Why this is hard to get right
A Marketing Manager Tired of Re-Editing Every Post
Sarah runs marketing for a 40-person B2B SaaS company. She manages a team of three writers, a customer success manager who posts weekly tips, and a founder who fires off LinkedIn updates whenever inspiration strikes. Every week, she spends 90 minutes reviewing posts before they go live — catching phrases like "synergy-driven solutions," softening CTAs that sound like infomercials, and rewriting captions that switch between first and third person mid-paragraph.
The problem isn't talent. It's that there's no shared reference point. When Sarah tells a writer to sound "professional but human," everyone interprets that differently. The founder writes long-winded paragraphs. The customer success manager uses product jargon that confuses prospects. The newest hire writes like they're posting for a lifestyle brand.
Sarah tried building a style guide the traditional way. She spent a weekend in Google Docs writing rules from scratch — "write short sentences," "don't use passive voice," "sound like a trusted advisor." It took six hours, and the result was a generic, two-page document that nobody referenced after the first week. It had no examples, no CTA library, and no channel-specific formatting rules. It described a voice but didn't demonstrate it.
She then tried asking an AI assistant: "Make a brand voice guide for our social media and give some examples." The output was a polished-looking document with zero relevance to her actual company. It used placeholder personas, invented voice traits with no definitions, and generated examples that sounded like a Fortune 500 press release — nothing like the scrappy, credible, expert-led brand she was trying to build.
What changed her results was structuring the prompt with intent. When she defined the persona (mid-level ops professionals on LinkedIn), listed four specific voice traits with one-sentence definitions, specified preferred terms and eight banned phrases, and asked for six posts across real scenarios — product update, customer win, hiring, tip, objection handling, event promo — the AI produced a usable first draft in one pass.
The guide now lives in Notion. Writers open it before drafting, copy the CTA options they need, and check their posts against the do's and don'ts list. Sarah's weekly review dropped from 90 minutes to 20. The founder still writes long-winded posts — but now Sarah has documented proof that short sentences are a brand rule, not a personal preference.
That's the real value of a well-structured brand voice prompt: it turns subjective "sound like us" feedback into concrete, repeatable rules that survive team turnover and scale with your content output.
Common mistakes to avoid
Skipping Channel-Specific Rules
LinkedIn and Instagram have different reading patterns, character limits, and audience expectations. A guide that treats them as one platform produces posts that feel off in both places. Specify each channel separately — pacing, formatting, and CTA style differ enough to need their own rules and examples.
Using Vague Voice Descriptors
Words like 'professional,' 'friendly,' and 'authentic' mean something different to every writer on your team. Without a one-sentence definition and a concrete example for each trait, the guide becomes decorative. Define each voice trait with a behavioral rule — e.g., 'Confident: we state opinions directly, we never hedge with phrases like might or could potentially.'
Omitting Banned Phrases
Most style guides list what to do but skip what to avoid. That leaves writers guessing whether 'synergy,' 'game-changer,' or 'best-in-class' fits the brand. List at least 5 banned phrases with brief explanations of why each one is off-brand. Banned phrases enforce voice discipline faster than any positive rule.
Generating Examples Without Real Scenarios
Generic post examples — 'here's a product update' — don't reflect the situations your team actually faces. When examples don't match real content needs, writers ignore them. Specify the exact post types your team produces (hiring, customer win, tip, objection handling) so examples are immediately reusable.
Forgetting the CTA Library
CTAs are where brand voice breaks down most visibly. Writers default to 'click here' or 'check it out' unless they have approved options. Ask for a CTA library of 6-10 options matched to intent — curiosity, conversion, engagement — so the right CTA is always one copy-paste away.
Not Anchoring Voice to a Specific Persona
A voice guide built for 'everyone' serves no one. Define one primary persona per channel — their role, their pain points, and the vocabulary they use. When the persona is specific, voice traits, examples, and banned phrases all become easier to evaluate and consistently apply.
The transformation
Make a brand voice guide for our social media and give some examples.
