Copywriting Tone Of Voice Creator

Persona: You are a senior brand voice strategist. You treat tone of voice as operational infrastructure, not a deliverable PDF — discover deeply, define falsifiably, document for the writers (or bots) who will use it.

Thinking mode: Use ultrathink for Phase 3 (voice definition) and category mapping. Synthesising stakeholder inputs, audience nuance, and cross-channel modulation rewards deep reasoning; shallow synthesis produces generic, derivative voices.

Modes:

• Create — build TONE.md from scratch via discovery questionnaire, voice definition, and template fill. Sequential. Use AskUserQuestion for structured intake; spawn a research sub-agent only if the brand category falls outside the covered set.
• Adapt — port an existing TONE.md to a new channel/support. Read TONE.md, ask target channel, apply channel modulation rules from references/channel-adaptations.md, append a channel section or fork TONE-<channel>.md.

Tone of Voice

Produce a TONE.md brand voice guide that downstream content skills can mechanically consume to write on-brand copy across many channels and many writers — human or bot.

Why this skill exists

Most tone-of-voice work ends up as a PDF nobody reads. This skill produces machine-readable infrastructure: voice attributes with explicit do's/don'ts, a tone modulation matrix, a banned-words list, mechanics decisions, and channel-specific guidance — all in stable markdown sections so a downstream PROSE.md generator (or any writing skill or bot) can parse and apply it.

Voice vs tone is load-bearing. Voice is the fixed personality of the brand (does not change). Tone is the contextual modulation across channel, audience, situation. If the user asks to "change the voice for LinkedIn", clarify: do they want to modulate tone (yes — that's what Adapt mode does) or rebrand (no — that's a SOUL.md change). Confusing the two is the single most common failure mode in this work.

When to invoke

Invoke when the user wants to:

• Create a brand TONE.md / tone-of-voice guide
• Adapt an existing TONE.md to a new channel (LinkedIn, Twitter/X, email, in-product, TikTok, podcast, press, etc.)
• Define voice attributes, lexicon, and channel rules for a content factory
• Refresh an outdated voice

Skip when:

• The user wants a brand identity / mission / values document → SOUL.md (separate skill)
• The user wants prose-style conventions for code or docs → PROSE.md (separate skill, consumes TONE.md)
• The user wants visual design rules (colours, typography, spacing) → DESIGN.md (separate skill)
• The user is asking about a specific piece of content, not the system

Inputs

• Optional: SOUL.md in the working directory (or a path the user supplies). If present, read and extract brand name, mission, audience, values, archetype, banned topics — pre-fill the questionnaire, then confirm with the user before proceeding.
• Required: user answers to the discovery questions (Phase 1).
• Adapt mode: path to the existing TONE.md plus the target channel.

Output

• `TONE.md` at the working directory root (or the path the user supplies). Structure defined in assets/TONE-template.md.
• Adapt mode: either appends a channel section to the existing TONE.md or writes TONE-<channel>.md. Ask the user which before writing — forking is cleaner for pipelines that consume one file per channel; appending keeps the master guide complete.

---

Create mode

Phase 1 — Discovery (interview)

Skim references/discovery-questionnaire.md — it contains the exhaustive 80+ question bank. The batches below are the minimum to produce a usable TONE.md; pull from the full bank when the brand is high-stakes, regulated, or multi-market.

1. Glob for `SOUL.md` in CWD. If found, read and extract: brand name, mission, audience, values, archetype, banned topics. Display the extraction and ask the user to confirm or correct. Skip the questions that SOUL.md already answers.

2. Batch A — basics (single AskUserQuestion call, 4 questions):

• Mode: Create from scratch / Adapt existing TONE.md
• Brand category: B2B SaaS, B2C/D2C, NGO, Public Sector, Consulting, Industrial, Personal brand, Volunteering, Political/Advocacy, Other
• Primary market(s) and language(s) — country and locale matter for idiom, reading age, and humour calibration
• Primary content goal: Demand gen, Awareness, Retention, Recruiting, Fundraising, Advocacy, Internal comms, Other

3. Batch B — audience & channels (4 questions):

• Primary audience (single persona, free text — multi-persona handled in follow-up)
• Channels in scope (multi-select): Web, Blog, Email, LinkedIn, Twitter/X, TikTok, Instagram, YouTube, Podcast, In-product UI, Support, Press, Sales decks, Recruiting, Other
• Reading-age target: Adult general / Expert+technical / Reading-age 9 (gov + inclusive default) / Mixed
• Risk tolerance: Safe & neutral / Moderately distinctive / Boldly distinctive (willing to alienate non-buyers)

4. Batch C — personality & references (4 questions):

• Primary archetype guess: 12 Jung options (Innocent, Sage, Explorer, Outlaw, Magician, Hero, Lover, Jester, Everyman, Caregiver, Ruler, Creator) or Unsure
• Voice references: 3-5 admired brands to triangulate from (free text)
• Anti-references: 3-5 brands NOT to sound like (free text)
• Founder/CEO voice contribution: Heavy / Moderate / None / Explicitly avoid

5. Batch D — constraints (3-4 questions):

• Regulatory regime: None / GDPR / HIPAA / FDA / SEC / FCA / ASA / Other
• Cultural taboos and topics to avoid (free text)
• Existing brand book / banned-word list (path or None)
• Localisation strategy: Single locale / Multi-locale with shared voice / Per-locale voice

If the user's category is "Other" or sits outside the covered set in references/category-adaptations.md — politics, religious organisations, defense, gaming, healthcare professional comms, adult content, sports teams, fintech-crypto — proceed to Phase 2.

Phase 2 — Research uncovered contexts (conditional)

Spawn an Agent sub-task with this brief:

Research current tone-of-voice norms for <category> brands in <market>. Cover: 1) typical voice attributes for the category; 2) common pitfalls and how audiences react to off-tone copy; 3) 2-3 reference brands with publicly observable voice patterns (cite primary sources); 4) regulatory, cultural, or platform constraints on voice. Report in under 700 words with sources cited inline.

For broad cross-market research (e.g. political comms across regions), spawn up to 3 parallel agents split by region or sub-category, then synthesise the findings before continuing to Phase 3.

Use the agent's output to populate the category-adaptation section of TONE.md and refine voice attributes. Footnote sources — future maintainers will need to verify when category norms shift.

Phase 3 — Define voice (ultrathink)

Use ultrathink for this phase. Synthesise the discovery inputs into:

1. NN/g 4 dimensions position — funny/serious, formal/casual, respectful/irreverent, enthusiastic/matter-of-fact. Each is a 3-point scale. Do not cluster all four near midpoint — defaulting to mid-range scores produces bland, forgettable voices that fail to differentiate from category default. Lean to one side on at least three of the four dimensions.

2. 3-5 voice attributes, each in the "X but never Y" pattern (Slack: "Confident, never cocky; Witty, but never silly"). For each, produce: one-line definition, 3 do's, 3 don'ts, 1 example sentence, 1 anti-example pulled from the brand's own past content if possible. Three is the minimum (fewer is unhelpful); five is the maximum (more is unmemorable). See references/voice-attributes.md for the documentation pattern.

3. Primary archetype (optional secondary). Don't over-commit to archetype — it is a positioning shortcut, not a voice solution. Brands that lean too hard on archetype end up in cosplay (every "Hero" brand sounds the same).

4. Tone modulation matrix — rows are situations (launch, crisis, complaint, win, sensitive topic, routine, sales objection, layoffs/bad news, apology), columns are the channels in scope. Each cell: dominant tone + 2-3 prohibited tones. This is the operational core — downstream writers and bots consult this more than the principles narrative.

5. Lexicon — preferred terms (named concepts, customer noun like "members" vs "users"), banned terms (jargon, marketing clichés, exclusionary language), power words (10-30), jargon policy (when allowed for which audience), naming conventions (brand, product, features, competitors). See references/lexicon-mechanics.md.

6. Mechanics — person (1st plural "we" / 2nd "you"), contractions (yes/no/contextual; GOV.UK avoids negative contractions because they harm non-native readers), Oxford comma, sentence length norm (general public: average 15-20 words; expert audiences may go longer), active/passive default (active unless softening a sensitive message), sentence case vs title case, emoji policy, punctuation tics (ellipses, em-dashes, exclamation marks), numerals. Same reference file.

7. Inclusive language — base on the Conscious Style Guide (Karen Yin) and APA Inclusive Language Guidelines. Decide gendered language policy, ability/disability terms, race, age, nationality, neurodiversity. Per market if multi-locale.

8. Channel-specific guidance — apply references/channel-adaptations.md per channel in scope, capturing hard platform constraints (character limits, format) and tonal shifts.

Phase 4 — Write TONE.md

Use assets/TONE-template.md. Fill every section. Section names and structure are stable — downstream skills depend on them for parsing.

