Simulated Shopee and TikTok Shop buyers across Southeast Asia and Asia-Pacific rank an accurate seller-shot hero first and ungated AI listing video last. Put real product footage in the hero slot, and hold any AI creative until a buyer-panel check clears it. An AI-generated label alone does not.
Figures marked PAI simulation · directional are simulation outputs, not census or survey estimates. How we label data.
Ungated AI listing video is already a buyer-trust problem, and the blame lands on the seller.
On 25 August 2026, Channel NewsAsia reported that sellers on Shopee and TikTok Shop were calling out auto-generated and affiliate AI product videos that misrepresented what they sell. Reported cases included a warped book cover, a young-adult novel sold as a fun children's adventure, and a calendar shown at the wrong size with stickers in the wrong colour. Sellers said they could not reliably stop the clips, and authors blamed publishers for videos the publishers had not made.
In March 2026 the Advertising Standards Authority of Singapore said false and misleading advertising is unacceptable regardless of the tool used or the disclosure added, and that vague AI labels do not help shoppers tell what is real.
So what: An AI marketing agent that recommends "generate and publish" or "add an AI label" is recommending the two policies this panel ranks last. The check has to happen before the clip ships.
25 Aug 2026report
Secondary sourceOn 25 August 2026, Channel NewsAsia reported that sellers on Shopee and TikTok Shop were calling out auto-generated and affiliate AI product videos that misrepresented size, colour, or the tone of the product.
23 Mar 2026guidance
Secondary sourceOn 23 March 2026, the Advertising Standards Authority of Singapore said false and misleading advertisement content is unacceptable regardless of the tools used or the disclosures added, and that vague AI labels are not useful to shoppers.
7cases
Secondary sourceThe Advertising Standards Authority of Singapore received 7 pieces of feedback in 2025 about advertisements that used generative AI, more than double the 3 received in 2023 and 2024 combined, out of 379 pieces of advertisement feedback that year.
An accurate seller-shot hero leads purchase intent, usability, and credibility. Ungated AI and a badge alone collapse.
The simulated panel scored five policies for the same kind of listing-linked product video. Seller-shot means a short real phone video of the product with a hand or scale in frame. Seller-approved AI ships only after a named person watches the full clip and confirms colour, size, materials, and claims. Validate-before-ship holds the clip until a buyer-panel simulation clears a kill bar.
| Policy | Purchase intent (/5) | Top-two score (4 or 5) | Usability | Credibility |
|---|---|---|---|---|
| Accurate seller-shot hero | 3.99 | 78.3% | 4.41 | 4.55 |
| Seller-approved AI | 3.38 | 51.9% | 3.49 | 3.86 |
| Hold until a buyer simulation passes | 2.99 | 35.4% | 2.70 | 3.48 |
| AI badge only | 2.08 | 3.7% | 2.02 | 2.04 |
| Ship-as-is AI listing video | 1.76 | 0.8% | 1.27 | 1.42 |
Exhibit 1. Seller-shot leads intent and trust. Ship-as-is and badge-only lose on every primary measure. Preferences AI simulation, listing-video product test, n = 357; directional, not a survey of real shoppers.
Hold-until-simulation ranks third on intent and first on uniqueness (2.87 out of 5). Shoppers notice it as a distinct policy. It is not a substitute for real footage when the hero slot can be filmed. In the same test, 98.0% of preference votes went to seller-shot footage.
Before running the test, we set a condition for rejecting the hypothesis: if holding for a buyer simulation, or requiring seller approval, did not clearly beat ship-as-is and badge-only on purchase intent, we would not treat validate-before-ship as the method to recommend. Both cleared that bar. Seller-shot remains the hero to ship now.
So what: Film the next heroes you can film. Use seller approval or a pre-ship buyer check as the gate for AI creative, not as a reason to publish unchecked auto video.
3.99/ 5
PAI simulation · directionalIn a simulated product test of listing-video policies, an accurate seller-shot hero scored 3.99 out of 5 on purchase intent.