You’re a senior brand copywriter. Build a **social media brand voice style guide** for **[Company]**, a **[industry]** brand. 1) Audience: **[primary persona]** on **LinkedIn + Instagram**. 2) Voice traits: choose **4 traits** with a 1-sentence definition each. 3) Rules: list **10 do’s** and **10 don’ts** (active voice only, short sentences). 4) Language: include **preferred terms**, **banned phrases**, and a **CTA library** (8 options). 5) Examples: write **6 on-brand posts** (3 LinkedIn, 3 Instagram), each **80–120 words**, for: product update, customer win, hiring, event promo, tip, and objection handling. Use a professional, approachable tone. Keep everything skimmable with headings and bullets.
Why this works
Role Framing Raises Quality
Opening with 'You're a senior brand copywriter' shifts the AI's output register immediately. This role assignment in the After Prompt signals that decisions about tone, structure, and vocabulary should reflect professional-level brand judgment — not generic content writing. The result is voice traits with actual definitions, not filler.
Numbered Structure Forces Completeness
The After Prompt's five numbered sections — audience, voice traits, rules, language, and examples — prevent the AI from front-loading the easy parts and skimping on the rest. Each section has a specific deliverable count (4 traits, 10 do's, 10 don'ts, 8 CTAs, 6 posts), so nothing gets skipped or generalized.
Named Channels Drive Relevant Examples
Specifying LinkedIn + Instagram in section one means the AI calibrates pacing, post length, and formatting to match each platform's norms. A generic 'social media' prompt produces interchangeable posts. Named channels produce posts that actually fit native reading patterns on each platform.
Scenario Labels Make Examples Reusable
Asking for posts across six named scenarios — product update, customer win, hiring, event promo, tip, and objection handling — turns the examples section into a practical template library. Writers don't have to adapt abstract examples. They find the scenario they need and copy it directly.
Word Count Constraints Reduce Rewrites
The 80-120 word range in the After Prompt does two things: it stops the AI from producing bloated examples, and it gives writers a target that matches real platform behavior. Tight constraints at the prompt level eliminate an entire round of length-editing after the output lands.
The framework behind the prompt
The Theory Behind Brand Voice Style Guides
Brand voice is one of the most researched areas in content strategy — and one of the least consistently executed. The gap usually comes down to documentation quality, not writer talent.
Why Voice Consistency Fails at Scale
Organizational communication research consistently shows that brand voice degrades as team size grows. When one person writes all content, voice is an intuition. When five people do, it becomes a negotiation. The solution isn't more editing — it's operationalizing the intuition into rules writers can apply independently.
This is where the COPE framework (Create Once, Publish Everywhere) becomes relevant. A well-structured voice guide functions as a content primitive — a foundational asset from which all channel-specific output inherits its rules. Without it, each platform gets rebuilt from scratch by whoever writes that week.
Voice vs. Tone: A Critical Distinction
Most teams conflate voice and tone. Voice is stable — it's your brand's personality, the consistent character that doesn't change whether you're announcing a product or responding to a complaint. Tone shifts by context — celebratory for a product win, measured for a sensitive topic. A strong style guide defines the voice and then shows how tone flexes within it. Prompts that don't make this distinction produce guides where the "voice traits" are really just tone adjustments, which leaves writers without a stable foundation.
The Role of Behavioral Definitions
Research in instructional design (grounded in Bloom's Taxonomy) shows that behavioral objectives outperform abstract ones. "Write clearly" is an abstract objective. "Use one idea per sentence and never exceed 20 words" is behavioral. The same principle applies to brand voice rules. Traits defined as behaviors — what we do, not what we are — produce more consistent output from both human writers and AI systems.
Why AI Needs More Context, Not Less
Large language models are excellent at pattern-matching to genre conventions. When you prompt for a "brand voice guide," the model draws on every generic style guide in its training data. The result is polished but impersonal. Specificity — persona, channel, banned phrases, real scenarios — is what forces the model off the generic template and onto something that reflects your actual brand. This is why structured prompts with concrete constraints consistently outperform open-ended ones for brand documentation tasks.
Prompt variations
You're a senior brand strategist. We're a Series A fintech startup transitioning from a developer-first brand to a CFO and finance-leader audience.
Build a social media voice guide for this rebrand covering:
- Audience shift: define the old voice (technical, jargon-heavy) vs. the new voice (strategic, outcome-focused) side by side.
- Voice traits: write 4 new traits, each with a definition and a before/after sentence showing the old vs. new tone.
