How to keep AI-generated content on brand
Keep agents working from current brand guidance, give each task the right rules and files, and use checks of the finished work to improve the next task.
TL;DR
Keep current brand guidance and assets available to every agent. Give each task the rules, examples and files it needs, then check the finished work against the guidance. Correct failures and review again, using what you find to improve the output, the task instructions or the brand guidance.
Turn brand preferences into rules
Give the agent decisions it can apply:
Instead of Friendly and professional.
Point to approved writing for this situation and explain the decisions to follow: how directly to state the problem, which words to use and when humour belongs.
Instead of Use consistent terminology.
Give the full approved product name and any retired spellings that should stay out of new work.
Instead of Use our brand colours.
Name the permitted text and background pairings. Say where an accent may appear and where it must not.
Instead of Use the logo correctly.
Link the original artwork, its permitted background and the size and clear space required for this placement.
Explain why each rule applies and pair approved examples with near-misses, following Anthropic's guidance. Keep campaign headlines and launch directions with their task, separate from enduring brand rules.
Make the current version available
Saved prompts and settings can retain old rules. Rainbrand maintains shared identity, guidance and assets that agents retrieve through MCP, the Model Context Protocol, an open standard connecting AI applications to tools and data. Products can retrieve the context from code.
For several client brands, keep their material separate and name the intended brand in each request.
Give each task usable inputs
Select the guidance and files for the actual task:
| Work | Guidance to supply | Files and examples to use |
|---|---|---|
| Copy: emails, posts, product text | Voice, terminology, product facts and claims you can support | Approved writing examples |
| Layouts: slides, documents, web pages | Exact colours, typography and logo rules | Logo files, font files and relevant templates |
| Visuals: images and video | Subject, style, composition and exact required words | Approved reference images, logo artwork and type specimens |
For visuals, OpenAI recommends assigning reference roles such as subject, style and background; Gemini accepts several references. Describe what to make and what each file contributes. Spell unusual names letter by letter if needed, and place the original logo rather than asking a model to redraw it.
A Rainbrand Brief selects guidance and assets for the task. Pass them with the original request to your creation tools.
Check the finished work
For juice brand Eager, Claude made a blackcurrant-flavour campaign concept. The task requested a purple carton; Claude also made the page headline purple, which the brand's text rules did not allow.
A colour check flagged the headline. Claude changed it to black and submitted the revised page for another check:
The headline then passed the colour check. Rainbrand supplied guidance and checked the result; Claude made the work and corrections. The requested carton colour did not become a rule for page text.
Compare your finished work with the task and guidance it used:
Copy
- Names, product terms and capitalisation match the guidance.
- Claims are supported, and banned words or phrases are absent.
- Required headlines, calls to action and legal lines appear as written.
Visuals
- The logo is an unaltered approved file, in an allowed colour.
- Colours and type follow the brand's rules for where they are used.
- Imagery follows the guidance for subject, style and composition.
A Rainbrand Review returns brand findings with available evidence. Your agent makes corrections with its tools and requests another check. Inspect uncertain findings or supply clearer references.
Use failures to improve the next task
Before repeating the task, identify what caused the failure:
- If the guidance and files were correct but the output broke a rule, revise the output and check it again.
- If the task left out required copy or supplied the wrong files, fix those inputs for the next request.
- If the maintained guidance is outdated or incomplete, update the brand so later tasks receive the correction too.
For a lasting change, inspect and apply a brand edit, then request a fresh Brief. Retire outdated assets from the Brand Library too. Keep the Brief, output and Review together when trying your next deliverable:
Request a Brief for [brand] for the attached image task, using its audience, dimensions and exact headline. Make the image using the guidance and relevant reference images, and place the approved logo file. Request a Review of the finished composition, use the findings to correct it, then check the revised output.More in AI tools
All 8 AI tools articles4 min read
How to give Claude Code and Cursor your brand, not just your design system
Coding agents read your design system and instruction files, but a landing page also needs your voice, product claims, imagery and logo rules. Keep code conventions in the repo, fetch the current brand for each task, and check the page before it merges.
3 min read
How to use ChatGPT for marketing and stay on brand
Use ChatGPT to research, draft, analyse and make campaign assets from real sources. Keep projects for campaign work, fetch current brand context through a connection, and check the deliverables before publication.
3 min read
How to use Claude for marketing without losing your brand
Give Claude a real marketing job, the sources behind it and current brand guidance. Use Projects, skills and connectors for their different jobs, then inspect the finished work before it goes out.