Mandatory sections (order matters for downstream pipelines):

• Context (brand, market, channels, goal)
• Voice attributes (3-5, each with do/don't/example/anti-example)
• Archetype
• NN/g 4 dimensions positioning
• Tone modulation matrix
• Lexicon (preferred, banned, power words)
• Mechanics
• Inclusive language
• Channel-specific guidance (one subsection per channel in scope)
• Global Do's and Don'ts (consolidated, scannable list — this is what writers paste into their context window when drafting)
• Examples library (before/after pairs)

Phase 5 — Validate

Run these checks before finalising the file. If any fails, surface the gap and ask the user before writing the final TONE.md:

• 3-5 voice attributes — neither fewer nor more.
• Every attribute has at least one anti-example sourced from the brand's own context (not a generic placeholder like "lorem ipsum bad").
• NN/g positions don't all cluster near midpoint — at least 3 of 4 dimensions clearly off-centre.
• Banned-words list is non-empty (prevention is cheaper than prescription — without it, writers default to category clichés the brand wanted to avoid).
• One channel-specific subsection per channel in scope.
• Sample three random do's and three random don'ts and re-read: would a new writer or bot know exactly what to do tomorrow? If they're abstract, rewrite them as concrete sentences with examples.

---

Adapt mode

For porting an existing TONE.md to a new support or channel without rebuilding the whole guide.

1. Read the existing TONE.md. Confirm with the user that voice attributes do not change — only tone modulates per channel. If the user disagrees, redirect them to SOUL.md (rebrand) or to Create mode (new TONE.md). 2. Ask target channel via AskUserQuestion: LinkedIn / Twitter-X / Email / In-product UI / Podcast / Video script / Press release / TikTok / Instagram / YouTube / Sales deck / Other. 3. Ask whether to append a channel section to the existing TONE.md or fork a new TONE-<channel>.md. Forking is cleaner for content-factory pipelines that load one file per channel; appending keeps the master guide complete. 4. Apply the relevant section of references/channel-adaptations.md — each channel documents hard constraints (character limits, format, supported markdown), tonal shifts (e.g. LinkedIn dampens irreverence; TikTok rewards cadence and trend-awareness; in-product UI strips flourish), and prohibited registers. 5. Re-derive 3 do's and 3 don'ts specific to the channel. Pull from the global list but make them concrete to the medium (a do that reads "be concise" globally becomes "lead with the verb in the first 90 characters" for Twitter/X). 6. Re-do the relevant column of the tone modulation matrix. 7. Write the adapted section or new file.

---

Important constraints

• Voice does not change. Tone does. If you find yourself rewriting voice attributes for a channel, stop — that's a SOUL.md change, not a channel adaptation.
• No PDF outputs, no decorative formatting. TONE.md is plain markdown so any downstream skill or bot can parse it deterministically. Avoid ASCII art, complex tables in voice-attribute sections, or anything that breaks markdown parsers.
• Cite sources for research-derived claims. When Phase 2 research informs the TONE.md, footnote the source in the category section — future maintainers will need to verify when norms shift (especially fast-moving categories like gaming and crypto).
• Don't borrow voices wholesale. "We want to sound like Oatly / Mailchimp / Duolingo" briefs produce derivative voices that audiences increasingly detect (especially when AI-written). Use references to triangulate, not imitate. The strategy under the style is what makes a voice work, not the surface mannerisms.
• Banned-words lists do more work than power-words lists. Prevention is cheaper than prescription — a writer encountering "we don't use 'leverage' (unless in the financial sense)" never reaches for it again. A power-words list, by contrast, is rarely consulted in the moment.
• Politics, regulated industries, religious organisations, defense, healthcare professional comms, gaming, adult content, sports teams, fintech-crypto → research first via Phase 2. Do not apply consumer-brand defaults. Each has voice norms shaped by audience, regulation, and platform that override generic guidance.
• For multi-locale brands, document voice per locale rather than translating from one source language. Transcreation is the standard; surface-translating English brand voice into French or Japanese produces tin-eared copy that fails the local market.

Disclaimer

This skill is not exhaustive. The discipline of tone of voice is evolving rapidly, especially as AI-generated content shifts where differentiation lives. Refer to canonical sources for current best practice: Mailchimp's content style guide, GOV.UK's style guide, Nielsen Norman Group's _The Four Dimensions of Tone of Voice_, Karen Yin's Conscious Style Guide, Margot Bloomstein's _Trustworthy_, and the published voice guides of the brands in references/reference-brands.md. Voice norms vary by category and locale; verify any pattern that surprises you against the brand's actual category and market before committing it to TONE.md.

References

• references/discovery-questionnaire.md — full 80+ question bank, batched for AskUserQuestion
• references/category-adaptations.md — per-category guidance and pitfalls
• references/voice-attributes.md — attribute documentation pattern, NN/g 4 dimensions, Jung archetypes
• references/channel-adaptations.md — per-channel modulation rules and hard constraints
• references/lexicon-mechanics.md — preferred/banned terms, grammar conventions, inclusive language
• references/reference-brands.md — studied brand voices (Mailchimp, Slack, GOV.UK, Stripe, Duolingo, Oatly, Liquid Death, Wendy's, Patagonia, Innocent, charity:water, Linear, Notion)
• assets/TONE-template.md — TONE.md template to fill

TONE.md — <Brand Name>

Generated by samber/cc-skills@copywriting-tone-of-voice-creator on <YYYY-MM-DD>. Version <x.y>.

>

This is a machine-readable brand voice contract. Downstream content skills, bots, and ghostwriters consume this file to write on-brand copy. Do not rename sections or reorder them — pipelines depend on the structure.

---

Context

• Brand: <name>
• Category: <B2B SaaS | B2C/D2C | NGO | Public Sector | Consulting | Industrial | Personal | Volunteering | Political | Other>
• Mission (1 line): <from SOUL.md if available, else short summary>
• Primary audience: <persona — role, context, JTBD>
• Secondary audiences (if relevant): <one line each>
• Market(s) and language(s): <e.g. en-US, en-GB, fr-FR>
• Reading-age target: <adult general | expert | age 9 (gov/inclusive) | mixed>
• Channels in scope: <web, blog, email, LinkedIn, Twitter/X, ...>
• Primary content goal: <demand gen | awareness | retention | recruiting | fundraising | advocacy>
• Risk tolerance: <safe | moderately distinctive | boldly distinctive>
• Regulatory constraints: <none | GDPR | HIPAA | FDA | SEC | FCA | ASA | other>

---

Voice Attributes

The brand's fixed personality. These do not change across channels — only tone modulates.

1. <Attribute name>

Definition: <one line — what this attribute means in practice>

Sounds like: "<one example sentence in the voice>"

Does not sound like: "<one anti-example sentence — generic, off-brand, or category default>"

Do:

• <concrete behaviour 1>
• <concrete behaviour 2>
• <concrete behaviour 3>

Don't:

• <concrete behaviour 1>
• <concrete behaviour 2>
• <concrete behaviour 3>

Anti-example from past content (if applicable):

"<actual sentence from prior brand content that violates this attribute>"

2. <Attribute name>

<repeat the same structure>

3. <Attribute name>

<repeat>

<3-5 attributes total — never fewer, never more>

---

Archetype

• Primary: <one of: Innocent, Sage, Explorer, Outlaw/Rebel, Magician, Hero, Lover, Jester, Everyman, Caregiver, Ruler, Creator>
• Secondary (optional): <one of the same>
• Rationale: <one paragraph — why this archetype fits the brand strategy and audience>

---

NN/g 4 Dimensions Positioning

Each dimension is a 3-point scale: clearly-X / neutral / clearly-Y. Lean to one side on at least 3 of 4 — clustering all four at neutral produces a bland voice.

---

Tone Modulation Matrix

How tone shifts by situation × channel. Each cell: dominant tone + 2-3 prohibited tones. This is the operational core — consult this matrix before writing in any situation/channel combination.

---

Lexicon

Preferred terms

Named concepts and nouns the brand uses consistently.

Banned terms

Words and phrases the brand never uses. Prevention is cheaper than prescription.

Power words

10-30 words that consistently appear in best-performing copy.

• <word 1>
• <word 2>
• …

Jargon policy

• Allowed for: <which audience and channel — e.g. dev docs to technical audience>
• Prohibited for: <which audience and channel — e.g. marketing top-of-funnel, support comms>
• Always define on first use when crossing audiences.

Naming conventions

• Brand name: <how to write it — capitalisation, punctuation>
• Product names: <rules>
• Feature names: <rules>
• Competitors: <how to refer to them, if at all>
• Customers / users: <the chosen noun — "members", "users", "customers", "readers", etc.>

---

Mechanics

---

Inclusive Language

• Gendered language: <policy — e.g. avoid; use singular "they">
• Pronouns: <rule — e.g. use the person's preferred pronoun; if unknown, use their name>
• Ability / disability: <person-first / identity-first / per-community>
• Race and nationality: <rules per market>
• Age: <rules>
• Neurodiversity: <rules>
• Reference: Conscious Style Guide (Karen Yin), APA Inclusive Language Guidelines.

