A production proposal can promise lifelike AI footage without answering the brand’s actual question. What is the generated footage meant to do for the audience: develop a visual idea, create an imaginary world, explain a physical product, or stand in for a person?
For overseas brands coordinating samples and filming with a Chinese supply chain, that distinction changes the brief and the budget. ONCE recommends approving the role of AI before commissioning a batch of finished assets.
An October 22, 2026 Snapchat introduction to research with Ipsos and WPP offers a useful prompt. It describes 4,261 monthly AI users aged 18–44 in five countries and a separate group of 300 creators and marketers. Among consumer respondents, 48% reported daily AI use, while 56% described entirely AI-generated content as “soulless.” Familiarity and reservations can coexist. Neither percentage predicts your campaign’s sales.
Read the research at the level it actually measures
That consumer sample covers the US, UK, France, Saudi Arabia and India. It excludes non-users, older consumers and many markets. The industry respondents are a separate group. Treating these as universal consumer acceptance figures would overstate the evidence.
Gallup’s August 19, 2026 US analysis, Americans Aren’t Sold on Businesses Using AI in Advertising, illustrates why the task matters. With disclosure specified, 75% accepted AI assistance for brainstorming or early ad drafts, while 62% rejected creating people or voices for advertising. These are stated attitudes toward uses, not observed responses to your finished film.
Ipsos’ May 18, 2026 article, AI Ads Are Good Enough — And That’s the Problem, reports an assessment of 20 ads with 3,000 US respondents. Its AI counterparts used reverse-engineered briefs and generation without human creative intervention. The outcomes were advertising-assessment measures. That design cannot establish how every human-directed hybrid production performs, and predicted effectiveness is not realized client revenue.
These sources answer different questions. They support more precise commissioning questions; they do not supply a universal verdict on AI advertising.
ONCE’s creative-role brief
This is a proposed production method, not a validated scoring system. For each generated sequence, record the audience’s decision, AI’s assigned task, the material that must remain dependable, and the condition that would cause the team to reject the pilot.
| Assigned role | What the commissioning team should preserve and inspect |
|---|---|
| Backstage exploration: concepts, boards and mood studies | A clear brand proposition; assess the quality of the proposed idea rather than the number of generations |
| Deliberate imagination: abstract settings or visual metaphors | Cues that make the imaginative treatment understandable; check what viewers remember about the brand |
| Product presentation: compositing or scene expansion | Reliable physical-product references; inspect geometry, material, accessories and actions that influence expectations |
| Human expression: characters, voices and performance | An explicit role and intended meaning; do not fabricate an actual person’s experience or independent judgment |
The point is to give different creative tasks different acceptance criteria. An imaginary landscape can be valuable precisely because it is impossible. Footage presented as an actual product result needs a different basis for approval.
Specify what human direction does, too. Who assesses the idea? Who compares the image with the sellable product? Who reviews the language and performance for the target market? A line reading “human reviewed” does not describe those deliverables.
Separate the imaginative job from the product-information job
Consider a hypothetical footwear launch, not an ONCE customer example. The overseas marketing team wants an adventurous visual identity, but shoppers still need to understand the upper material and how the shoe flexes.
One proposal sends the shoe through an unfolding abstract blue landscape. Another shows the real material and a physical flexing action. The first can create an expressive world; the second addresses a concrete product question. They may work together in one film. A striking landscape does not resolve uncertainty about the material.
If accurate product footage already exists, the next commission might cover only a short exploratory world-building sequence. If the available references cannot establish the product details, obtain the relevant sample footage or reliable product material first. If the missing ingredient is a compelling story, more environments will not automatically supply it.
Ask suppliers to distinguish creative development, reference preparation or filming, generation, product checking and local-market review. Compare that combined scope. This article assumes no market price or guaranteed AI saving.
Buy a pilot that can change a decision
Before expanding production, state one question the pilot should answer. For example: with the same real product footage and proposition, does the abstract treatment improve brand recall without introducing a wrong inference about the shoe?
Prepare two reasonably comparable edits: a restrained setting and the proposed imaginative sequence. Keep the product shots, wording, offer, duration and viewing conditions as consistent as practical. Handle and record the disclosures each version needs. Concealing AI involvement is not a sound way to create a comparison.
An initial review with people in the target market can identify misunderstandings: which brand do they remember, what do they think the product does, and what remains unclear? A few interviews can inform revisions, but cannot establish a statistically reliable winner. A subsequent performance experiment needs an agreed sample, allocation, measures and stopping conditions; ordinary platform delivery is not automatically random assignment.
The result concerns these particular executions. It does not prove that consumers like AI in general or that a model outperforms cinematography. Connect the finding to a purchasing action: extend the treatment, revise it, or redirect work toward unresolved product information.
Retain the decision record, not just the exported video
For repeated launches, keep the assigned role, reviewed version, audience context, misunderstandings found and resulting production decision. That record can help a later project reuse an effective approach or avoid an irrelevant one. It is more informative than counting generated clips, although its commercial value still needs to be demonstrated.
To scope a project with ONCE’s production team, start with the intended market, one purchasing concern and the product assets already available. Define the role and pilot first, then decide the appropriate production volume.
Sources checked October 23, 2026: Snapchat for Business, Navigating the Age of AI: How AI is Rewiring Human Behavior, October 22, 2026; Gallup, Americans Aren’t Sold on Businesses Using AI in Advertising, August 19, 2026; Ipsos, AI Ads Are Good Enough — And That’s the Problem, May 18, 2026. The Snapchat official blog text was retrieved through the search index; direct page retrieval failed. We did not review the full underlying report or raw data. The cover is an AI-generated editorial concept.
