What should you compare before approving AI-enhanced Snapchat product creative?

01

What should you prepare before starting the review?

Use a fixed reference set so the approver is not judging the enhancement from memory.

Open three items together: the locked source asset, the current product page or specification sheet, and the enhanced output. Confirm that the reference material covers the advertised SKU, variant, package generation, included components, and current offer.

Snap’s announcement describes Smart Upscale, Image-to-Video, and Background Image Enhancement as advertiser-selected tools, with advertisers choosing generated assets to run in Ads Manager. That control makes source-versus-output review an advertiser responsibility rather than a step to assume the tool has completed.

1. Confirm the source file is the approved, unedited reference.
2. Confirm the SKU, variant, size, finish, flavor, and packaging version.
3. Identify every enhancement applied: upscale, generated background, image-to-video, or multiple stages.
4. Assign one approver who can verify product details and another reviewer when claims, rights, or safety are uncertain.

02

Does the enhanced product retain the correct identity and geometry?

The output should represent the product a buyer will receive, not a cleaner-looking approximation of it.

Compare silhouette, proportions, openings, seams, buttons, ports, handles, closures, edges, and component count. Upscaling can create plausible-looking detail that was not present in the source, while generative processing can straighten, enlarge, remove, or duplicate features.

Inspect transparent packaging, reflective surfaces, fine hardware, and partially obscured areas at high magnification. Reject the output if a reasonable shopper could expect a materially different physical product.

1. Verify the exact SKU, model, package, and variant.
2. Check dimensions and relative proportions against the source.
3. Look for invented, missing, merged, or duplicated components.
4. Confirm that reflections and transparent areas do not depict false contents.
5. Reject any output whose product identity cannot be reliably verified.

03

Are color, material, labels, and included accessories accurate?

Visual polish must not change what the product appears to be made from or what comes with the offer.

Compare product and packaging color with the approved reference rather than with another generated image. Check whether matte surfaces became glossy, plated finishes appear solid, or one textile now resembles a more expensive material.

Read logos, ingredient panels, warnings, certifications, measurements, and package text. Generated text that merely resembles a label is not acceptable. Props such as stands, cables, lids, refills, and applicators should not appear included unless they ship with the advertised offer.

1. Confirm product, packaging, and variant colors.
2. Check material, finish, texture, transparency, and reflectivity.
3. Read all visible label and package text at full resolution.
4. Verify logos, certifications, warnings, and measurements.
5. Remove or clarify accessories and props that are not included.

04

Does the generated background change scale or imply performance?

Evaluate a new background as part of the claim, not as neutral decoration.

Check perspective, contact points, reflections, shadows, and gravity. A small product placed beside undersized props can appear larger; an inaccurate shadow can make it look suspended or differently shaped.

The setting must also agree with real usage and safety instructions. Water, heat, food, outdoor conditions, impact, or luxury surfaces can imply waterproofing, heat resistance, food safety, ruggedness, outdoor suitability, or premium materials even when the copy makes no such statement. The FTC’s net-impression guidance calls for evaluating the complete ad, including its visual depictions.

1. Compare apparent product scale with recognizable scene elements.
2. Check that contact points, perspective, reflections, and shadows are plausible.
3. Confirm that the environment represents an appropriate use case.
4. Check whether props falsely imply additional items are included.
5. Reject backgrounds that create an unsupported performance or material claim.

05

Does image-to-video motion show only real product behavior?

Review every frame because motion can turn a decorative enhancement into an apparent product demonstration.

Watch the video at normal speed, then scrub it frame by frame. Product shape, color, label, texture, and component count should remain stable throughout. Camera movement must not reveal an invented underside, interior, connector, opening, or accessory.

Pay particular attention to lids opening, liquids pouring, devices operating, garments stretching, and parts moving. The FTC has warned that a depiction presented as a product demonstration must truthfully represent what the product can do; a brief fictionalization disclaimer may not correct the resulting impression.

1. Watch the complete video instead of approving from its thumbnail.
2. Pause on transitions, rotations, reveals, and final frames.
3. Confirm the product remains structurally and visually stable.
4. Verify that every demonstrated action is physically reproducible.
5. Check that splashes, glow, particles, speed, and transformations cannot be mistaken for product performance.