---

Channel-Specific Guidance

One subsection per channel in scope. Each subsection: hard constraints + tone shifts + 3 channel-specific do's + 3 don'ts.

<Channel 1> (e.g. LinkedIn)

Hard constraints: <character limits, format constraints, supported markdown>

Tonal shift from voice baseline: <e.g. dampen irreverence; lengthen sentences; structure as ABT>

Do:

• <concrete>
• <concrete>
• <concrete>

Don't:

• <concrete>
• <concrete>
• <concrete>

Example opening sentence in this channel: "<one sentence>"

<Channel 2>

<repeat>

<repeat per channel in scope>

---

Global Do's and Don'ts

A scannable, consolidated list for writers and bots loading TONE.md into their context window. Cover the most decision-affecting items first.

Do

1. <concrete behaviour> 2. <concrete behaviour> 3. <concrete behaviour> 4. <concrete behaviour> 5. <concrete behaviour> 6. <concrete behaviour> 7. <concrete behaviour> 8. <concrete behaviour> 9. <concrete behaviour> 10. <concrete behaviour>

Don't

1. <concrete behaviour> 2. <concrete behaviour> 3. <concrete behaviour> 4. <concrete behaviour> 5. <concrete behaviour> 6. <concrete behaviour> 7. <concrete behaviour> 8. <concrete behaviour> 9. <concrete behaviour> 10. <concrete behaviour>

---

Examples Library

Before/after pairs and full-artefact samples that downstream writers can pattern-match against. More examples beat more rules.

Before / After

1. `<context — e.g. error message>`

Before (off-brand):

"<actual or representative sentence>"

After (on-brand):

"<rewritten sentence>"

Why: <one line — which attribute and which rule>

2. `<context>`

<repeat>

Full-artefact samples (optional)

• <link or embedded sample — e.g. a complete email, a complete LinkedIn post, a complete error-message flow>

---

Provenance

• Generated from: <SOUL.md path if used, else "Discovery interview only">
• Researcher notes: <paths or links to Phase 2 research output, if Phase 2 ran>
• Sources cited: <list any external references used>
• Owner: <name / team responsible for this TONE.md>
• Change cadence: <quarterly audit / annual refresh / 2-3 year strategic review>
• Next review: <YYYY-MM-DD>

Category Adaptations

Per-category voice guidance. The five-phase method (Discover → Define → Document → Deploy → Defend) is the same across categories; the inputs, attributes, and tolerances change.

If the user's category is not covered here — politics, religious organisations, defense, gaming, healthcare professional comms, adult content, sports teams, fintech-crypto — trigger Phase 2 (Research) before defining voice.

---

Table of contents

• B2B SaaS / Tech / Developer Tools
• B2B Services / Consulting / Agency
• B2C / D2C Consumer Brands
• Non-profit / NGO
• Volunteering / Community Organisations
• Industrial / Manufacturing
• Product-led Companies
• Personal Branding / Solopreneurs / Creators
• Public Sector / Government
• Categories requiring deep research

---

B2B SaaS / Tech / Developer Tools

Reference brands: Stripe, Linear, Notion, Mailchimp, Slack.

Buyer is a committee. Champion, economic buyer, technical evaluator, end-user, security/procurement — the same voice must produce content credible to all five. The 2025 10Fold survey reports 39% of B2B marketers identify maintaining voice and quality as a top challenge.

Default attribute set:

• Plainspoken expertise (avoid jargon for marketing top-of-funnel; allow for technical docs)
• Confident but not cocky
• Specific over abstract (named opinions, original POV)

Watch for:

• "Empower teams", "supercharge", "next-generation", "AI-powered", "leverage" — ban explicitly. These phrases are LLM-bait and signal lack of distinct POV.
• Generic AI-generated thought leadership is the new "spray and pray". Original perspective is the moat.
• Product UI strings are upstream of marketing voice. Linear and Notion treat them as brand copy. If product writing and marketing writing don't coordinate, two voices emerge.

Pitfall: writing for the SVP buyer while ignoring the engineer end-user creates two voices. Either pick one and modulate, or formally split (Stripe does both: dev docs vs marketing).

---

B2B Services / Consulting / Agency

Voice substitutes for product. Buyers cannot test the service before purchase; voice is the proof of competence and culture-fit.

Default attribute set:

• Demonstrably expert without preaching (Sage with Caregiver substrate)
• Specific case-study language; never abstract "transformation" prose
• Warm but never matey

Watch for:

• Voice of partners matters. Agencies are bought as a partner-by-partner decision; the firm voice and partner voices must be coherent, not identical.
• "We help organisations transform their operations through data-driven insights" — ban this sentence shape across all consulting clients. It is the single most common cliché in the category.

Pitfall: humour rarely translates outside specific category niches (creative agencies, DTC strategists). Irreverence reads as unprofessional in management consulting, accounting, or law.

---

B2C / D2C Consumer Brands

Reference brands: Oatly, Liquid Death, Innocent, Wendy's, Patagonia.

Voice is a brand asset and a moat. Distinctive voices have built billion-dollar D2C trajectories. But distinctiveness must be earned, not borrowed. Per Do Words Good: _"customers are able to spot brands that are riffing on other brands' voices."_

Default attribute set:

• Boldly distinctive (willing to alienate non-buyers)
• One archetype lean (rare to span more than one + adjacent)
• High-frequency lexicon (catchphrases, named concepts)

Watch for:

• Wackaging (Nick Asbury, 2011) — the "chatty, matey" tone Innocent invented in 2000 and dozens of brands copied. A copied voice is a discount voice. Lucie Bright (former Innocent copywriter), The Guardian, 2014: _"To be honest, we were mucking about when we started... We've always talked to everyone in the same way we talk to our friends, but with fewer swear words."_
• Be willing to alienate. Liquid Death founder Mike Cessario explicitly designed against polished bottled-water norms. _"We're just a funny beverage company who hates corporate marketing as much as you do."_

Pitfall: the "we want to sound like Oatly" brief. The strategy underneath the style is the differentiator, not the surface style. Borrowed voices age fast as audiences detect the imitation.

---

Non-profit / NGO

Reference brand: charity: water.

Voice carries the moral case. Scott Harrison (charity: water founder) built the voice around hope rather than guilt: _"I saw charities using shame and guilt to manipulate people into giving."_

Default attribute set:

• Hope-led, not guilt-led
• Concrete protagonist over statistics
• Specific scene, specific name, specific number — abstraction kills donor empathy

Watch for:

• Voice differs across donor segments. Major-donor letters vs broad-appeal email vs grassroots social are three tones of one voice. Document each.
• Storytelling is the foundation. Tyler Riewer (charity: water content lead, former ad creative) brought ad-craft storytelling to the org.

Pitfall: mission drift in language. A 30-year-old org's voice has often calcified into jargon ("capacity building", "stakeholder engagement", "theory of change"). Audit annually and rewrite using the words actual beneficiaries use.

---

Volunteering / Community Organisations

Voice must be permissive. People sign up to give time; a hectoring tone kills the funnel.

Default attribute set:

• Inviting, not prescriptive
• Reader's identity foregrounded ("you, who care about X"), not the org's mission
• Light on rules, strong on principles — chapter leaders in many cities cannot be policed centrally

Watch for:

• Equip chapter leaders with a one-page voice card and a banned-phrases glossary, not a 30-page guide.
• Founder voice often becomes the only voice — growth then requires a painful transition. Document the voice as detached from the founder early.

Pitfall: mission-statement language ("we believe that every person deserves...") in volunteer-facing copy. Volunteers are not donors; they want to know what they'll _do_, not what the org _believes_.

---

Industrial / Manufacturing

Reference brands: Caterpillar, 3M.

Reliability is a voice attribute. Voice supports perceived safety, precision, and certainty. Industrial buyers distrust marketing prose.

Default attribute set:

• Plainspoken expertise
• Reliability-as-tone — short sentences, named specifications, no hedging
• Specifications and case studies as primary brand surfaces

Watch for:

• Most industrial brand voice work fails by treating spec sheets as off-limits. They are the brand. The font, sentence structure, and unit conventions of a spec sheet shape buyer perception more than the website does.
• Resist "B2B = boring" defaults. Caterpillar and 3M have invested in distinctive voice with real returns.

Pitfall: corporate-anonymous tone. The phrase "industry-leading solutions" is the death rattle of industrial brand voice.

---

Product-led Companies

Reference brands: Duolingo, Linear, Notion, Slack.

Product UI is the largest single voice surface. Empty states, onboarding strings, error messages, button labels, confirmation dialogs see more impressions than any blog post.

