Define Amazon REC’s Creative Inputs Before Enabling AI Optimization

01

Why REC changes the creative approval unit

Responsive optimization expands the approval problem from a finished ad to a changing collection of source material and possible combinations.

Amazon describes Responsive eCommerce Creative, or REC, as an Amazon DSP display format that can automatically create and resize ads for eligible placements. Its commerce treatments can combine product information with product images, custom images, logos, headlines, disclaimers, video, and dynamic variations. Depending on configuration, live details such as price, promotions, coupons, reviews, availability, and Add to Cart treatments may also appear. The current specifications allow as many as 20 associated ASINs in one creative. (Amazon REC guidelines and specifications)

Amazon also documents advanced dynamic creative optimization that can work with as many as 15 uploaded images and 15 headlines. When advanced DCO is enabled, Amazon says it can use content from the advertiser’s asset library and generate additional headline or image variations. (Amazon on personalized responsive ecommerce creatives) Newer creative tools can also source or enhance material from existing campaigns, Brand Stores, and product detail pages, as well as create video through generative AI. (Amazon REC creative-tools overview)

That means approval of one attractive preview is insufficient. The brand has to govern the ingredients, their transformations, and their permissible relationships. Amazon’s documentation explains available inputs, previews, optimization, and policy requirements, but it does not describe a complete brand-side manifest that approves every possible combination. The workflow below is therefore an operating recommendation, not a claim about a native Amazon governance feature.

02

1. Establish an ASIN allowlist

Start with the products Amazon is allowed to advertise, not with the available artwork.

Create one row for every approved marketplace and ASIN. Record the parent or child variation, product owner, destination, packaging version, launch status, inventory concern, and allowed REC treatments. Exclude discontinued products, embargoed launches, low-stock items, obsolete formulas, and seasonal variants unless the release owner has explicitly accepted the risk.

This product-level control matters because REC commerce treatments react to changing catalog and offer conditions. Amazon states that an out-of-stock product will not serve an Add to Cart variation and that inactive coupons are removed from optimization. The format may also display live commerce information such as price, deals, promotions, reviews, and availability-dependent treatments. (Amazon REC guidelines and specifications)

Do not assume that a product-family approval covers every child ASIN. A headline about an ingredient, size, color, compatibility, age range, or included accessory may be correct for one variation and incorrect for another. The allowlist should therefore identify the precise products compatible with each claim and asset.

1. Record marketplace, ASIN, variation, landing page, product owner, and packaging version.
2. Mark each ASIN approved, conditional, temporarily held, or prohibited.
3. List the images, headlines, videos, disclaimers, and commerce treatments permitted for that ASIN.
4. Define re-review conditions for inventory, offer, packaging, formula, variation, or destination changes.
5. Require the release owner to revoke or pause affected inputs when a condition is no longer true.

03

2. Audit PDP and Brand Store dependencies

Treat the PDP and Brand Store as part of the creative-input surface, not only as destinations.

Amazon says its REC tools can source or enhance assets from product detail pages, Brand Stores, and prior campaigns. It also says conversational creative can derive questions from product detail pages, Brand Store content, and campaign information. (Amazon REC creative-tools overview) A destination-page edit can therefore affect more than post-click conversion; it may alter material available to creative automation.

Capture a dated snapshot or field-level record of the product title, main and secondary images, bullets, description, claims, warnings, A+ Content, price, promotion, videos, and relevant Store modules. Record which fields have been reviewed for advertising use and which are merely acceptable on the destination.

This distinction is important because Amazon explains that product detail pages may incorporate product information contributed by manufacturers and sellers, and that submitted information can be combined. (Amazon guide to improving product detail pages) If a catalog value changes without the creative team noticing, an automated treatment could inherit an outdated or unapproved premise.

1. Save a snapshot date and responsible catalog owner.
2. Identify every PDP or Store field on which an approved claim depends.
3. Flag contributions that the brand does not directly maintain or cannot reliably verify.
4. Re-run review after material PDP, Store, offer, packaging, or product changes.

04

3. Separate asset sourcing from transformation permission

An asset can be approved for ordinary use without being approved for AI enhancement, animation, or product-in-use generation.

Give every image, logo, and video a stable asset ID. For each item, record its source, status, compatible ASINs, marketplaces, placements, rights expiration, reviewer, and evidence of approval. Then add separate permissions for crop, resize, animation, background generation, enhancement, headline generation, image generation, and video generation.