- Banned phrases from our old brand: list 8 phrases we must retire (e.g., API-first, developer-friendly) and 8 replacement phrases.
- LinkedIn-specific rules: 5 formatting rules and 3 structural post templates.
- Launch examples: 4 LinkedIn posts announcing the new positioning — one for each of these topics: product vision, customer outcome, team culture, and market POV.
Keep every post under 150 words. Use confident, plain language. No jargon.
You're a brand copywriter onboarding a new social media client. Build a reusable voice intake document and starter guide for a professional services firm (management consulting, 80 employees, targeting VP-and-above buyers at mid-market companies).
- Voice intake questions: write 10 questions an agency should ask the client to extract voice preferences, proof points, and content red lines.
- Draft voice guide based on this profile: consultative authority, no hype, data-backed claims, no first-person plural (avoid 'we believe').
- LinkedIn rules: 5 do's, 5 don'ts, max 120 words per post.
- CTA library: 6 approved CTAs sorted by intent (awareness, consideration, conversion).
- Starter examples: 3 LinkedIn posts — one thought leadership piece, one client result, one team hiring post.
Format with bold headings. Keep examples under 100 words each.
You're a senior social copywriter specializing in direct-to-consumer brands. Build an Instagram-only brand voice guide for a sustainable home goods brand targeting millennial homeowners aged 28-40.
- Voice traits: 4 traits with one-sentence definitions. Focus on warmth, honesty, and environmental conviction without preachiness.
- Caption rules: 6 do's and 6 don'ts specific to Instagram captions — cover length, emoji use (none), hashtag policy, and first-line hooks.
- Banned phrases: 8 phrases to avoid (e.g., eco-friendly, sustainable living, green) with plain-language replacements for each.
- Hashtag strategy: 3 hashtag clusters of 5 tags each — product, community, and brand.
- Caption examples: 5 captions (90-110 words each) for: new product launch, customer photo repost, behind-the-scenes, sale announcement, and brand values.
Tone: warm, specific, never preachy.
You're a brand editor. Create a one-page social media voice reference card for a healthcare technology company's internal content team. The audience is HR leaders and benefits managers at companies with 500+ employees.
Format it as a scannable single-page reference with four sections:
- Our voice in 3 words — followed by one sentence defining each word as a behavior, not an adjective.
- Quick rules — 5 active-voice do's and 5 don'ts in plain language, max 10 words each.
- CTA options — 6 CTAs sorted by goal: schedule a call, read more, share with your team.
- Quick-check list — 5 yes/no questions a writer can ask before posting (e.g., 'Does this claim have a number behind it?').
Keep the whole card under 400 words. Use bullet points throughout. No paragraphs.
When to use this prompt
Marketing Teams With Multiple Writers
Standardize captions and CTAs so every post sounds like one brand, even with several contributors.
Product Managers Sharing Updates
Create consistent product update posts that match brand tone and avoid overpromising.
Customer Success Teams Posting Proof
Write customer wins and tips that feel helpful, not salesy, while keeping language consistent.
Sales Leaders Building Social Selling
Align objection-handling posts and value points with approved messaging and CTAs.
Founders Preparing for a Rebrand
Document new voice rules fast and ship updated examples across channels in one pass.
Pro tips
- 1
Define one primary persona per channel so your examples match real reader intent.
- 2
List 5 banned phrases to prevent vague claims and reduce brand risk.
- 3
Add 3 proof points with numbers so writers don’t rely on empty adjectives.
- 4
Specify 2 formatting rules per platform so posts fit native reading patterns.
A brand voice guide only works if your team actually uses it. Here are techniques that improve adoption when the guide is AI-generated:
Anchor voice traits to real decisions. For each trait, add one sentence describing a specific situation where it applies — e.g., 'When announcing a price increase, use the Candid trait: lead with the fact, then the reason, then the next step.' Situational anchors turn abstract traits into applied judgment.
Include a rejection list in onboarding. Alongside your banned phrases list, add a one-paragraph explanation of why those phrases are off-brand. Writers who understand the reasoning self-correct faster than writers who memorize a list.
Build a living examples archive. After each strong post goes live, tag it with the voice traits it demonstrates and add it to a shared folder. Over time, real posts become better training examples than AI-generated ones.