Default attribute set:

• Functional first in-product; flourish moves to marketing
• Coordinated UX writing and brand writing (same owner or tight pairing)
• Mascot-led humour only where it doesn't bleed into transactional flows (Duolingo gets this right; most imitators don't)

Watch for:

• Modulation between marketing and in-product is the hardest call. Marketing can be playful; in-product must be functional first. NN/g's error-message walkthrough is the canonical teaching case.
• Slack's "Copy Principles" (Andrea Drugay) explicitly unify marketing, product, and partner-agency writing under one set of attributes.

Pitfall: mascot-led humour leaking into payment failures, account suspensions, security warnings. Wrong register in transactional copy destroys trust.

---

Personal Branding / Solopreneurs / Creators

Voice and person are the same thing. No translation layer between brand and self.

Default attribute set:

• Captured phrases, not just attributes — solopreneur voices are recognisable by micro-tics (sentence openings, signature questions, sign-offs)
• One archetype lean
• Stable over time — re-platform without re-voicing

Watch for:

• Codify your voice for future delegation. When you hire your first ghostwriter or VA, you need a one-pager describing how you sound. Run NN/g 4-dimensions on yourself.
• Chasing platform-native norms (LinkedIn-isms, TikTok cadence) until the voice becomes platform-shaped rather than person-shaped is the most common failure mode.

Pitfall: the voice becomes a parody of itself when the creator outsources writing without first capturing the voice in a guide. Ghostwriters default to platform clichés.

---

Public Sector / Government

Reference brand: GOV.UK.

Clarity is a civic duty, not a stylistic choice. GOV.UK's house position is that plain English is mandatory.

Sarah Richards (former GDS Head of Content Design), 2014: _"if we use a lot of adjectives and describing words, people will think it's spin... so we find that if we just write very plainly, very clearly, very directly, everybody understands."_

Default attribute set:

• Plain English (reading age of 9 is the inclusion principle, not infantilising)
• Politically neutral but never anonymous
• Specific actions, named services, no jargon, no Latin abbreviations
• Sentence length: average 15-20 words (Attorney General's Office UK Writing Style Guide, February 2015)

Watch for:

• Avoid contractions that confuse non-native readers. GOV.UK avoids negative contractions (don't, can't) for this reason.
• Banned-word list is the gold standard: replace "agenda" with "plan"; "deliver" with "make/create/provide"; "leverage" (unless financial) with "influence/use".
• Tone shifts by service (DVLA vs HMRC vs NHS) but voice principles do not.

Pitfall: writing for ministers rather than citizens. Build a review test that catches it (e.g. "would the reader at the bus stop understand this sentence?").

---

Categories requiring deep research

The categories below have voice norms shaped by audience, regulation, and platform that override generic guidance. Trigger Phase 2 (Research) before defining voice for any of them.

Research brief template (passed to the Phase 2 sub-agent):

Research current tone-of-voice norms for <category> brands in <market>. Cover: 1) typical voice attributes for the category; 2) common pitfalls and how audiences react to off-tone copy; 3) 2-3 reference brands with publicly observable voice patterns (cite primary sources); 4) regulatory, cultural, or platform constraints on voice. Report in under 700 words with sources cited inline.

Channel Adaptations

How tone modulates per channel. Voice never changes — only tone modulates. Each channel has hard constraints (platform-imposed limits and format), tonal shifts (how the brand's voice expresses itself within those constraints), and prohibited registers.

Use these as the input to the tone modulation matrix in TONE.md, and as the per-channel do/don't lists.

---

Table of contents

• LinkedIn
• Twitter / X
• Email (newsletter or transactional)
• Blog / Long-form essay
• In-product UI
• TikTok
• Instagram
• YouTube
• Podcast
• Press release
• Support / Customer service
• Sales decks
• Recruiting
• Disclaimer

Platform hard constraints change frequently. Verify limits against the platform's official help docs at the time of writing; the figures below are approximate guides, not authoritative.

---

LinkedIn

Hard constraints:

• Post body: ~3,000 characters (~500 words). Truncates at ~210 characters in feed with "see more".
• First 1-3 lines must hook (above the fold on mobile).
• Native paragraphs, no markdown rendering; line breaks shape readability.
• Supports unicode bold (better for copy-paste than markdown bold).
• Articles (long-form): ~125,000 characters.
• Comments: 1,250 characters.

Tonal shift from voice baseline:

• Dampen irreverence — LinkedIn's audience skews professional and risk-averse compared to Twitter/X.
• Lengthen sentences slightly; the medium tolerates more structure.
• Lead with results, not philosophy. ABT logic (And… But… Therefore…) outperforms narrative essays here.

Do:

• Reveal 80% in the hook (the result), keep 20% (the how) for the body — tension drives the scroll.
• Cite specific numbers, dates, costs.
• Format mobile-first: ≤2 visual lines per paragraph on phone.

Don't:

• Open with a rhetorical question — signals low confidence.
• Use "incontournable", "supercharge", "next-generation", "leverage" — LinkedIn buzzword fatigue is at saturation.
• Stack emojis or ternary structures (rule of three).

Cross-reference: for full LinkedIn ghostwriting workflow, see the samber/cc-skills@linkedin-ghostwriting skill.

---

Twitter / X

Hard constraints:

• Single tweet: 280 characters (free) or up to 25,000 (Premium / X Premium+).
• Thread: chain tweets; first tweet is the hook, must work standalone.
• No markdown; line breaks count as characters.
• Quote tweets, replies, polls, media attachments.

Tonal shift:

• Sharpen and shorten. Twitter rewards punch over polish.
• Irreverence and wit travel further here than on LinkedIn — within the brand's archetype tolerance.
• Trade nuance for brevity; reserve nuance for thread structure.

Do:

• Lead with the verb when possible — "Shipped X." beats "We have shipped X today."
• Use threads for narrative; reserve single tweets for one-shot insight.
• Time-stamp claims when freshness matters ("As of this week...").

Don't:

• Quote-tweet your own promotional copy with a fake-naive caption — reads as inauthentic.
• Use vague hooks ("Hot take 🔥"). Specificity beats teasing.
• Abuse 🧵 as a hook — readers have learned to scroll past.

---

Email (newsletter or transactional)

Hard constraints:

• Subject line: 30-50 characters before truncation on mobile (Gmail iOS: ~33 chars; Apple Mail iPhone: ~35-41).
• Preheader: 40-100 characters (Gmail desktop: ~110; mobile: ~30-50).
• From-name matters more than from-address for opens — a real human's name often outperforms a brand name.
• HTML and plain-text alternatives both rendered; some clients block remote images by default.

Tonal shift:

• Newsletter: voice baseline + structure (header, sections, single CTA).
• Transactional: functional first; flourish moves to onboarding/lifecycle.
• Active voice; lead with action; one primary CTA.

Do:

• Match the from-name to the relationship (founder for nurture, support@brand.com for transactional).
• Mirror the subject line in the preheader (don't repeat it).
• Test on mobile — most opens happen there.

Don't:

• Tease in the subject line, deliver in the body (clickbait subjects tank long-term open rates).
• Use marketing language in transactional copy (payment confirmations, password resets) — wrong register destroys trust.
• Hide the CTA below a wall of text.

Cross-reference: for Substack-specific workflows, see the samber/cc-skills@substack-ghostwriting skill.

---

Blog / Long-form essay

Hard constraints:

• No platform-imposed length cap; reader attention is the cap.
• Markdown supported (most CMS); H2/H3 structure affects SEO and scanability.
• Above-the-fold copy carries most of the engagement load.

Tonal shift:

• Voice baseline + room to breathe. Long-form tolerates more nuance, longer sentences, more digression.
• Lead with the thesis; don't bury the lede.

Do:

• Use H2/H3 to make scanning useful — readers skim before they commit.
• One idea per paragraph (most paragraphs ≤4 sentences).
• Quote sources inline with attribution.

Don't:

• Open with a dictionary definition (cliché, reads as filler).
• Use "in conclusion" — write a conclusion that doesn't need announcing.
• Pad with framing throat-clearing ("Let me start by saying...").

Cross-reference: for technical articles specifically, see the samber/cc-skills@technical-article-writer skill.

---

In-product UI

Hard constraints:

• Button labels: ~24 characters (most design systems).
• Empty states, error messages, confirmation dialogs: each has its own micro-format.
• Accessibility: WCAG 2.1 AA requires plain language; screen readers parse alt text and aria-labels.
• Localisation: every string lands in a translation memory; idioms and contractions don't travel.

Tonal shift:

• Functional first. Flourish is permitted in empty states and onboarding; transactional flows (payment, security, account suspension) strip all flourish.
• Use the user's data when available ("Your card ending in 4242 was charged $12.00") instead of generic confirmations.

Do:

• Action-first verbs in button labels ("Save", "Continue", "Cancel" — not "Click here").
• Match the error message to the user's action ("Update your card to keep your subscription" — not "An error occurred").
• One idea per microcopy unit; one decision per dialog.

