The structured brand brief: a template for AI inputs
The standard advice is to "put your brand guide in the prompt." What that produces is a wall of description the AI reads, acknowledges, and largely ignores — not because AI is obstinate, but because descriptive guidance doesn't constrain output. The model selects the most plausible continuation of your prompt. If your brief describes how your brand feels, the model generates content that plausibly describes how your brand feels, which is not the same as content that follows your brand's actual rules.
The fix is structural. A brief built around named input fields — not paragraphs — constrains what AI can produce in ways that description cannot. This post provides the template.
Why field structure outperforms prose
An AI system processes your brief as context and generates content that satisfies it. Prose descriptions are satisfied by content that sounds similar to the prose. Named fields with specific constraints are satisfied only by content that actually follows the constraints.
The difference in practice: "We use plain language" is satisfied by any output that doesn't use obvious jargon. "Terminology ban: do not use leverage, ecosystem, scalable, robust, or seamless" is satisfied only by output that contains none of those words. The second version is checkable. The first is not.
Research on instruction-following in language models establishes a related point: compact, named principles produce more consistent adherence than diffuse descriptions. The structure does the work that the quality of the description cannot.
The template
The structured brand brief has seven sections. Use all of them. Leave a field empty rather than filling it with vague description — an empty constraint is neutral. A vague constraint actively misleads.
Section 1: Voice rules
Structural directives for how content is written. Not "be direct" — specify what directness means in your content.
Format: named rule + specific behavior
| Rule name | What it means |
|---|---|
| Lead rule | State the conclusion first. Don't bury the claim. |
| Hedge rule | Use assertions, not hedges. Replace "may help" with "reduces." |
| Length rule | Maximum sentence length: 25 words. Break longer sentences at the conjunction. |
| Pronoun rule | Use "you" throughout. Never "clients," "users," or "customers." |
Fill this with the rules your content actually needs. Three focused rules outperform twelve aspirational ones.
Section 2: Terminology bans
Words and phrases that must never appear, regardless of context.
Format: a flat list
Do not use: leverage, optimize, seamless, robust, innovative, comprehensive,
ecosystem, scalable, cutting-edge, world-class, best-in-class, solution,
synergy, unlock, empower, journey, transform
Review AI content for these words in your first round of editing. Any that appear repeatedly belong on the ban list.
Section 3: Required terms
Words and phrases that must appear in specific contexts.
Format: term + where it belongs
| Term | Usage |
|---|---|
| [Brand name] | First mention in every piece. Never abbreviated. |
| [Product name] | Use exact product name; do not substitute generic category terms. |
| [Signature phrase] | Use in CTAs and closing paragraphs. |
Keep this list short. Required terms that appear in every sentence stop feeling like brand and start feeling like stuffing.
Section 4: Brand facts
Concrete claims about your organization or product that must be stated accurately. These are the facts AI is most likely to hallucinate or generalize.
Format: named fact + exact phrasing
| Fact name | Exact text to use |
|---|---|
| Founded | Founded in [year] |
| Headquarters | Based in [city], [state] |
| Core claim | [Specific, numbered claim — no generics] |
| Differentiator | [The specific feature or approach that distinguishes you] |
AI produces "established" when you say "founded," "located in" when you want "based in," and "improves efficiency" when you have a specific metric. Named facts with exact phrasing prevent each of these substitutions.
Section 5: Audience rules
What your audience already knows, what they don't need explained, and how they want to be addressed.
Format: direct statements
Audience: [Title / role], working in [context]
Assumes: [What they already know — skip the basics]
Never explain: [Concepts that are beneath this audience's level]
Address as: [Peer / expert / someone who has already decided / etc.]
This section prevents the two most common audience failures: over-explaining to experts and under-explaining to generalists. Be specific about both what to include and what to skip.
Section 6: Format rules
Structural patterns for how content is organized.
Format: named rule + specification
| Rule | Specification |
|---|---|
| Lead paragraph | State the problem and the specific answer within the first 3 sentences. |
| Section headers | Sentence case. No generic headers ("Introduction," "Conclusion"). |
| List items | No more than 7 items per list. Never a list of single words. |
| CTA | One per piece, final paragraph, [specific phrasing]. |
Format rules are where AI most consistently produces acceptable-but-wrong output. The model knows what good articles look like in general. It does not know what your content looks like specifically.
Section 7: Off-brand examples
Specific before/after pairs from your actual content. These are the most powerful section and the most commonly omitted.
Format: labeled before/after
❌ Off-brand: "Our innovative AI solution leverages advanced technology to empower
marketing teams with seamless content creation capabilities."
✓ On-brand: "The platform generates on-brand content by injecting structured brand
context into every generation — voice rules, terminology bans, and verified facts."
One pair per major failure mode. The more specific the example, the more useful it is. Generic before/after pairs (fluffy vs. direct) are less useful than examples taken from real content that failed review.
How to fill out the template
Work backwards from content that failed. Take a piece of AI-generated content that went through multiple rounds of revision and identify every edit you made. Each edit is a constraint the brief didn't capture.
Categorize each edit:
- A substituted word → terminology ban
- A restructured sentence → voice rule
- A corrected fact → brand fact
- A removed explanation → audience rule
- A changed header or structure → format rule
Do this for three to five pieces. The brief that emerges will address the actual failure modes, not the hypothetical ones.
How the brief connects to AI generation
Copper Sun's brand context system stores the brief as structured data and injects it into every content generation — not as a paragraph in a prompt, but as named fields the generation process treats as constraints. The effect: you fill out the brief once, and every piece of content your team generates through the platform applies it automatically.
The template above works as a standalone prompt input for any AI tool. In Copper Sun, it becomes persistent brand context that applies across modules and projects without re-pasting. See brand constitution vs. brand guide for the underlying model, and how to capture your brand voice for AI for the process of developing the voice rules section in depth.
Frequently Asked Questions
How is this different from a brand guide?
A brand guide describes your brand for humans who will exercise judgment about application. The structured brief specifies constraints for AI that will follow them literally. The sections are different, the format is different, and the purpose is different — even if some of the underlying content overlaps. You can have both: the brand guide for human onboarding, the brief for AI generation.
How specific do the brand facts need to be?
Specific enough that there is only one correct version. "We are a leading provider" is not a brand fact; any AI can generate that without input. "Founded in [year], serving [specific number] customers across [specific category]" is a brand fact. If a fact can be stated in multiple correct ways, pick one exact phrasing and use that.
How often should the brief be updated?
Update the terminology ban whenever you catch a new word in your AI content that doesn't belong. Update brand facts when facts change. Update voice rules when you notice a structural failure mode your current rules don't address. The brief is a living document — but add to it deliberately, not continuously. A brief updated every week becomes inconsistent. Quarterly review is a reasonable cadence once the initial version is stable.
Can one brief cover multiple content types?
One brief covers one content voice. If your brand produces content in significantly different registers — a technical integration guide and a social campaign, for example — you may need separate briefs with shared terminology bans but different voice and format rules. Start with one brief and split it only when you have documented evidence that a single set of rules produces consistent failures in one content type.