Build an Approval Matrix Before Meta Automatically Rewrites or Resizes Your Creative

01

1. Define what each approval state permits

The matrix works only if every state has an operational meaning. Write those meanings into the process so that a media buyer, designer, founder, and legal reviewer interpret them consistently.

Allow means the feature may be enabled without another creative review, but only inside documented limits. For example, image expansion might be allowed when the product is centered, the surrounding area is a plain color, no packaging or disclosure is near an edge, and the generated area cannot alter the product.

Manual review means a named owner must inspect the generated result in context before publication. Approval of the source image does not automatically approve every generated crop, background, animation, text combination, or soundtrack.

Prohibit means the feature must remain disabled for that asset. Use this state when an output could alter a regulated claim, misrepresent the product, violate a creator agreement, obscure a disclosure, or change an intentionally silent piece of creative.

Permissions should follow the exact asset version. A clean packshot and a testimonial image may appear in the same campaign but carry different risks. The packshot might permit bounded image expansion; the testimonial may prohibit text generation because the quotation, attribution, and qualification must remain exact.

1. Write the three state definitions at the top of the matrix.
2. Require a short condition or reason beside every Allow or Prohibit decision.
3. Reject incomplete rows before Ads Manager setup begins.

02

2. Start with a conservative enhancement matrix

Use this recommended matrix as a starting point, then tighten it for your category, claims profile, brand rules, licensing agreements, and product presentation requirements.

Text variations — default: Manual review. Allow only for low-risk, evergreen language drawn from an approved message bank. Review any variation involving price, promotions, comparisons, testimonials, quantified outcomes, product-performance language, deadlines, or qualifiers. Prohibit outputs that introduce unapproved health, safety, financial, environmental, or legal claims.

Image expansion — default: Manual review. Meta describes image expansion as generating additional pixels to fit more aspect ratios. Consider allowing it only when expansion affects simple background space and cannot modify product edges, labels, logos, people, demonstrations, or disclosures. Prohibit it where generated pixels could change apparent product size, texture, included components, packaging, or results. (Meta Advantage+ creative)

Generated backgrounds — default: Manual review or Prohibit. Allow only when the product cutout is locked and the setting is immaterial to product function or meaning. Review lighting, shadows, scale references, surfaces, and implied locations. Prohibit generation for before-and-after evidence, regulated presentations, sensitive creator imagery, or photography governed by strict licensing terms.

Full-image generation — default: Prohibit unless a deliberate review workflow exists. Broader generated variations can depart further from the approved source than simple expansion. Require comparison against the product record, brand rules, and original brief before accepting any result.

Static-image animation — default: Manual review. Meta says its tools can animate a static image without requiring an existing video asset. Decorative background movement may be acceptable, but motion involving the product, packaging, spokesperson, demonstration, disclosure, or text needs review. Prohibit motion that could imply an unsupported function, usage method, endorsement, or result. (Meta Advantage+ creative)

Music — default: Manual review. Meta documents automatic music selection based on ad content and also provides advertiser selection options. Review with sound on whenever tone, voice, creator rights, premium positioning, or message interpretation matters. Prohibit music when silence is intentional or contractual restrictions apply. (Meta Advantage+ creative)

03

3. Build the matrix around the source asset, not the campaign

Each row should make it possible to reconstruct what the brand supplied, what Meta was permitted to change, what reviewers saw, and what ultimately launched.

Record these identity fields: asset ID; exact source filename and version; product or SKU; source-file link; product-media reference; campaign and ad identifiers; approved copy version; claim or substantiation references; and intended placements.

Add a separate state column for text variations, image expansion, generated backgrounds, full-image generation, animation, and music. Do not combine them into one Advantage+ creative field. A single asset may allow expansion, require music review, and prohibit generated text.

Add governance fields for creative owner, claims or legal owner, paid-social owner, approval date, approval expiration, preview evidence, approved automated changes, rejected outputs, exceptions, and reapproval status.

For Shopify stores, the product-media record can serve as one source reference because Shopify associates media with products and makes those media records available through the admin. It should not replace the approved master file, but it gives reviewers another place to verify the represented product. (Shopify product media)

A practical row might read: DTC-SERUM-014 v3; image expansion approved for Stories and Reels after preview review; text and generated backgrounds prohibited; animation allowed only in the background while the bottle and label remain static; music restricted to reviewed selections; signed by Creative, Claims, and Paid Social on YYYY-MM-DD.