In a simulated product test, 78.3% of the panel gave an accurate seller-shot hero a top-two purchase-intent score (4 or 5 out of 5).
4.41/ 5
PAI simulation · directionalIn a simulated product test, an accurate seller-shot hero scored 4.41 out of 5 on usability, the highest of the five listing-video policies.
4.55/ 5
PAI simulation · directionalIn a simulated product test, an accurate seller-shot hero scored 4.55 out of 5 on credibility, the highest of the five listing-video policies.
3.38/ 5
PAI simulation · directionalIn a simulated product test, seller-approved AI scored 3.38 out of 5 on purchase intent, second to an accurate seller-shot hero.
2.99/ 5
PAI simulation · directionalIn a simulated product test, holding a listing video until a buyer simulation passes scored 2.99 out of 5 on purchase intent.
2.87/ 5
PAI simulation · directionalIn a simulated product test, holding a listing video until a buyer simulation passes scored 2.87 out of 5 on uniqueness, the highest of the five policies.
2.08/ 5
PAI simulation · directionalIn a simulated product test, an AI badge with no accuracy gate scored 2.08 out of 5 on purchase intent.
1.76/ 5
PAI simulation · directionalIn a simulated product test, ship-as-is AI listing video scored 1.76 out of 5 on purchase intent, the lowest of the five policies.
In a simulated product test, 98.0% of preference votes (350 of 357) went to an accurate seller-shot hero.
Note on simulated figures: Directional output from a Preferences AI Digital Population, not a survey of real shoppers. The panel is concentrated in Singapore, so read it as a regional directional screen, not a multi-market census. Methodology
Yes for the next listing: accuracy gates take every vote. Over six months, validate-before-ship becomes the default for AI creative.
On the follow-up simulated survey, after the same kind of marketplace context, two-thirds chose an accurate seller-shot hero as the next listing and one-third chose seller-approved AI. Ship-as-is, badge-only, and hold-until-simulation received no votes as the next listing.
The motives are physical fidelity and authenticity. Every simulated respondent said colour, size, and claims need to match the real product. Almost all said they trust a real phone video more than an AI host. None said speed of new videos matters more than an accuracy gate.
Asked which policy should be the default in six months, the largest share chose validate-before-ship: an audience simulation, or an equivalent buyer test, must pass before AI creative goes live.
| Six-month default | Share of simulated respondents |
|---|---|
| Validate-before-ship for AI creative | 58.7% |
| Seller-shot real hero preferred or required | 22.2% |
| Seller-approved AI required | 18.8% |
| Ship-as-is auto or affiliate AI stays normal | 0.3% |
| AI badge required, no accuracy gate | 0% |
Exhibit 2. The next listing locks onto seller-shot footage. The six-month default seats validate-before-ship as the standard for AI creative. Preferences AI simulation, follow-up survey, n = 356; directional.
So what: Treat seller-shot as the hero default, and validate-before-ship as the gate whenever AI creative is in the stack.
In a simulated survey, 66.3% of respondents chose an accurate seller-shot hero as the next listing policy.
In a simulated survey, 33.7% of respondents chose seller-approved AI as the next listing policy.
In a simulated survey, 0% of respondents chose ship-as-is AI, an AI badge alone, or hold-until-simulation as the next listing policy.
In a simulated survey, 100% of respondents said they need confidence that colour, size, and claims match the real product. The question allowed more than one motive.
In a simulated survey, 97.2% of respondents said they trust a real phone video more than an AI host. The question allowed more than one motive.
In a simulated survey, 0% of respondents said the speed of new videos matters more than accuracy gates. The question allowed more than one motive.
In a simulated survey, 58.7% of respondents chose validate-before-ship as the six-month default for listing-linked product video: a buyer test must pass before AI creative goes live.
In a simulated survey, 22.2% of respondents chose a seller-shot real hero as the six-month default for the hero slot.
In a simulated survey, 18.8% of respondents chose seller-approved AI as the six-month default before publish.