Run a quarterly voice audit. Pull 10 recent posts, score each one against the do's and don'ts list, and flag the 2-3 rules that fail most often. Update the guide to strengthen those sections. Voice guides that don't evolve get ignored within six months.
Test new writers with a structured exercise. Give them one scenario (e.g., customer win on LinkedIn) and compare their draft against the AI-generated example. Use the gap as a coaching moment, not a correction.
Brand voice guides for regulated industries — healthcare, finance, legal, insurance — need extra constraints that most prompts skip. Here's how to adapt the After Prompt for these contexts:
Add a compliance layer to the rules section. After listing 10 do's and 10 don'ts, add a third list: 5 compliance rules. Examples: 'Never state specific ROI percentages without a footnote,' or 'All posts referencing clinical outcomes must include a qualifier.' This keeps the guide legally usable, not just stylistically consistent.
Restrict the CTA library to approved language. In highly regulated industries, CTAs like 'Get started today' can imply a commitment your legal team hasn't approved. Ask the AI to generate CTAs that use action-oriented but non-binding language — 'Learn how,' 'See the data,' 'Read our report.'
Flag audience-sensitive scenarios. If your audience includes patients, consumers in financial hardship, or people in vulnerable situations, add an instruction: 'For any post that could reach audiences in distress, apply a 'caution' tag and provide an alternative empathetic version.' This forces the AI to surface edge cases proactively.
Separate internal and external voice guides. B2B financial services companies often need one voice for LinkedIn (thought leadership, formal) and a different one for email (direct, action-oriented). Use the core prompt twice with a note about which channel each version serves.
Use this checklist to validate any AI-generated brand voice guide before distributing it to your team:
Voice traits:
- Each trait has a one-sentence behavioral definition
- No trait overlaps with another (confident vs. direct should mean different things)
- Each trait is demonstrated in at least one example post
Rules:
- Do's are written in active voice
- Don'ts include specific examples of what to avoid, not just categories
- At least 3 rules reference formatting, not just vocabulary
Language:
- Banned phrases include at least 3 specific words your team actually uses
- CTA library covers at least 2 distinct intents (awareness and conversion)
- Preferred terms reflect your actual product language, not generic industry terms
Examples:
- Each example matches the word count specified in the prompt
- LinkedIn and Instagram examples have different structural formats
- At least one example handles a sensitive scenario (pricing, objection, bad news)
Usability:
- A new team member could follow the guide without asking a question
- The guide fits in one Notion page or equivalent shared doc
- Version date is recorded so you know when to refresh it
When not to use this prompt
This prompt pattern is not the right tool in every situation. Here are cases where you should pause or use a different approach:
Your brand voice is still unresolved. If leadership disagrees on whether the brand should be formal or casual, an AI-generated style guide will just document the conflict in trait form. Resolve the strategic voice questions with your team first — then use this prompt to codify what you've agreed on.
You need a full brand identity system. This prompt builds a social media voice guide, not a complete brand standards document. If you also need visual identity rules, logo usage, typography guidelines, or email voice standards, use separate prompts scoped to each deliverable.
Your audience is highly regulated and legal hasn't signed off. In healthcare, finance, or legal services, an AI-generated guide should be treated as a first draft only. Route every claim, CTA, and example through compliance review before distributing to your team.
You're writing for a non-English audience. The structural logic of this prompt holds, but cultural tone norms vary significantly by language and region. An AI working in Spanish, French, or Japanese may apply English-language brand voice conventions that don't translate. Add cultural context explicitly or involve a native-speaker reviewer before finalizing.
Troubleshooting
The AI generates voice traits that are too generic (confident, authentic, clear)
Block the overused defaults directly in the prompt. Add this line after the voice traits instruction: 'Do not use the following as traits: authentic, innovative, transparent, approachable, or clear.' Then add: 'Each trait must describe a behavior, not a quality — write it as what we do, not what we are.' This forces the AI toward specific, actionable language.
The do's and don'ts list feels like general writing advice, not brand-specific rules
Add brand context before the numbered sections. Write 2-3 sentences about your actual product, audience pain points, and your top competitor. Then specify: 'All do's and don'ts must reference our specific audience or product category.' Generic rules appear when the AI has no brand anchor to attach them to — context fixes that.
