Free editable template

AI ad creative consistency kit: brand bible and QA scorecard

Write your brand facts once and reuse them in every AI image or video prompt. The workbook assembles a prompt block, an avoid block and a video add-on from your colours, product facts, character and style, then scores each finished ad on ten checks, four of them pass-or-fail, before it ships.

By GaaS editorial · Sources checked · 10 worksheet rows

Open the CSV in Excel, or import it into Google Sheets and choose comma-separated values. Save your working copy in a controlled location. The CSV contains the rows below and space for your notes; it has no macros, formulas or live account connection. The separate Excel workbook contains editable example inputs and formulas that update the forecast.

How to use the template

Write the brand bible

List colours with hex codes, fonts, logo rules, exact product facts, the person or character, the style, three to five do and don't items and the approved claims. The workbook turns these into one prompt block you paste at the start of every prompt.

Collect a reference pack

Photograph the product from three angles on a plain background, keep a transparent logo file and cut-out packshot, make a character sheet with front, three-quarter and profile views, and choose two or three style references. Get written consent before using a real person's likeness.

Generate scenes and composite the truth

Use your tool's reference features for style, product and character. Ask the generator for the scene with a plain label area, then add the real logo, packshot, headline, offer and price in an editor.

Score every ad before it ships

Score each check 0, 1 or 2. Logo, product, text and claims are critical and must score 2. Then the total must reach your threshold. Log the tool, model version, references and seed for each kept image so you can repeat it.

Preview the worksheet

Every row shown here is included in the download. Scroll the table horizontally on a small screen.

CheckWhat to look forHow to test
Logo exactEvery letter, colour and proportion matches the logo file; no extra or ghost logosOverlay the real logo file at 50% opacity; zoom to 200%
Product accurateShape, size, colour, label, pack count and contents match what the customer receivesCompare side by side with the product reference photos
Text, price and offer exactSpelling, numbers, currency, dates and terms match the live offerRead every word aloud at 200% zoom; check against the offer page
Claims approvedEvery claim is on the approved list with proof on file; no implied results you can't supportTick each claim against the approved claims list
Hands and anatomyFive fingers, natural joints, nothing melted, merged or floatingZoom on hands, teeth, ears and where objects touch skin
Face or characterMatches the character sheet across every ad in the setPut this ad next to the character sheet and the last approved ad
Colours in paletteMain colours within the brand palette after gradingUse an eyedropper on 3 key areas; compare with the hex codes
Style matches referencesLight, lens, setting and mood feel like the same campaignView with 3 approved ads in a grid
Placement fitKey content inside safe zones for each placement; readable on a phonePreview in each placement (feed, Stories, Reels) on a real phone
AI disclosure checkedPlatform AI labels and any disclosure your category or market requiresCheck Meta and Google rules for this ad type; note the decision

Worked review example

Illustrative static ad for a fictional tea brand: logo 2, product 2, text and price 1, claims 2, hands 1, face 2, colours 2, style 2, placement 1, AI disclosure 2. The total is 17 of 20, or 85%.

The ad still fails because a critical check scored 1: the offer reads 2O% instead of 20%. The fix is to retype the offer in the editor and score the ad again, not to regenerate the whole image.

What this template does and does not establish

Tool features and limits were checked against official documentation on 16 November 2026 and change often; re-check before relying on a number. Drift ratings and the 85% threshold are editorial examples, not measured rates. The scorecard supports a human review; it does not replace legal review of claims in regulated categories.

Primary references

Continue the work