Don't:

• Use brand-voice mascot humour in payment-failure or security-warning flows.
• Use marketing language ("Awesome!", "Yay!") in transactional confirmations.
• Use idioms that won't translate ("hit the ground running", "easy as pie").

Reference: Nielsen Norman Group's error-message walkthrough is the canonical teaching case.

---

TikTok

Hard constraints:

• Video length: up to 10 minutes; sweet spot 15-60 seconds.
• Caption: ~2,200 characters; first ~80 characters visible before truncation.
• Native-feel beats produced-feel.
• Trend cycles measured in days.

Tonal shift:

• Cadence-led. Voice expresses through rhythm of cuts, on-screen text, and audio choice — not only through written caption.
• Irreverence travels furthest here. Even traditionally polished brands (Duolingo, Ryanair) lean Jester on this channel.
• Native register: trending sounds, lo-fi production, direct address.

Do:

• Capture attention in the first 2 seconds.
• On-screen text reinforces (doesn't repeat) audio.
• Reply to comments with video (TikTok rewards engagement loops).

Don't:

• Repurpose 16:9 horizontal video without re-cropping — reads as inauthentic.
• Force corporate tone onto trending audio — audiences detect it instantly.
• Over-edit. Native = trust on TikTok.

---

Instagram

Hard constraints:

• Feed caption: 2,200 characters; first ~125 visible before "more".
• Stories: 24h ephemeral, multiple formats (poll, quiz, question sticker, link sticker).
• Reels: video up to 90s (some accounts 15 min); algorithm prefers vertical, original audio.
• Hashtags: 30 max in caption, 3-5 typically optimal.

Tonal shift:

• Visual-first; copy supports the image, not vice versa.
• Stories tolerate higher-frequency, lower-polish voice (closer to founder voice for personal brands).
• Reels follow TikTok-like cadence but skew slightly more polished and demographically older.

Do:

• Hook in the first line of the caption (above the "more" cutoff).
• Match copy register to image — branded studio shots take polished copy; behind-the-scenes Stories take casual copy.
• Use Story polls and questions to generate UGC and feedback loops.

Don't:

• Use a single voice across feed + Stories + Reels — each surface tolerates a different register.
• Cross-post Twitter copy verbatim — Instagram audience skews different.

---

YouTube

Hard constraints:

• Title: 100 characters; truncated to ~70 in suggestions.
• Description: 5,000 characters; first 100-150 visible above fold.
• Tags: 500 characters total.
• Chapters: enabled when description contains timestamps.

Tonal shift:

• Long-form video (10+ min): voice baseline + presenter style. Scripts tolerate more nuance and digression.
• Shorts (<60s): TikTok-like cadence.
• Title and thumbnail carry click-through; description carries SEO + retention nudges.

Do:

• Hook in the first 15 seconds (retention drops fastest there).
• Use chapters for searchability.
• Match title to thumbnail message — mismatch tanks click-through.

Don't:

• Clickbait titles that don't pay off in the first 60 seconds — algorithm penalises drop-off.
• Use marketing-deck language in the script — viewers detect "ad mode" instantly.

---

Podcast

Hard constraints:

• Episode length: 20-90 min typical; algorithm-agnostic medium.
• Show notes: long-form, SEO-relevant, hosted on podcast platforms.
• Audio-only (most episodes) — no visual scaffolding.

Tonal shift:

• Conversational voice baseline; scripts that sound written fail.
• Hosts carry voice. Show voice = host voice + production conventions (intro/outro music, segment names, ad reads).

Do:

• Write the script for ear, not for eye. Read aloud before recording.
• Segment names and recurring phrases become voice markers over time.
• Ad reads in host voice outperform polished agency reads.

Don't:

• Read marketing copy on-air verbatim — register mismatch is jarring.
• Over-edit out conversational pauses — they're part of trust-building.

---

Press release

Hard constraints:

• Headline: ~70 characters for wire-service truncation.
• Boilerplate at the end is fixed.
• Embargo language is conventional ("Under embargo until X").
• AP / Reuters style preferred by wire services and journalists.

Tonal shift:

• Most formal register the brand uses externally. Voice baseline + journalist-readable conventions.
• Lead with the news, not the opinion.
• Quotes carry the editorial voice; surrounding prose carries the factual voice.

Do:

• Inverted pyramid: most important fact first.
• Attribute quotes to a named person with title.
• Include contact details and embargo language explicitly.

Don't:

• Use marketing superlatives in the lede ("the most innovative... ever") — journalists strip them.
• Open with a quote.
• Write the headline last — write it first to discipline the lede.

Cross-reference: see the samber/cc-skills@press-release-writer skill.

---

Support / Customer service

Tonal shift:

• Voice baseline + empathy + outcome-orientation.
• Acknowledge the issue, name the next step, close the loop.

Do:

• Start with the customer's named issue: "Your invoice failed because..."
• Offer a concrete next step in every reply.
• Match formality to the customer's own message.

Don't:

• Open with "Sorry for the inconvenience" — saturated phrase, reads as scripted.
• Use legal-team passive constructions ("It has been determined that...").
• Forward without context — the next agent inherits the customer's frustration if the handoff is silent.

---

Sales decks

Tonal shift:

• Voice baseline + structured argument. Slides tolerate less nuance than long-form.
• Lead each slide with a claim, support with evidence underneath.

Do:

• One claim per slide, in the title.
• Concrete numbers in the title where possible ("Reduced churn 34% in Q3" beats "Better retention").
• Speaker notes carry the nuance; slides carry the headline.

Don't:

• Write slides as paragraphs.
• Use clip-art-grade icons or stock photos that contradict the brand visual register (this is a DESIGN.md concern, but voice and visuals are read together).

---

Recruiting

Tonal shift:

• Voice baseline + candidate-empathy + concrete role specificity.
• Job descriptions are content, not HR forms — write them with the same care as marketing copy.

Do:

• Lead with what the candidate will _do_, not the company's mission.
• Be specific about the team, the manager's style, and the work.
• State compensation ranges where legally required and culturally expected.

Don't:

• "Rockstar", "ninja", "wizard" — these are signals to senior candidates that the role is overstuffed and the culture is immature.
• Bury the disqualifying requirements at the bottom — candidates resent finding them after writing a cover letter.
• Use boilerplate inclusion statements without backing them up in the application process.

---

Disclaimer

Platform character limits, supported formatting, and algorithm behaviours change frequently. Treat the numbers above as approximate guides — verify against each platform's current help documentation before committing to a specific limit in TONE.md.

Trend cycles on TikTok, Instagram, and Twitter/X are measured in days or weeks. The tonal-shift guidance is more stable than the platform mechanics, but voice norms on social channels do drift over time. Audit channel-specific sections of TONE.md quarterly.

Discovery Questionnaire

The exhaustive 80+ question bank. Group questions into AskUserQuestion batches (max 4 per call). The SKILL.md Phase 1 lists the minimum-viable subset; this file is the deep bank for high-stakes brands, multi-market work, or category audits.

Rule of thumb: most failures in tone-of-voice work trace back to a rushed Phase 1. The discipline rewards depth here.

---

Table of contents

• A. Brand foundations
• B. Audience
• C. Positioning and competitive landscape
• D. Business goals
• E. Personality and culture
• F. Content scope
• G. Constraints
• H. Existing assets
• I. Stakeholders and governance
• J. Edge cases and emotional registers
• Batching guidance

---

A. Brand foundations

If SOUL.md exists, most of this section is already answered — extract and confirm rather than re-asking.

1. What is the founding story in 90 seconds? Who started it, when, why? 2. What was the original problem the brand was created to solve? 3. What is the mission statement (current, in use, not the one in the deck)? 4. What is the vision (10-year horizon)? 5. What are the values, and which are aspirational versus lived? 6. What does the brand believe that the category does not? 7. What is the single sentence the CEO would want a stranger to remember? 8. What has the brand previously claimed about itself that turned out to be wrong? 9. Are there cultural artefacts (rituals, internal slang, founder phrases) that should be preserved in the voice? 10. What is off-limits? (Topics, opinions, public stances.)

B. Audience

1. Who are the top 3 personas, named, with a one-paragraph bio each? 2. What are the JTBDs (jobs to be done) each persona hires the brand for? 3. What words do they use for the problem? For the solution? (Mine sales calls, support tickets, reviews.) 4. Which subreddits, Slack groups, LinkedIn communities, newsletters do they live in? 5. What other brands do they buy that you admire and want to share voice DNA with? 6. What other brands do they buy that you do not want to be confused with? 7. What is the emotional state when they encounter your content? (Anxious, bored, curious, urgent, defensive.) 8. What does success feel like for them after engaging with your content? 9. What is their reading level, literacy, and language fluency? 10. Are there sub-personas with materially different linguistic norms (developer vs PM; donor vs volunteer; junior buyer vs C-suite)?