1. Create a stable internal asset ID before uploading anything to Ads Manager.
2. Link the approved master rather than relying on a filename copied into a spreadsheet.
3. Record permissions separately for every enhancement.
4. Add the placement scope to every conditional approval.
5. Require reviewer names and timestamps instead of an anonymous approved status.

04

4. Put generated text behind the strictest claim gate

Automated text should recombine approved ideas—not invent new promises. Build a controlled message bank before enabling text variations.

Meta says its text-generation tools use submitted primary text, headline, campaign information, and other text inputs to produce alternatives. Depending on the selected enhancements, text may appear in different fields or combinations, including primary text, headlines, descriptions, or overlays. That means review must cover the complete rendered ad rather than one copy field in isolation. (Meta Advantage+ creative)

The message bank should contain approved product names, descriptors, benefits, supported performance statements, current price and promotion wording, required qualifications, approved calls to action, prohibited superlatives, overlay restrictions, and expiration dates. Link each objective claim to its supporting record or legal approval.

The FTC states that advertising must be truthful and non-deceptive and that advertisers need evidence supporting objective claims. Its health-products guidance also stresses substantiation and the complete impression created by words, images, qualifications, and disclosures. (FTC advertising basics; FTC health-products guidance)

Adopt a rejection rule: if a generated output introduces a number, guarantee, comparison, scientific phrase, customer result, health or safety benefit, environmental benefit, urgency statement, or promotion not present in the approved bank, it does not launch. Do not rewrite it informally in Ads Manager; send the revised copy back through the controlled copy version and approval process.

1. Attach the approved copy version and message bank to the asset row.
2. Identify copy elements that must remain verbatim.
3. Review primary text, headline, description, overlay, and destination-page consistency.
4. Reject unsupported express claims and new implied claims.
5. Save the accepted wording as a new controlled version if it becomes reusable.

05

5. Protect product truth before allowing visual generation

Generated visuals can change an ad’s implied meaning even when the main product cutout appears untouched. Define protected regions and prohibited contextual changes before reviewing previews.

Mark the product silhouette, packaging, label copy, logos, trademarks, color-critical surfaces, included accessories, anatomy or product-contact areas, before-and-after regions, demonstration steps, safety equipment, disclosures, and creator likenesses as protected where applicable.

Context is part of the claim. A generated clinical setting can imply professional or medical credibility. Extreme weather can imply durability. Nearby objects can change perceived scale. Shadows can change apparent shape, and animation can imply that a static component moves. The FTC’s small-business advertising guidance explains that both words and pictures contribute to the express and implied claims communicated by an ad. (FTC Advertising FAQs)

Compare generated output with authoritative, high-quality product imagery. Shopify recommends high-quality product media and notes that presentation can vary across themes and channels, reinforcing the need to review the actual destination and rendered placement rather than assuming one master image will always display identically. (Shopify product-media types)

A useful distinction is background-only freedom versus product-pixel freedom. If a feature may generate only blank surrounding space, approval can be relatively narrow. If it can alter the product, model, demonstration, packaging, or evidence area, apply a higher level of review or prohibit it.

1. Create a marked reference image showing protected regions.
2. List contextual implications that are unacceptable for the product.
3. Compare product shape, color, label, scale, accessories, and usage method.
4. Reject outputs that require a reviewer to guess whether the product remains accurate.

06

6. Require placement-by-placement preview sign-off

A toggle audit confirms configuration; a preview audit confirms output. Require both before publication.

Meta’s Advantage+ creative materials describe choosing individual features and reviewing variations in the Preview area. Meta also notes that certain enhancements may be active when its system predicts a benefit, so the media buyer should inspect all relevant enhancement controls rather than assuming the initial setup reflects the matrix. (Meta Advantage+ creative)

Review Feed, Stories, Reels, and every other selected placement available in the setup. A disclosure that survives a feed crop may become unreadable in a vertical treatment. Expanded pixels can appear harmless in one ratio but introduce a product or scale problem in another.