06

Are people and usage contexts authorized and truthful?

A generated person must not create a false endorsement, testimonial, result, or usage claim.

Confirm rights for every depicted likeness, voice, or recognizable lookalike. Snap’s general requirements call for advertisers to hold the necessary rights and permissions for their ad elements and prohibit unauthorized or false celebrity usage.

Check fit, body position, skin contact, age, environment, and product handling. Do not present a generated person as a real customer, expert, celebrity user, or independent reviewer. The FTC’s Endorsement Guides apply truth-in-advertising principles to fabricated endorsers, including virtual influencers.

1. Verify permission for each person, likeness, voice, and lookalike.
2. Confirm the product is worn, held, or operated correctly.
3. Remove unsupported before-and-after results or suitability implications.
4. Do not attribute invented experience or opinions to a generated person.
5. Escalate uncertain endorsement, testimonial, or publicity-rights issues.

07

Do the text, price, offer, and destination agree?

The creative and destination should describe the same product and offer without contradictory fine print.

Read every generated headline, caption, badge, call to action, and piece of packaging copy. Verify price, discount, quantity, subscription terms, shipping statements, availability, and inventory language against the live destination immediately before approval.

Snap reviews ad creative together with its associated destination. FTC guidance similarly explains that fine print cannot contradict a misleading headline or visual representation. Put necessary qualifications close to the relevant claim rather than relying on a distant landing-page explanation.

1. Verify headlines, captions, badges, CTA text, and packaging copy.
2. Match price, discount, quantity, terms, and shipping statements to the live offer.
3. Confirm the advertised SKU and variant are available on the destination page.
4. Make qualifications clear, prominent, and close to the claim.
5. Reject creative whose visual promise conflicts with its copy or landing page.

08

How should the source and enhanced derivative be stored?

Freeze the source, preserve it unchanged, and save each enhancement as a separately identified derivative.

Give the derivative its own asset ID and record the source ID, SKU, tool, generation date, operator, enhancement type, prompt or settings when available, approver, decision, and rejection reason. Preserve useful rejected versions because their failure modes can inform later briefs.

C2PA guidance models edited files as derivatives of source assets and supports recording ingredients, actions, prompts, and other provenance information at significant lifecycle events. Preserve available Content Credentials or comparable metadata rather than stripping or replacing that history.

Use the Ecommerce Asset Version Control guide to structure source freezing and the AI Creative Provenance Workflow for derivative records. Whether consumers need an AI-related label is a separate, context-specific assessment; use the AI Edited Ad Label Decision Checklist rather than assuming every enhancement requires the same disclosure.

1. Mark the verified source read-only or otherwise prevent replacement.
2. Export the enhancement under a new derivative ID and filename.
3. Record its parent source, enhancement method, operator, date, and approver.
4. Save approve, revise, reject, or escalate as an explicit status.
5. Retain available provenance metadata and document why rejected derivatives failed.

09

When should the asset be approved, revised, rejected, or escalated?

Use a decision based on material accuracy and claim risk, not merely production quality.

Approve when there is no material deviation and no new unsupported claim. Revise when a cosmetic artifact or contextual ambiguity can be corrected without changing the concept. Reject wrong products, invented functionality, false demonstrations, unauthorized people, misleading inclusions, inaccurate text or price, and outputs that cannot be compared confidently with the source.

Escalate regulated products, health or safety implications, uncertain rights, testimonials, and materially altered depictions. In ATIYO, teams can keep the source, derivative, brief context, iterations, decisions, and reusable learnings connected. Media performance remains in the ad platform; ATIYO preserves creative context and learnings rather than acting as the source of delivery, conversion, or ROAS reporting.

1. Approve: accurate product, context, motion, text, rights, and offer.
2. Revise: correctable artifact or ambiguity with no underlying concept problem.
3. Reject: material mismatch, unsupported claim, false demonstration, or unverifiable output.
4. Escalate: regulated claim, safety issue, unclear rights, testimonial, or substantial synthetic alteration.