AI Campaign Review Kit

1. Check the product before you judge the styling

Compare each candidate with the real product reference. Check shape, proportions, color, material, label, hardware, and included accessories. Then inspect the scene: a hand can imply the wrong scale, a reflection can show a different product, and a prop can suggest an unsupported benefit.

Look closely, then at phone size, then in the intended crop. For a handbag, trace the handle from one attachment point to the other. Check the grip separately. A beautiful image with the wrong handle is not ready to use.

Prompt 1 · Find the mismatches

Review these campaign images against the attached original product references and approved brief. For each image, list: visible product mismatches, uncertain details requiring human verification, hierarchy, brand fit, and suitability for [channel and crop]. Quote the relevant brief rule for each concern. Do not invent product facts or claim an image will convert. Recommend select, revise, or reject, with one specific next action. If image detail is not visible enough to check, say so.

Returns a proposed review list. Check each observation yourself against the real product; AI can miss details or flag things that are correct.

2. Turn feedback into a decision

Separate accuracy, clarity, and personal preference. “I dislike green” is a preference. “The green reflection makes stainless steel look painted” is an accuracy concern.

Select when the product is accurate and the image fulfills its brief. Revise when the idea works and the correction is specific. Reject when the concept misses the message or the product cannot be represented accurately in that version.

Prompt 2 · Build your selection worksheet

Organize this feedback for [campaign] into accuracy, clarity, and preference: [paste feedback]. Use this brief: [paste brief]. Create one row per candidate with: candidate name, intended placement, accuracy concern and reference, message clarity, product visibility, select/revise/reject recommendation, one next action, and what must stay unchanged. Leave final approval blank for me. Separate observable concerns from opinions and unknowns.

Returns a proposed selection worksheet for your approval.

Write corrections as visible changes: “Restore the short handle in the reference. Preserve the approved lighting, background, and product position.” Inspect the entire revised image again before changing its status.

3. Review the campaign as a set

Place the homepage, email, social, and product-page candidates beside one another. Keep product truth and the chosen visual direction consistent, while giving each image a distinct job.

A homepage image might establish the setting; a detail image might explain the clasp. They can use different crops without looking like unrelated campaigns. Check light, palette, materials, styling, readable text, and the space each placement needs.

Prompt 3 · Check the whole campaign

Review this set for [campaign]: [identify each image and its intended placement]. The shared direction is [palette, lighting, materials, and styling]. The customer takeaway is [message]. For each image, explain its job, whether it fits the shared direction, and whether its intended crop preserves the product and readable copy. Identify contradictions and unnecessary repetition. Suggest one specific revision where needed. Do not force identical compositions or treat an unverified product detail as accurate.

Returns proposed consistency and placement corrections. Preview the actual layouts yourself before approving them.

4. Compare one creative choice at a time

Name the question before producing more versions. For example: does a tighter crop make the closure easier to understand? Keep the product, light, background, and copy fixed while comparing the crop.

An editorial comparison helps you choose credible candidates. It does not establish which image sells more. For a customer-facing test, decide the measure, observation window, and decision rule with the person running the channel before results arrive.

Prompt 4 · Plan a useful comparison

Create a one-variable comparison sheet for [campaign]. My question is [specific question]. Version A is [description]. Version B is [description]. The only intended difference is [variable]. List what must stay fixed, including copy, offer, audience, placement, and comparable timing where relevant. Add fields for the primary measure, observation window, decision rule, results, limitations, and next action. Leave undecided fields blank. Flag other differences that would prevent me from isolating the creative choice. Do not declare a sales winner from visual judgment.

Returns a comparison plan. Small, uneven, or conflicting results may leave the question unresolved.

5. Count the images you can actually use

Track all generation, review, correction, and final preparation time, including work on rejected images. Add direct production charges and a consistent allocation of subscription costs without counting the same charge twice.

Cost per usable image = total production cost ÷ images approved for their intended use. Value your time at the same internal hourly rate when comparing methods. If nothing is approved, there is no cost-per-usable-image result yet.

Prompt 5 · Calculate the usable cost

Build a cost worksheet for [campaign or batch]. Use these actual figures: direct production charges [amount], allocated subscription cost not already counted [amount], total working hours including rejected versions [hours], internal hourly rate [rate], and images approved for their intended use [count]. Show labor cost, total cost, and cost per usable image with the arithmetic. Identify missing inputs rather than guessing. If approved images equal zero, report that the ratio cannot yet be calculated. For a comparison with [second method], check that both methods include the same cost categories, scope, and approval standard.

Returns a cost worksheet based on your figures, not a pricing estimate.

Set a stopping rule before another revision. For example, after two focused attempts, reassess whether an accurate photographed product with an AI-created setting would better serve the image.

The thing that breaks the review: approving a beautiful image while a product-accuracy question is still unresolved.

Prompts are reusable review templates, not claims of tested results. Product examples are illustrative; no paid client campaign or performance outcome is represented.