In a simulated survey, 0.3% of respondents chose ship-as-is auto or affiliate AI as the six-month default.
No. On this panel a badge without a gate protects no one, and a mismatch collapses trust in the seller.
Asked which better protects them, 78.4% of simulated respondents chose an accuracy gate even without a fancy badge, 21.6% wanted both a disclosure and a gate, and none chose an AI badge alone.
After an AI video mismatches size, colour, or tone, trust in that seller's other listings averages 1.88 out of 5. Every simulated respondent rated that trust 1 or 2. If a marketplace keeps allowing ship-as-is misleading AI after flags, 84.6% said they would switch to listings with real or accuracy-gated video, and another 13.2% said they would shop less on that marketplace (97.8% combined).
So what: Disclose when a label is required, and do not treat the label as a substitute for colour, size, and claims that match. A mismatch is a trust incident, not a polish issue.
In a simulated survey, 78.4% of respondents said an accuracy gate protects them better than a badge, even without a fancy AI label.
In a simulated survey, 21.6% of respondents said they want both an AI disclosure and an accuracy gate.
In a simulated survey, 0% of respondents said an AI badge or disclosure alone is enough to protect them.
1.88/ 5
PAI simulation · directionalIn a simulated survey, mean trust in a seller’s other listings after an AI video mismatched size, colour, or tone was 1.88 out of 5.
In a simulated survey, 100% of respondents rated trust in a seller’s other listings as 1 or 2 out of 5 after an AI video mismatched the product.
In a simulated survey, 84.6% of respondents said they would switch to listings with real or accuracy-gated video if a marketplace kept allowing misleading ship-as-is AI after flags.
In a simulated survey, 13.2% of respondents said they would shop less on a marketplace that kept allowing misleading ship-as-is AI after flags.
In a simulated survey, 97.8% of respondents said they would either switch listings or shop less if a marketplace kept allowing misleading ship-as-is AI after flags.
Yes. Every simulated respondent wanted some pre-ship buyer check, and most wanted it for all AI creative.
The kill bar set before the runs was this: if holding for a buyer simulation, or requiring seller approval, did not clearly beat ship-as-is and badge-only on purchase intent and on the forced next listing, do not claim validate-before-ship as the method to recommend. It cleared. Both gated paths beat ungated ship and badge-only on intent. The forced next listing gave the losers 0%.
On the survey, 74.7% said hold AI creative until a buyer panel or simulation passes, for all creative. Another 25.3% said yes for risky categories only, such as products where size or colour is easy to get wrong. None were neutral, opposed, or unsure. That is 100% in favour of some pre-ship validation.
So what: For an AI marketing agent, the recommendation is validate-before-ship for AI creative, with seller-shot footage as the preferred hero when filming is feasible. It is not "generate, badge, publish."
In a simulated survey, 74.7% of respondents said brands should hold AI creative until a buyer panel or simulation passes, for all creative.
In a simulated survey, 25.3% of respondents said brands should validate AI creative with a buyer panel only for risky categories, such as size- or colour-sensitive goods.
In a simulated survey, 100% of respondents wanted brands to validate AI creative with a buyer panel before it ships, either for all creative or for risky categories only.
3.38/ 5
PAI simulation · directionalIn a simulated product test, seller-approved AI scored 3.38 out of 5 on purchase intent, second to an accurate seller-shot hero.
In a simulated product test, 0.8% of the panel gave ship-as-is AI listing video a top-two purchase-intent score (4 or 5 out of 5).
In a simulated product test, 3.7% of the panel gave an AI badge with no accuracy gate a top-two purchase-intent score (4 or 5 out of 5).
Stop recommending ungated AI, default the hero to real footage, and hold AI creative until a buyer check passes.
In a simulated survey, 58.7% of respondents chose validate-before-ship as the six-month default for listing-linked product video: a buyer test must pass before AI creative goes live.
In a simulated survey, 66.3% of respondents chose an accurate seller-shot hero as the next listing policy.