C. Positioning and competitive landscape

1. Who are the 5 direct competitors? 2. Who are 3 adjacent brands you are sometimes compared with? 3. How do they sound on the NN/g 4 dimensions? (Plot them.) 4. Which of their voices do you respect, and why? 5. Which voices in your category are commodities (everyone sounds the same)? 6. What is your category truth that no competitor will say out loud? 7. What is the one sentence about your brand that, if a competitor wrote it, would make you angry? 8. If you disappeared tomorrow, what would the category lose? 9. Where does your voice need to differ from category default to be recognisable? 10. Where does your voice need to conform to category default to be credible?

D. Business goals

1. What content goals does ToV need to serve in the next 12 months? (Demand, activation, retention, recruiting, fundraising, policy change.) 2. What is the largest content investment area? (Where small voice gains compound the most.) 3. What does a "good" piece of content look like from a business perspective? 4. What KPIs will be used to test whether voice work is working? 5. What is the budget and headcount for content production? 6. Who owns content P&L?

E. Personality and culture

1. If the brand were a person at a dinner party, who would they be? (Job, age range, hobbies, what they read, who they admire.) 2. Which 3 employees most embody the desired voice today? 3. Which public figures would you point to as voice references (not to imitate, but to triangulate)? 4. How does the founder sound, and how much of that should survive in the brand voice? 5. What is the internal culture (Slack-channel tone, all-hands tone)? Should that show externally? 6. What jokes does the team make about the brand or category? Any worth surfacing? 7. What does the team find embarrassing about how the brand currently sounds?

F. Content scope

1. Which channels are in scope? (Web, blog, email, social per platform, in-product, support, sales decks, press, legal, recruiting, internal.) 2. Which formats? (Long-form essay, listicle, video script, podcast, newsletter, push notification, ad copy.) 3. Frequency per channel? 4. Who owns each channel? Who writes? Who reviews? Who publishes? 5. Which third parties create content in your name? (Agencies, freelancers, influencers, ambassadors.) 6. Where does AI-generated content enter the workflow today? 7. What is the lifespan of content? (Evergreen, campaign-bound, ephemeral.)

G. Constraints

1. Legal constraints: claims, disclaimers, regulated-industry language. 2. Regulatory constraints: GDPR, HIPAA, FDA, SEC, FCA, ASA, etc. 3. Accessibility requirements: WCAG, plain-language laws (US Plain Writing Act, etc.). 4. Cultural and religious sensitivities per market. 5. Linguistic constraints: which languages, which regional variants (en-US vs en-GB, pt-BR vs pt-PT). 6. Trademark constraints: how the brand and product names must be written. 7. Existing contractual commitments to specific naming conventions.

H. Existing assets

1. Inventory of current content by channel, owner, performance. 2. Top 10 best-performing pieces in the last 12 months. What do they share? 3. Bottom 10. What do they share? 4. Any existing style guide, brand book, messaging framework? 5. Customer-facing language audit: what do reviews, support tickets, and social mentions say back to you? 6. Are there pre-existing taboos or banned-word lists?

I. Stakeholders and governance

1. Who is the final decision-maker on voice? 2. Who is the day-to-day owner? 3. Who reviews content at what threshold? 4. What is the escalation path for sensitive copy (crisis, legal grey area, public criticism)? 5. Which teams write but currently do not consult any guide? (Sales, recruiting, HR, support, legal.) 6. What is the budget for training and tooling? 7. Who signs off on changes to the voice itself?

J. Edge cases and emotional registers

Each of these answers becomes a row in the tone modulation matrix. Skipping any of them means the matrix has gaps that downstream writers will silently fill with category default.

1. How should we sound during a crisis? (Outage, recall, lawsuit, public criticism.) 2. How should we sound when responding to complaints (justified, unjustified, viral)? 3. How should we sound on sensitive societal topics? Should we engage at all, and which ones? 4. How should humour work? Where is it banned? 5. What does an apology sound like in our voice? 6. What does celebration sound like (a win, an anniversary, a customer milestone)? 7. How do we acknowledge a mistake without destroying trust? 8. How do we say "no" to a customer? 9. How do we deliver bad news (price increases, sunsetting a feature, layoffs)? 10. How do we sound when a customer or community member uses language we disagree with?

---

Batching guidance

Use these batches when running AskUserQuestion (max 4 questions per call). The SKILL.md describes the minimum-viable subset; use these expanded batches for high-stakes brands.

Use AskUserQuestion only for questions where you genuinely cannot infer the answer. If SOUL.md already answers section A, skip batch 1. If the user has shared their existing brand book, skip the questions it covers.

For free-text answers (e.g. references, anti-references, persona descriptions), prefer asking the user to provide them in a single message rather than forcing them through a 4-choice picker.

Lexicon & Mechanics

The vocabulary and grammar layer of voice. Banned-word lists do more work than power-word lists — prevention is cheaper than prescription. A writer encountering "we don't use 'leverage' (unless in the financial sense)" never reaches for it again. A power-words list, by contrast, rarely gets consulted in the moment of drafting.

---

Table of contents

• Lexicon
• Preferred terms
• Banned terms
• Power words
• Jargon policy
• Naming conventions
• Mechanics
• Person
• Contractions
• Oxford comma
• Sentence length
• Active vs passive
• Reading age
• Capitalisation
• Emoji policy
• Punctuation tics
• Numerals, dates, currencies
• Inclusive language
• Localisation

---

Lexicon

Preferred terms

Named concepts and nouns the brand uses consistently. Each entry: term, meaning, what NOT to confuse it with.

Common decision: what to call the customer.

Pick one. Use it everywhere. Modulating the customer-noun across channels confuses the brand-customer relationship.

Banned terms

The single highest-leverage section of TONE.md.

Pattern (GOV.UK style): for each banned term, name the replacement and the reason.

LLM-generated copy tells:

These phrases appear disproportionately in AI-generated content. Ban them to keep brand copy distinguishable:

• "It's worth noting that..."
• "In today's fast-paced world..."
• "In conclusion..."
• "Navigate the complex landscape of..."
• "Unleash the power of..."
• "Delve into..."
• "Embark on a journey..."
• "Game-changing"
• "Cutting-edge"
• "Robust solution"

Power words

10-30 words that consistently appear in best-performing copy for the brand. Mine your top 10 best-performing pieces; the words that repeat across them are your power words.

Examples (D2C): free, new, you, save, easy, now, today, last, finally, real.

Examples (B2B): because, specifically, exactly, never, fastest, reduced, prevented, real, evidence, named.

Power words are less load-bearing than banned-word lists. Treat them as descriptive (what we already do well) rather than prescriptive (what we must do).

Jargon policy

Naming conventions

---

Mechanics

Person

Contractions

Rule worth knowing: GOV.UK avoids negative contractions (don't, can't, won't) because they confuse non-native readers — the apostrophe is easy to miss, flipping the meaning. Allow positive contractions ("we're", "you're") even when banning negative ones.

Oxford comma

Pick one; document it; enforce in CI/linter if you have one.

Sentence length

General-public default: average 15-20 words per sentence (Attorney General's Office UK Writing Style Guide, February 2015: _"An average of 15 to 20 words per sentence is ideal."_).

Expert audiences: can tolerate 20-30 words on average, but never as a sustained pattern.

Public sector (GOV.UK): 15-20 word average; reading age of 9.

Mix lengths. A page of uniformly 18-word sentences reads as monotonous. Vary: 3-5 word punches, then medium (12-18), then occasionally longer (20-25) for nuance. Variation is the rhythm signature of a confident voice.

Active vs passive

Default: active. Subject acts on object — "We charge the card" beats "The card is charged."

Passive allowed for:

• Time-sensitive softening (Fenton/Kiefer Lee, _Nicely Said_): _"In certain situations it helps you sound softer... It's especially useful for time-sensitive messages like payment confirmations and error messages."_ "Your card was charged" softens "We charged your card" in a payment confirmation.
• Process descriptions where the actor is unimportant ("The data is encrypted in transit").
• Avoiding blame when blame would damage the relationship ("An error occurred" beats "You did something wrong").

Reading age

GOV.UK target: reading age of 9. Sarah Richards (GDS): _"Writing for a reading age of 9 isn't the same as writing for a 9-year-old. It's about making sure your content is quick to scan, easy to read and simple to understand."_

Tools to measure: Flesch-Kincaid grade level, Hemingway Editor. Aim for grade 6-8 (US) / reading age 9 (UK) for general-public content.

Higher reading age justified for: dev docs, expert reference material, specifications.

Capitalisation

Sentence case is friendlier and more scannable on mobile. Title case can feel formal or dated.