Use distinct controls for brand uploads, PDP assets, Brand Store assets, prior-campaign material, the Amazon asset library, and Amazon-generated outputs. A global “AI allowed” checkbox is too coarse. A studio packshot may be safe to resize but not to place in a generated usage scene. A licensed lifestyle photograph may be approved for display advertising but not for synthetic modification.

Amazon’s additional product terms state that generative tools and outputs are supplied “as is”; inputs must not contain personal data, and outputs cannot be used as deepfakes. The terms also place responsibility on advertisers to comply with law, intellectual-property rights, advertising policies, and applicable provenance controls. They state that submitted inputs may be used internally to improve products and models and that some inputs may be sent to supporting vendors. (Amazon Ads additional terms) Operationally, do not submit confidential production files, personal data, unreleased talent material, or assets lacking documented advertising and transformation rights.

1. Assign a stable ID to each image, video, and logo.
2. Record source as brand library, PDP, Brand Store, prior campaign, Amazon library, or generated output.
3. Set status to approved, conditional, prohibited, expired, or revoked.
4. Specify allowed transformations individually rather than granting blanket AI permission.
5. Attach rights evidence, usage limits, expiration, reviewer, and approval date.

05

4. Approve headlines as product claims

Review each headline for product applicability and evidence, not only tone and grammar.

Amazon’s REC specifications limit a custom headline to 50 characters. They also prohibit pricing or savings language, forceful urgency, unsupported superlatives, and Subscribe & Save messaging in that field. (Amazon REC guidelines and specifications) These platform rules are only the first layer of review; the brand still needs to decide whether a statement is true for the exact ASIN, market, formulation, and depicted use.

Attach every headline ID to eligible ASINs, approved markets, claim evidence, required disclaimer, incompatible imagery, expiration date, and reviewer. This makes it possible to revoke a claim without deleting unrelated creative work.

Amazon’s advertising policies require claims to be truthful and supported. Product claims should identify the applicable products clearly, with substantiation available on the product detail page or packaging where required. (Amazon Ads creative acceptance policies) Do not allow AI-generated headlines into the eligible pool merely because they fit the character limit. They need the same product, policy, legal, and substantiation review as human-written copy.

1. Create a unique ID for every headline.
2. Map it to exact ASINs, markets, evidence, dates, and reviewers.
3. Record prohibited images, products, or contexts.
4. Review generated headlines before adding them to the approved pool.
5. Expire the headline when its evidence, formula, packaging, or market authorization changes.

06

5. Make disclaimers dependent, not free-floating

A disclaimer should be mapped to the claim and product that require it.

Amazon’s custom disclaimer field is limited to 60 characters and is intended for products that require a disclaimer. (Amazon REC guidelines and specifications) Treat that field as a dependency rather than another interchangeable text asset.

For each disclaimer, record the associated ASIN, claim, market or jurisdiction, headline, image or depicted use, destination disclosure, and legal approval date. If a claim cannot be adequately qualified within the available treatment, remove the claim from that combination rather than relying on an unreadable or incomplete disclaimer.

Amazon’s quick-reference policy guidance requires necessary disclosures to appear in the ad or on the landing page and identifies illegible ad copy or disclaimers as unacceptable. (Amazon Ads policy quick reference) Preview disclaimer-bearing combinations in the smallest relevant format and in animated states, not only in a large desktop mockup.

07

6. Encode prohibited pairings

Two approved components can still create a misleading or noncompliant implication when combined.

Build negative rules alongside the allowlists. A health-oriented headline may become a treatment implication when placed over a particular usage scene. An ingredient benefit may be inaccurate for another flavor or formula. A lifestyle image may show accessories that are not included with the selected ASIN. A sustainability statement may no longer be supported after packaging changes.

Amazon states that custom images must not contradict the destination. When multiple identifiable products appear, the image should accurately represent products available on the landing page. (Amazon REC guidelines and specifications) Pairing rules help translate that requirement into executable review instructions.

Write rules in plain language and, where possible, reference asset IDs: “Headline H-07 must not pair with ASIN B,” or “Image I-12 requires disclaimer D-03 when used with headline H-04.” Include an owner who can resolve ambiguous cases. A warning without an accountable decision-maker tends to become documentation that everyone assumes someone else checked.