In a simulated survey, 100% of respondents said they need confidence that colour, size, and claims match the real product. The question allowed more than one motive.
| Statistic | Value | Source |
|---|---|---|
In a simulated product test, 51.9% of the panel gave seller-approved AI a top-two purchase-intent score (4 or 5 out of 5). | 51.9% | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, seller-approved AI scored 3.49 out of 5 on usability. | 3.49/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, seller-approved AI scored 3.86 out of 5 on credibility. | 3.86/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, 35.4% of the panel gave hold-until-simulation a top-two purchase-intent score (4 or 5 out of 5). | 35.4% | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, holding a listing video until a buyer simulation passes scored 2.70 out of 5 on usability. | 2.70/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, holding a listing video until a buyer simulation passes scored 3.48 out of 5 on credibility. | 3.48/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, an AI badge with no accuracy gate scored 2.02 out of 5 on usability. | 2.02/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, an AI badge with no accuracy gate scored 2.04 out of 5 on credibility. | 2.04/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, ship-as-is AI listing video scored 1.27 out of 5 on usability, the lowest of the five policies. | 1.27/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated product test, ship-as-is AI listing video scored 1.42 out of 5 on credibility, the lowest of the five policies. | 1.42/ 5 | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
In a simulated survey, 38.8% of respondents selected a motive that an AI-generated label alone is not enough if the video can still mislead. The question allowed more than one motive. | 38.8% | Source: Preferences AI simulation: marketplace listing-video survey, 3 October 2026 |
In a simulated survey, 91.3% of respondents said they shop on Shopee or TikTok Shop at least monthly, matching the intended audience screen. | 91.3% | Source: Preferences AI simulation: marketplace listing-video survey, 3 October 2026 |
In a simulated survey, 94.4% of respondents said they usually or sometimes watch product videos or short listing clips before buying. | 94.4% | Source: Preferences AI simulation: marketplace listing-video survey, 3 October 2026 |
In the simulated product-test panel, 92.4% of respondents were placed in Singapore (330 of 357), so the screen is not a balanced Southeast Asia census. | 92.4% | Source: Preferences AI simulation: marketplace listing-video product test, 3 October 2026 |
Two directional Preferences AI studies, run on 3 October 2026, on one population frame.
The press figures in the opening section are from Channel NewsAsia (25 August 2026, updated 26 August 2026) and the Advertising Standards Authority of Singapore (23 March 2026). They are not Preferences AI results.
Panel mix on the product test: Singapore 92.4% (330 of 357), Indonesia 4.8% (17), with a handful in China, Malaysia, Thailand, and Vietnam. Age was spread from 18–24 through 45–54, led by 25–34 (32.2%). About 35% were coded middle income and 34% lower-middle. Shopping style was 45.7% research-heavy.
Directional simulation evidence from a synthetic panel, not a survey of real shoppers and not a census of Shopee or TikTok Shop buyers. The panel is concentrated in Singapore (92.4% on the product test). Do not read the figures as a balanced Southeast Asia result. The audience screen was otherwise close to the brief: 91.3% said they shop these apps at least monthly, and 94.4% usually or sometimes watch listing videos. Seller-shot, seller-approved AI, and validate-before-ship were not tested against each other with an explicit production cost or turnaround time. Scores are panel means on a 1–5 scale. A top-two score is the share rating 4 or 5. Several survey answers piled onto a single option, including 0% for ungated policies as the next listing and 100% wanting some pre-ship check. Treat those as directional signals of a sharp split, and review them for mode collapse before citing them as a market rate.
We refresh this page roughly monthly. The URL stays the same.
Preferences AI. Southeast Asia Marketplace AI Listing Video 2026: Why Seller-Shot Footage Beats Ungated AI. Last updated 3 October 2026. https://preferencesai.io/research/sea-marketplace-ai-listing-video-2026
Go beyond the averages
Run a quick concept test on an AI Digital Population for your APAC market, or compare options from your coding agent with preferences-validate.