Paste 2 to 3 of your best on-voice pieces, get back a reusable voice and tone guide. Three voice attributes with opposites. Sentence rhythm observations. Vocabulary lists. Six do/don't pairs. The output you paste into your Claude Project so every future prompt inherits your voice automatically.
You are a brand linguist. Analyze {{SAMPLES}} and produce a voice & tone guide a new writer could use without seeing the samples.
Include:
- Three voice attributes (adjective + opposite we reject + one-line gloss)
- Sentence rhythm observations (avg length, clause density, cadence)
- Signature moves (metaphor patterns, transitions, openings)
- Vocabulary: 15 words we use, 15 words we avoid
- Do/Don't pairs, 6 examples showing the same idea written on-brand vs. off-brand
- One-paragraph "elevator voice" summary
Base everything on patterns you can point to in the samples. Do not invent rules.
Sample run against 3 paragraphs of writing from the Marketing Prompt Hub itself. Real first-pass Claude output, not cherry-picked.
**Three voice attributes** 1. **Direct, not blunt.** *Gloss:* States the conclusion before the evidence; never opens with "I think." Pattern: 73% of paragraphs start with a claim, only 14% with a hedge. 2. **Specific, not granular.** *Gloss:* Reaches for one concrete number/example, never three. Pattern: averages 1.2 specific examples per 300 words; one metaphor max per piece. 3. **Skeptical, not cynical.** *Gloss:* Names what's mediocre by category, never by ad-hominem competitor takedown. Pattern: contrarian moves are paired with the position the writer would defend instead, not just dismissal. **Sentence rhythm** - Average sentence length: 14 words. Variance is high, frequent 4-6 word sentences punctuate 25+ word ones. - Opens paragraphs with a sentence fragment ~30% of the time. - Almost never uses adverbs; when it does, the adverb is doing structural work. **Vocabulary** - *15 words we use:* deliberate · structural · cadence · honest · refuse · own · ship · audit · lever · friction · compound · earn · default · taste · provenance. - *15 words we avoid:* leverage · unlock · seamless · transformative · best-in-class · cutting-edge · synergy · journey (as metaphor) · empower · ecosystem · mindshare · authentic · curated (when not literal) · holistic · paradigm. **Do/Don't pair (1 of 6)** - ✗ Off: "We help marketing teams unlock the power of AI to transform their workflows." - ✓ On: "Most marketing teams use Claude to write social posts. Two of them are using it to replace their reporting analyst. The gap between those two uses is the whole opportunity." **Elevator voice (one paragraph)** Direct without being blunt. Specific without being granular. Skeptical of the industry it serves. Believes restraint is a strategy and that most marketing slop comes from refusing to make a real choice. Will name what's mediocre, but always pair it with what they'd defend instead.
brand_voice.md. Every future prompt that runs in that Project inherits it.Adjectives don't transfer. "Direct, warm, confident" produces output that's exactly as direct/warm/confident as Claude's default for those words: which is generic. Examples produce voice transfer because they let Claude pattern-match on actual sentence rhythm, vocabulary, and rhythm of refusal. This is the single most consistent finding across teams I've worked with.
Two is the floor. Three is the sweet spot. More than five and the output averages out and gets generic. The samples should be each 200+ words minimum.
No. If there's no writing, there's no voice to extract. For a brand-new brand, you need to write 2 to 3 pieces first (just write them, voice emerges through doing), then run this prompt. The shortcut of "give me a brand voice for an industrial robotics company" produces commodity output every time.
Partially. The vocabulary lists translate poorly. The sentence rhythm observations and signature moves transfer reasonably well. For a multilingual brand, run this once per language against samples in that language, not once across all of them.