AI product images: a brand consistency checklist before campaign delivery
An AI-generated product image can look convincing while still showing the wrong product. Brand review needs to look beyond the overall mood and inspect the details that make the item recognisable. Define an approval reference before generating a whole campaign, then use the same checks across every selected asset.
AI-generated imagery from Martinelli’s Valentine’s campaign. Production: PHOCUS Creators. Director: Nora Solaz. View the project and credits.
Create a source of truth for the product
Gather approved views of the actual item, packaging, logos and colour variants. Include the details that must not change: a seam, closure, sole shape, label position or surface finish. Mark which references are current and whether any depict a prototype.
Ask the brand team to distinguish essential accuracy from styling freedom. A campaign may allow an imagined setting while requiring the product itself to remain exact. Writing that distinction down gives the creative team an assessment standard beyond “it feels close”.
Check shape and construction first
Compare the generated asset with the reference at useful magnification. Inspect proportions, edges, fasteners, stitching and joins. Look for details that change halfway around the object or disappear when the viewpoint changes. Approve the object before spending time refining the surrounding world.
For a series, compare images together. A detail that looks plausible in isolation may be inconsistent across the set. Use one review sheet showing the approved reference alongside each proposed angle, with a named reviewer and an explicit decision.
Separate branding, material and colour checks
Review logos and text for spelling, shape and placement. Check the material cues: leather, fabric, glass and metal should support the intended product representation. Compare colour against the approved workflow and intended output; a pleasing screen preview is not a substitute for the brand’s colour approval.
Inspect reflections and shadows around the object. Ask whether the image implies features or behaviour that the real product does not have. Route factual product questions to the person authorised to approve them rather than resolving them solely as aesthetic choices.
Test the final use, not just the working image
Review the actual crop and delivery size. A small social preview can hide a defect that becomes obvious on a larger product page. Leave space for intended copy and verify that a vertical crop does not remove the detail the asset is meant to communicate.
For moving image, inspect transitions and frame-to-frame continuity as well as selected stills. A product that changes shape during motion needs further work even if the first and last frame look acceptable. Define the required movement before choosing a production approach.
Track decisions and provenance separately
Keep an asset identifier, reference version, review notes and approval status. Distinguish “approved creatively” from “approved for product accuracy” and “cleared for release”. This is a recommended workflow for avoiding ambiguous handovers, not a claim that one standard covers all approvals.
Where available, preserve Content Credentials and relevant source records. C2PA’s explainer describes provenance information and integrity checks; those records do not replace visual product review or establish that a depicted claim is true.
Decide when to change the technique
If a critical detail cannot be controlled consistently, reconsider the approach. A photographed product, a different composition or a hybrid treatment may better fit the brief. Read our comparison of live action, AI and hybrid production before scaling a promising test into a full asset package.
The imagery shown here comes from the Martinelli project produced by PHOCUS; this checklist is general guidance, not a report of that campaign’s review process. Send PHOCUS your product references and intended uses to discuss a suitable production approach.