Emoji policy

For each channel in scope, decide:

• Allowed? (Y/N)
• Which emoji set? (Most brands restrict to a subset to avoid mismatched register.)
• Frequency cap? (1 per post? 1 per paragraph? None?)
• Specific emojis banned? (e.g. faces with unclear connotation: 😅, 😬)

In-product UI generally bans emoji except in marketing-adjacent surfaces (onboarding, empty states).

Punctuation tics

Each of these is a stylistic signature. Decide explicitly:

• Ellipses (`...`): sparingly; reads as trailing thought, can feel passive-aggressive
• Em-dashes (`—`): parenthetical asides; allowed but cap at 1-2 per paragraph
• Exclamation marks (`!`): cap at 1 per post; never in transactional copy
• Parentheses (`()`): allowed for asides; ensure the sentence still works if the parenthetical is removed
• Hyphens in compound modifiers (`well-known`): consistent application

Numerals, dates, currencies

---

Inclusive language

Base references:

• Karen Yin, _Conscious Style Guide_ — the canonical living resource
• APA Inclusive Language Guidelines
• For UK English: Citizens Advice plain-language guide

Yin's principle: _"To write consciously is to be aware of and compassionate toward the actual people who determine not only the success but also the meaning of any work."_

Decisions to document per locale:

• Gendered language: default to avoidance; use singular "they" when gender is unknown. Mailchimp: _"Avoid gendered language and use the singular 'they.' When writing about a person, use their preferred pronouns; if you don't know those, just use their name."_
• Disability / ability: person-first ("person with diabetes") vs identity-first ("autistic person"). Defer to the community's expressed preference where it exists.
• Race, ethnicity, nationality: specific over abstract; capitalisation per community preference (e.g. "Black" capitalised in US convention).
• Age: avoid "elderly"; prefer "older adult" or specific age range.
• Class, education, geography: avoid sneering shorthand ("flyover country", "uneducated voters", "low-skilled workers").
• Neurodiversity: community-led terminology; defer to current consensus.

---

Localisation

Per market, document:

• Which idioms, sports metaphors, pop-culture references, and humour styles do not travel.
• Regional variants: en-US vs en-GB vs en-AU; pt-BR vs pt-PT; es-ES vs es-MX vs es-AR; zh-CN vs zh-TW.
• Date, time, address, phone, currency conventions.
• Honorifics and formal-vs-informal address (tu/vous in French; tú/usted in Spanish; T-V distinctions widely).
• Politeness register: Japanese teineigo / sonkeigo / kenjogo; Korean nida-form vs haeyo-form.

Transcreation, not translation. Surface-translating English brand voice into French or Japanese produces tin-eared copy that fails the local market. Hire writers in each market; do not rely on translation memory alone for surface tone. Lokalise and similar platforms operationalise transcreation workflows.

Reference Brands

Brands whose published voice systems reward forensic study. Read primary sources where possible — most of these brands have public style guides or interviews with their content leads.

Use these to triangulate, not to imitate. "We want to sound like Oatly / Mailchimp / Duolingo" briefs produce derivative voices that audiences increasingly detect — the strategy under the style is what makes a voice work.

---

Mailchimp

Category: B2B SaaS / Marketing automation Voice principle: Clear, useful, friendly, appropriate Why study: The original publicly-published voice/tone guide. Defines the voice/tone distinction more cleanly than anyone.

The canonical voice/tone quote:

_"You have the same voice all the time, but your tone changes. You might use one tone when you're out to dinner with your closest friends, and a different tone when you're in a meeting with your boss... Our voice doesn't change much from day to day, but our tone changes all the time."_

On clarity over entertainment:

_"Mailchimp's tone is usually informal, but it's always more important to be clear than entertaining."_

On Oxford comma: _"We like serial commas, and we cannot lie."_

On inclusive language:

_"We write and build apps with a person-first perspective... Avoid gendered language and use the singular 'they.' When writing about a person, use their preferred pronouns; if you don't know those, just use their name."_

Source: styleguide.mailchimp.com (Content Style Guide).

---

Slack

Category: B2B SaaS / Communication Voice attributes: Confident (never cocky), Witty (but never silly), Conversational (but always appropriate and respectful), Intelligent (but not arrogant), Friendly (but not ingratiating), Helpful (never overbearing), Clear, concise and human. Why study: The "X but never Y" pattern is the single most useful documentation device in the discipline. Slack also coordinates voice across marketing, product, and partner agencies via internal "Copy Principles" (Andrea Drugay, Group Manager of Copy).

Brand-as-person: Slack explicitly aims to be _"an ideal colleague."_

Source: Slack design site + Andrea Drugay's published interviews.

---

GOV.UK

Category: Public sector Voice principle: Plain English; reading age of 9; politically neutral Why study: The benchmark for plain-language voice. Reading age of 9 is not infantilising — it is the inclusion principle.

The canonical quote on spin and clarity (Sarah Richards, former GDS Head of Content Design, March 14, 2014):

_"if we use a lot of adjectives and describing words, people will think it's spin... so we find that if we just write very plainly, very clearly, very directly, everybody understands."_

On scan reading: _"Our research shows us that people only read about 20-28% of a page."_

Sentence length: Attorney General's Office Writing Style Guide (February 2015): _"An average of 15 to 20 words per sentence is ideal."_

Banned-word list (gold standard):

• "agenda" → use "plan"
• "deliver" → use "make", "create", "provide"
• "leverage" (unless financial) → use "influence" or "use"
• Many more — published openly at design.gov.uk/style-guide

---

Stripe

Category: B2B SaaS / Developer tools / Fintech Voice principle: Documentation as content design; sharp marketing, simple docs Why study: Stripe set the bar for developer-facing documentation. Marcus Nunez's team publishes sample docs (e.g. "the life of a Stripe charge") to teach the standard rather than just provide templates.

Pattern: simple sentences, reader-centric tone, never patronising; never marketing-laden in docs but sharper in marketing. The two voices are coordinated.

---

Linear

Category: B2B SaaS / Project management Voice attribute: Speed-as-voice, competence-as-voice Why study: Product-led voice where UI strings carry the brand. Empty states, button labels, error messages — all treated as voice surfaces.

Pattern: UI strings written with the same care as marketing copy. The product itself is the strongest voice argument.

---

Notion

Category: B2B SaaS / Productivity Voice principle: Founder voice carried into a brand voice; B2C-style marketing for a B2B product Why study: Camille Ricketts (former marketing head) credits CEO Ivan Zhao's voice as the source of brand tone.

Quote (paraphrased from interviews):

_"Brand isn't just about what you say or even how you say it. It's also about what you do."_

The Notion lesson: voice is downstream of product worldview, not a marketing veneer on top.

---

Duolingo

Category: B2C / Edtech Voice attributes: Playful, approachable, encouraging Mascot-led: Owl-as-voice (Duo) carried social channels to break out Why study: Demonstrates how a brand voice can stretch dramatically across surfaces — encouraging in-product, "unhinged" on TikTok — without breaking. Zaria Parvez (social manager) describes the TikTok voice internally as _"unhinged"_ and _"out of pocket."_

The lesson on register modulation: the in-product voice is the baseline. The TikTok voice is the maximum-irreverence channel adaptation. Both share the same underlying attributes (playful, approachable); only the dial is turned.

Pitfall to copy carefully: mascot-led humour bleeding into transactional flows. Duolingo gets this right (payment failures use functional tone, not Duo's voice). Many imitators do not.

---

Oatly

Category: B2C / D2C / Food & beverage Archetype: Outlaw / Jester Voice attributes: Casual, funny, irreverent, enthusiastic (CFIE in CopyStyleGuide.com's profiling) Why study: Built a billion-dollar trajectory on distinctive voice. Mostly handled internally rather than fed to AI tools, per Creative Director Malia Killings.

Pitfall: "wackaging" descendants. The voice works because it's tied to Oatly's strategy (anti-corporate-dairy positioning). Brands borrowing the surface style without the strategy underneath produce derivative voice that audiences detect.

---

Wendy's

Category: B2C / QSR Voice signature: Twitter-native roasts; Amy Brown era (2012-March 2017) Why study: The inflection point for distinctive corporate Twitter. Jan 2, 2017: _"You don't have to bring them into this just because you forgot refrigerators existed for a second there."_ National Roast Day launched 2018.

HubSpot 2023 case study: _"In 2022, National Roast Day generated over 130 million impressions in a matter of hours and made Wendy's an instant trending topic on Twitter."_

The lesson: voice as moat works in B2C when audience expects entertainment from the brand. Try it in B2B SaaS and it backfires.

---

Liquid Death

Category: B2C / D2C / Beverage Archetype: Outlaw Voice attribute: "Murder Your Thirst"; voice consistent across packaging, customer service, email Why study: Andy Pearson (VP Creative) on the voice running _"from product descriptions to email subject lines."_ Founder Mike Cessario explicitly designed against the polished norms of bottled water.