Listen to music with sound on. Read every text field and overlay. Inspect the destination page alongside the ad when an offer or qualification depends on landing-page details. Meta explains that its own ad-review process can examine creative elements and destinations, but that platform review remains separate from internal brand approval. Creative edits may also prompt another platform review. (Meta ad review policy)

1. Confirm the source asset ID and version in the matrix.
2. Open every relevant Advantage+ creative enhancement section.
3. Match each active control to Allow, Manual review, or Prohibit.
4. Preview every available variation and selected placement.
5. Check copy fields, overlays, crops, product details, labels, colors, disclosures, and calls to action.
6. Review music and animation at normal playback speed.
7. Capture dated screenshots or a screen recording.
8. Record accepted and rejected outputs.
9. Obtain all required sign-offs before publishing.

07

7. Keep an immutable record of approved automation

The final record should explain the boundaries of approval without requiring someone to reconstruct decisions from chat messages or Ads Manager history.

At minimum, the record should answer: Which source asset did Meta receive? Which enhancements were active? Which previews were inspected? Which placements were covered? Who approved each feature? Which claims, qualifications, and disclosures were checked? What conditions applied? Which version launched? Was anything edited after sign-off?

Store both accepted and rejected decisions. A rejected background can become a useful reusable learning: for example, do not place this product in clinical environments. A rejected animation may establish that the product cap cannot appear to rotate. These are operating rules, not merely comments on one ad.

Set explicit invalidation triggers. Reopen approval after a change to the source file, copy, offer, price, landing page, selected enhancement, product shown, music, rights status, or placement mix. Reapproval should also be required after the approval expires or when a claim-support record changes.

Use the internal Creative Brief Template to define the source concept and required message. Reference Keeping Every Brief On Brand With Your Brand Bible for verbal and visual constraints. Follow Ecommerce Asset Version Control so the sign-off remains attached to the exact source version.

1. Save the matrix row with its preview evidence.
2. Record the exact approved changes in plain language.
3. Lock the approval record after publication.
4. Create a new version rather than overwriting an approved row.
5. Assign a post-launch audit owner to confirm the live setup still matches approval.

08

8. Separate approval evidence from performance evidence

Approval tells you what automation was permitted to do. Performance tells you how the delivered ad performed. Do not treat those as the same record.

An approval log can establish that background generation was enabled and reviewed. It cannot establish that a particular generated background caused a performance result. Attribution requires reporting or a controlled test that isolates the variable.

Meta’s help material notes limits on breakdowns for individual creative variations and recommends testing when advertisers want to compare Advantage+ creative with standard ads. Avoid declaring a generated text, background, animation, or music track the winner unless the available platform reporting or test design actually identifies that variation. (Meta/Instagram music and creative reporting help)

For ATIYO, the boundary is straightforward: media performance remains in the ad platform and ATIYO preserves creative context and learnings. Teams can organize the roadmap, source brief, Brand Bible context, asset versions, approval decisions, owners, iterations, and reusable lessons. ATIYO does not connect to ad accounts, buy media, calculate ROAS, or independently know campaign performance unless a user records it.

A concise next step is to place the matrix beside the brief and asset history in ATIYO; a selected public plan includes a card-confirmed seven-day trial before billing begins.

09

9. Copy this prelaunch approval checklist

Use the checklist as the final release gate after the matrix has been completed and before the ad is published.

Source control: Asset ID and exact version recorded; product or SKU confirmed; source file and product-media references attached; approved copy and claim references attached; Brand Bible constraints documented.

Enhancement controls: Text variations, image expansion, generated backgrounds, full-image generation, animation, and music each marked Allow, Manual review, or Prohibit; conditions and prohibited combinations documented.

Preview review: Active enhancement areas inspected; available variations reviewed; selected placements checked; copy fields and overlays checked; product and packaging accuracy verified; claims, qualifiers, and disclosures checked; music reviewed with sound on; screenshots or a recording retained.

Sign-off and publication: Creative owner approved; claims or legal owner approved where required; paid-social owner matched settings to the matrix; approval date and expiration recorded; accepted automated changes entered in the log; post-launch audit owner assigned; reapproval triggers acknowledged.

The governing principle is not that all automation must be disabled. It is that automation should be allowed only when its possible output is bounded, reviewable, attributable to an exact source, and assigned to an accountable owner.