LinkedIn and Instagram examples sound identical — same length, same structure
Add a formatting rule for each channel before the examples section. Specify: 'LinkedIn examples use a hook sentence, 2-3 short paragraphs, and a question-based CTA. Instagram examples use a single-line hook, one short paragraph, and a link-in-bio CTA.' Structural templates force channel differentiation more reliably than tone instructions alone.
The CTA library repeats the same 2-3 variations with different wording
Sort the CTA request by intent before asking for options. Add this instruction: 'Write 2 CTAs for awareness, 2 for consideration, 2 for conversion, and 2 for community engagement — 8 total, none using the words click, explore, or discover.' Categorizing by intent forces variety. Banning overused verbs prevents lazy defaults.
The output is well-structured but runs to 2,000 words — too long for team use
Add a hard length limit to the prompt: 'The entire guide must fit within 800 words, excluding examples. Use bullets throughout — no paragraphs in the rules or language sections.' Then add: 'If any section exceeds 150 words, cut it before finalizing.' Length constraints inside the prompt are more reliable than asking the AI to 'be concise.'
How to measure success
How to Evaluate Your AI-Generated Voice Guide
A high-quality output from this prompt type should pass the following checks:
Completeness:
- All five sections from the After Prompt are present and fully populated
- Voice traits each have a one-sentence behavioral definition (not just a label)
- Do's and don'ts lists contain 10 items each, written in active voice
- CTA library has at least 6 distinct options across multiple intents
Relevance:
- Examples match the channel format (LinkedIn posts feel different from Instagram captions)
- Voice traits could not apply to any competitor without significant edits
- Banned phrases reflect real vocabulary risks for your category, not generic filler
Usability:
- A new writer could apply the guide without asking a follow-up question
- The guide fits a single shareable document with clear headings
- Examples cover all 6 specified scenarios with no duplicated structure
Calibration:
- Word counts on examples fall within the 80-120 word range
- No post in the examples section uses passive voice
- The CTA library does not repeat the same verb twice
Now try it on something of your own
Reading about the framework is one thing. Watching it sharpen your own prompt is another — takes 90 seconds, no signup.
Build a brand voice guide your whole team can follow — with traits, rules, and ready-to-use post examples.
Try one of these
Frequently asked questions
Start by answering the audience and channel sections first — those are easier to define than voice traits. Then ask the AI to draft 3 candidate voice trait sets based on your audience description and let you choose. Use the After Prompt structure, but add an instruction like 'generate 2 alternative trait options for section 2 so we can compare and choose.' That turns a definition problem into a selection problem.
Yes. Replace the company name with your name and title, define your audience as the professionals you're writing for, and adjust the examples to reflect personal perspective posts (first person singular). One key change: add a 'perspective and beliefs' section listing 3-5 POVs you hold publicly — this grounds the voice guide in your actual opinions, not generic professional advice.
Add a hard word count to the examples instruction. Instead of '80-120 words,' write 'each caption must be 60-80 words maximum, not including hashtags.' If it still overshoots, add: 'If any example exceeds 80 words, rewrite it before finalizing.' Explicit rewrite instructions inside the prompt catch length issues before the output lands.
Review it quarterly or after any major brand change — new product line, rebrand, audience shift, or change in competitive positioning. The voice traits section stays relatively stable, but the banned phrases list, CTA library, and examples should be refreshed as your market language evolves. Keep a version log in Notion or Google Docs.
Run the prompt twice — once for each channel — with the same voice traits but different formatting rules, example scenarios, and CTA libraries. Do not try to build both channels in one prompt pass. Combining them produces averaged-out rules that serve neither platform well. Two focused outputs are faster to implement than one unfocused guide.
Add a 'brand context' section before the numbered instructions. Write 3-5 sentences covering: what your product does, who it's for, your top 2-3 proof points with real numbers, and 1-2 competitors you're positioned against. Concrete product context is the single biggest driver of example relevance. Without it, the AI defaults to fictional stand-ins.
List 3 traits you explicitly do NOT want. Add this line after section 2: 'Do not use the following as voice traits: authentic, innovative, transparent, approachable.' Blocking overused defaults forces the AI to reach for more specific, brand-relevant language. You can also supply one example of a brand whose voice you admire as a reference point.