1. Block claims from incompatible formulas, sizes, age groups, or markets.
2. Block lifestyle scenes that imply unbundled accessories or unsupported uses.
3. Block expired reviews, seasonal treatments, rights-limited talent, and obsolete packaging.
4. Require specified disclaimers for regulated claims and depicted uses.
5. Prohibit combinations that make a required disclosure unreadable.

08

7. Review the combination space with a preview matrix

Preview enough representative combinations to test layout, meaning, commerce state, and transformation risk.

Amazon provides dynamic previews for uploaded images and headlines, while DCO can tailor combinations by audience and publisher. (Amazon on personalized responsive ecommerce creatives) A single default preview cannot demonstrate every possible treatment.

At minimum, inspect square, wide, tall, and small mobile compositions; onsite and offsite examples; product-image and custom-image treatments; multi-product combinations; every regulated or disclaimer-bearing ASIN; and each permitted AI transformation type. Where enabled, include video and connected-TV-oriented treatments in the matrix.

Also test commerce states such as coupon, deal, review, Add to Cart, and unavailable-product behavior. Amazon notes that some smaller REC sizes animate between the headline, logo, disclaimer, and product image, and that it may test animation, design elements, and formats. (Amazon REC guidelines and specifications) Review the sequence, not just the first frame: a disclosure that appears separately from the qualified claim may fail the brand’s standard even if both technically appear.

1. Define required formats and placements before review begins.
2. Select high-risk ASINs and pairings rather than reviewing only the easiest examples.
3. Capture every permitted AI generation or transformation class.
4. Record pass, fail, exception, reviewer, and date for each preview case.
5. Do not infer that an unreviewed combination is approved because a similar one passed.

09

8. Assign ownership and re-review triggers

Name one release owner who is accountable for the manifest’s current state.

The release owner should be authorized to approve newly sourced or generated material, resolve prohibited pairings, revoke assets, request a pause, maintain claim evidence, and coordinate PDP or Store corrections. Name a backup reviewer so a launch does not bypass governance when the primary owner is unavailable.

Set explicit triggers for re-review: a new ASIN; formula, packaging, or title changes; a PDP or Store edit; a new promotion; rights expiration; a new marketplace or placement; a new AI transformation type; a policy rejection; or generated material derived from an unapproved source. Version the manifest and retain the reason for each change.

Do not confuse optimization with approval. If a sourced or generated asset appears that is absent from the current manifest, treat it as unapproved until the named owner evaluates it. The goal is not to predict every rendering manually. It is to make the allowed input boundary, escalation path, and revocation decision explicit.

10

Recommended manifest fields

Keep the manifest specific enough that another reviewer can reconstruct what was permitted at release.

A practical schema is: manifest version; marketplace; campaign or line item; ASIN; ASIN status; landing page; PDP snapshot date; approved headline IDs; approved image IDs; approved video IDs; approved logo ID; required disclaimer ID; allowed ad variations; permission for asset-library sourcing; permission for prior-campaign sourcing; permission for AI headlines, images, animation, and video; allowed transformations; prohibited-pairing rules; placement scope; completed preview cases; claim-evidence location; rights expiration; review owner; backup reviewer; approval date; next review date; revocation status; and notes.

Keep IDs stable across the brief, asset library, review evidence, and iteration history. Avoid replacing a file under an existing ID after approval; create a new version so the reviewed artifact remains identifiable. The manifest can begin as a controlled spreadsheet or database, provided permissions, versioning, and ownership are clear.

11

How ATIYO fits the workflow—and where it stops

Use ATIYO to preserve the strategy, brief, approved assets, iteration history, decisions, and reusable lessons surrounding the REC launch.

ATIYO is a creative-strategy operating system for founder-led ecommerce brands, creators, and dropshippers. A team can organize the input manifest alongside its roadmap, brand context, creative brief, assets, review decisions, iterations, and reusable learnings. This gives the next launch a traceable starting point instead of forcing the team to reconstruct why an ASIN, claim, or generated treatment was allowed.

ATIYO does not connect to Amazon ad accounts, buy media, calculate ROAS, or automatically know which REC combination performed. Media performance remains in the ad platform and ATIYO preserves creative context and learnings. If performance observations matter to the next iteration, a user must record them in ATIYO with the relevant manifest version and asset IDs.

Use ATIYO as the creative operating record, not as a substitute for Amazon’s policy checks, previews, campaign controls, or reporting. If that separation matches your workflow, start a card-confirmed seven-day trial on the selected plan before billing begins.