of traffic arbitrage is over; what comes next is reputation against reputation. Those building AI tools and services should note: regulators' patience has a floor, and the Hangzhou ruling was already the starting gun.

Sources: Securities Times / 21st Century Business Herald, 26 October 2026 (the original circulated via Jinri Toutiao); Ye Weiming, associate professor at Peking University HSBC Business School; Huang Taiyu, lawyer at Guangdong Lianyue (Dongguan) Law Firm; Ma Anran, lawyer at Beijing Huinuo Law Firm; the Hangzhou Intermediate People's Court's notice of the country's first unfair-competition case involving AI ghost-written "seed-planting notes" (June 2026); the Measures for Labelling AI-Generated Synthetic Content (in force 1 September 2025); Article 222 of the Criminal Law.

Cost calculation: real cost of one AI store-visit video = 9.6 yuan of compute + 1.25 yuan of membership amortisation ≈ 10.85 yuan a video; against 2,000-5,000 yuan for a traditional store visit and 300-800 yuan for an agency.