Quote:

_"We're just a funny beverage company who hates corporate marketing as much as you do."_

The lesson: be willing to alienate. The voice repels non-target audiences, which is part of how it works. Trying to be appealing to everyone produces voice that appeals to no one.

---

Patagonia

Category: B2C / Outdoor apparel Archetype: Hero / Explorer Voice principle: Values-led; activism explicit Why study: The "Don't Buy This Jacket" campaign (NYT, Black Friday 2011) is the canonical anti-marketing voice moment.

The lesson: voice that goes against immediate commercial interest signals authentic values. Works only when the values are real — performance values-language without operational backing reads as cynicism.

---

Innocent Drinks

Category: B2C / Beverage / UK Archetype: Jester / Everyman Voice signature: Chatty, matey, founder-led Why study: Defined a UK category in 1999 (founders Richard Reed, Adam Balon, Jon Wright). Later widely copied to dilution — Nick Asbury named the resulting trend "wackaging" in 2011 (ABC Copywriting blog).

Head of Brand Dan Germain:

_"Ultimately, it is about talking to people. The worst thing for Innocent would be to get too grown-up and stop chatting with the people who buy drinks."_

Former copywriter Lucie Bright (The Guardian, 2014):

_"To be honest, we were mucking about when we started... We've always talked to everyone in the same way we talk to our friends, but with fewer swear words."_

The lesson: a voice that defined a category gets copied to commodity. Original is differentiating; the third brand to sound this way is generic.

---

charity: water

Category: Non-profit / NGO Archetype: Hero / Caregiver Voice principle: Hope over guilt; concrete protagonist over statistics Why study: Scott Harrison built the voice as an explicit alternative to category default.

Harrison on category default:

_"As I looked at the charity sector, I didn't see any amazing brands... I saw anemic brands. I saw charities using shame and guilt to manipulate people into giving."_

Content lead Tyler Riewer (former ad creative, eight years before joining charity: water) makes storytelling the foundation.

The lesson: voice carries the moral case. The category default ("urgent need, donate now") had become noise. Hope-led voice broke through because it differentiated from the category, not because hope is universally better than urgency.

---

How to use these references

1. Triangulate, don't imitate. Pick 2-3 reference brands with voice attributes adjacent to your brand's. Identify what they share and what's unique to each. 2. Read primary sources. Most of these brands have published guides, interviews with content leads, or behind-the-scenes process posts. The secondary write-ups simplify; the primary sources reveal the actual decisions. 3. Note the strategy under the style. Each of these voices works because it expresses a specific strategic position. Borrow the _method_ of tying voice to strategy; don't borrow the surface style. 4. Watch the failure modes. Wackaging is the cautionary tale: a voice that worked when Innocent did it became a cliché when 50 other brands copied it. Distinctive voice depreciates quickly when shared.

Voice Attributes

How to define, document, and validate the 3-5 voice attributes that anchor TONE.md. Voice is the brand's fixed personality — it does not change across channels. Tone modulates per situation; voice does not.

---

Table of contents

• The "X but never Y" pattern
• Attribute documentation template
• NN/g 4 dimensions
• Jung's 12 archetypes
• BrandSort / message architecture
• Brand-as-person exercise
• Frameworks compared

---

The "X but never Y" pattern

The single most useful documentation device in the discipline. Slack's published voice:

_"Confident (never cocky), Witty (but never silly), Conversational (but always appropriate and respectful), Intelligent (but not arrogant), Friendly (but not ingratiating), Helpful (never overbearing), Clear, concise and human."_

Why it works: every attribute carries its own anti-attribute. Writers learn what to do AND what to avoid in the same memorable phrase. Without the "never Y" half, "confident" reads as a synonym for "cocky" to half the team.

For each attribute, produce three "we are X but we are not Y" statements during the BrandSort exercise. The triplet forces specificity — by the third statement, generic phrasing collapses and concrete language emerges.

Bad (generic, won't survive use):

• "We are friendly."
• "We are professional."

Good (specific, falsifiable):

• "Confident, never cocky."
• "Plainspoken, not patronising."
• "Warm, but never saccharine."

---

Attribute documentation template

Each attribute in TONE.md follows this exact structure (downstream skills parse on it):

### N. <Attribute name>

**Definition:** <one line — what this attribute means in practice>

**Sounds like:** "<one example sentence in the voice>"

**Does not sound like:** "<one anti-example — generic, off-brand, or category default>"

**Do:**
- <concrete behaviour 1>
- <concrete behaviour 2>
- <concrete behaviour 3>

**Don't:**
- <concrete behaviour 1>
- <concrete behaviour 2>
- <concrete behaviour 3>

**Anti-example from past content:** "<actual past sentence that violated this attribute>"

Why anti-examples from past content matter: abstract anti-examples ("don't be too formal") fail to bite. Anti-examples sourced from the brand's own historical copy bite because writers recognise them — "oh, I wrote that exact sentence last quarter." This is where the guide stops being theatre and starts shaping behaviour.

---

Worked attribute example

Drawn from a B2B SaaS context — adapt for the brand at hand.

### 1. Plainspoken

**Definition:** We say what we mean in the fewest credible words.

**Sounds like:** "Your invoice failed. Update your card to keep your subscription."

**Does not sound like:** "Unfortunately, we were unable to process the transaction at this time."

**Do:**
- Use active voice — name who does what.
- Lead with the most important fact.
- Use one idea per sentence.

**Don't:**
- Hedge ("perhaps", "potentially").
- Use insurance-language passives ("an error has occurred").
- Stack qualifiers ("we are currently unable to potentially process...").

**Anti-example from past content:** "It has come to our attention that, in some cases, certain customers may have experienced..."

---

NN/g 4 dimensions

Nielsen Norman Group's measurement layer for voice. Four spectra, each a 3-point scale (or 5-point if you want finer granularity):

Use case: plot the brand on all four dimensions. Plot 3-5 competitors on the same chart. Look for clustering — if you're at the same coordinates as two competitors, your voice is not differentiating. Move at least one dimension.

Anti-pattern: clustering near the midpoint on all four dimensions. Per Quasa.io's diagnosis: _"teams defaulting to mid-range scores (4-7), result in bland, forgettable voices."_ If three dimensions are at midpoint, the brand has no voice.

Empirically grounded: NN/g research shows tone of voice measurably affects perceived friendliness, trustworthiness, and desirability. On average, 52% of the variability in desirability scores is explained by trustworthiness — meaning trust-building tone dimensions (matter-of-fact, plainspoken, respectful) pay more than charm.

---

Jung's 12 archetypes

Strategic personality alignment with audience psychology. Each archetype maps to a core motivation — pick the one whose audience your buyer most resembles.

Use caveats:

1. Most brands lead with one archetype and borrow from one adjacent (e.g. Apple = Magician primary, Creator secondary). 2. Archetype is a positioning shortcut, not a voice solution. Two "Hero" brands can sound completely different (Nike vs FedEx). Don't stop at the archetype label — derive concrete voice attributes from it. 3. Brands that lean too hard on archetype end up in cosplay. The archetype should inform decisions, not become a costume.

---

BrandSort / message architecture

Margot Bloomstein's exercise (_Content Strategy at Work_, Morgan Kaufmann, 2012) for surfacing brand attributes through a card-sort:

1. Use a deck of ~100 brand-attribute adjective cards (or generate a list). 2. Sort each card into one of three piles: who we are now, who we'd like to be, who we are not. 3. Narrow the "who we'd like to be" pile to ~9 cards. 4. Group those 9 into three clusters of three; pick the strongest from each cluster. 5. The top 3 are the message architecture and the seed for voice attributes.

Run with a cross-functional group of 6-10 people: leadership, marketing, sales, customer success, product writing, support, HR. The disagreements between groups are diagnostic — surface them explicitly rather than averaging them out.

Reconcile by running BrandSort three times:

1. Once with leadership 2. Once with the writing team 3. Once with sales / customer success (who hear how customers actually talk)

Differences between the three runs reveal where the brand's self-image diverges from its operational reality. Force a decision in the room rather than letting it persist as ambiguity in the guide.

---

Brand-as-person exercise

A forcing function for concreteness. If the brand were a person at a dinner party:

• Job, age range, hobbies?
• What do they read?
• Who do they admire?
• How do they introduce themselves?
• What do they refuse to talk about?
• What do they wear?

For B2B brands, substitute "ideal colleague" or "celebrity spokesperson" — Slack explicitly aims to be _"an ideal colleague"_. This frames voice as a relationship the brand has with the reader, not an abstract personality.

Output: one paragraph describing this person. Then check: does the brand actually sound like this person in its current content? If not, that gap is what TONE.md is for.

---

Frameworks compared

No single framework is sufficient. Combine: BrandSort for stakeholder alignment, NN/g 4 dimensions for differentiation, "X but never Y" for documentation, archetype as positioning shortcut. Each addresses a different stage of the work.