CREATIVE STRATEGY

AI-Generated Ads: What Should Still Be Filmed With Real People and Products?

Updated: 9/18/26

Short answer: AI can generate environments, transitions, storyboards, variations, and supporting scenes. Real capture should remain the default when the ad’s persuasive power depends on product behavior, a customer’s experience, a recognizable person, a regulated claim, or evidence the buyer must trust.

The strategic question is proof, not production

Multimodal tools can create polished ads faster than traditional production. That does not make every scene equally suitable for generation. The creative team should ask whether a scene is decorative, explanatory, or evidentiary. Decorative scenes tolerate more synthesis. Evidentiary scenes carry the burden of proof.

An attractive generated kitchen can establish mood. A generated before-and-after result, testimonial, product demonstration, or professional endorsement can create a materially misleading impression if viewers reasonably treat it as real evidence.

Film the product when behavior matters

Capture real footage when buyers need to see texture, size, fit, speed, sound, installation, durability, application, or use in context. Product fidelity problems are not limited to incorrect logos. AI can subtly change dimensions, ingredients, controls, packaging, accessories, or results.

Use generated scenes around the real demonstration rather than replacing it. A practical workflow is to lock the verified product shot, then create backgrounds, transitions, crops, or non-evidentiary lifestyle variants.

Use real people for testimony and authority

A customer story should come from an actual customer whose statement reflects their experience. A founder, clinician, technician, or spokesperson should authorize how their identity and words are used. Synthetic actors can perform fictional scenarios, but the ad must not imply that they are real customers or experts when they are not.

The FTC’s endorsement guidance emphasizes honest opinions, material connections, and claims that advertisers can support. AI does not change those obligations.

Keep high-risk claims connected to evidence

Health, financial, safety, performance, environmental, and comparative claims deserve the highest review level. A generated scene can make an unsupported outcome feel demonstrated. Require source documentation, qualified review, accurate disclosures, and final approval before release.

Use AI where variation is valuable

Creative element	Recommended source
Product demonstration	Real capture
Customer testimonial	Real customer and authorized edit
Background environment	AI or practical production
Storyboard and shot exploration	AI-assisted
Resizing and format adaptation	AI-assisted with QA
Regulated or comparative proof	Verified real evidence
Fictional lifestyle scenario	AI-generated if clearly non-testimonial

Build a hybrid production package

Plan one real proof shoot to capture the product, people, demonstrations, disclosures, and clean source plates. Then use AI to extend formats, locations, pacing, and supporting concepts. This approach preserves evidence while gaining production leverage.

Maintain a provenance record containing source files, releases, claims support, generated elements, prompts, edits, and approvals. The MTC multimodal creative QA workflow provides the broader release process.

Test trust, not only click-through rate

Generated polish may increase attention while reducing confidence after the click. Evaluate qualified conversion rate, product-page behavior, comments, returns, cancellations, and customer-service feedback. A high CTR is not a creative win when the ad establishes expectations the product cannot meet.

Practical takeaway: Generate atmosphere and variation. Capture the evidence that makes the claim believable.

Write a proof standard into the creative brief

Label every planned scene as atmosphere, explanation, demonstration, or evidence. Then specify the acceptable source for that category, the required documentation, and the reviewer. A product close-up might require an approved source photograph, while a testimonial requires a real participant, release, transcript, and claim review. Generated backgrounds may need only brand and realism checks.

This classification lets teams use AI aggressively without debating authenticity asset by asset. It also helps media buyers understand which variations can be produced quickly and which require a new capture session.

Frequently asked questions

Can an AI actor appear in an ad?

Yes, but the context should not falsely present the actor as a real customer, employee, expert, or endorser.

Should every generated element be disclosed?

Disclosure requirements vary by platform, jurisdiction, and use. Review current rules, especially when realism or identity could mislead.

What is the best hybrid starting point?

Film a verified product demonstration and real spokesperson, then use AI for supporting scenes and format variations.