groundswell-brand-voice

name

groundswell-brand-voice

description

Keeps Groundswell Juice Co.'s brand voice cohesive across social media captions, marketing emails, web copy, menu descriptions, and customer service replies. Use this any time content is being drafted from scratch for Groundswell, or an existing draft needs to be checked/reviewed for tone, terminology, and brand-voice violations before it goes out — even if the user doesn't use the words "brand voice" explicitly (e.g. "does this email sound right," "can you check this post," "write a caption for...").

Groundswell Brand Voice

Handles two jobs: drafting new copy in Groundswell Juice Co.'s house style, and reviewing an existing draft for brand-voice violations. The guide in references/brand-voice-guide.md is the source of truth — if anything in this file and the guide ever conflict, the guide wins.

First, figure out which job this is:

• No draft yet (write a caption, email, blurb, menu line, reply) → "Drafting new copy" below.
• A draft already exists (pasted copy, or an existing post/email/reply that needs a check, cleanup, or opinion on whether it's on-brand) → "Reviewing a draft" below.

Both jobs are graded against the same standard, covered once here so it isn't duplicated: "Voice rules" and "Hard rules" below.

Voice rules that apply to every format

Pull the full detail from references/brand-voice-guide.md; the compressed version:

• Audience is informed, not casual — explain well, don't dumb down, don't write short punchy copy that reads like it doesn't trust the reader with information.
• Name things: the farm, the partner, the actual number. A claim without a specific behind it is treated as a form of dishonesty by this brand, not just weak writing. See the Preferred terms table in the guide (e.g. "farm partners" not "suppliers," "compostable in [X] cities" not "eco-friendly").
• Say what's unfinished. If something only partly works (not every city, not every taste preference), say so plainly rather than smoothing it over.
• Warm, not performative. Sounds like a neighbor who remembers your order, not a brand voice that's excited because excitement is the house style.
• Practical about health claims — explain the mechanism, don't call things "superfoods" or make claims that read as medical advice.
• Never self-congratulatory about baseline responsibility.

Hard rules — check every piece of copy against these

• Banned words, ever, in any format: guilt-free, detox, cleanse, elixir, indulge / indulgent, treat yourself, superfood, self-care (as a marketing hook), any girlboss/hustle-culture phrasing.
• No em dashes. Anywhere, in any copy. Use a period, comma, or parentheses instead.
• Exclamation points: max one per piece, and often zero. Never stacked.
• No vague sustainability/health claims left unsupported — "eco-friendly," "sustainably sourced," "clinically proven" (unless there's an actual source), "at our core," or similar phrases that sound good and verify nothing.

Re-read the copy once specifically hunting for these five things — don't rely on having avoided them while drafting, and don't rely on skimming when reviewing.

Drafting new copy

Before drafting: get the specifics

Groundswell's voice runs on verifiable detail, not tone. A draft that's warm and grounded but has no farm name, no number, no real mechanism behind a claim will read as off-brand no matter how well it mimics the sentence rhythm. This is the most common way a first draft fails.

If the input doesn't already contain enough concrete detail to support the copy, ask for it before writing — don't invent specifics (a farm name, a poundage figure, a percentage of cities covered) that weren't given. It's fine to write a slightly shorter piece with real specifics than a fuller one with invented ones. If a genuine detail truly isn't available, write around it honestly (see the "honest about tradeoffs" pattern in the guide) rather than filling the gap with a vague claim.

If the output format isn't stated, ask which one it is — the formats below have real structural differences, not just a length difference.

The formats

Social captions

Read references/social-examples.md. 3–4 sentences, no hashtags, no emoji. Open on the fact or the moment, include one concrete specific even at this length, close on a plain confident line rather than a stacked-exclamation CTA.

Emails

Read references/email-examples.md. Needs: an understated subject line (often framed as "why we did X" / "what X actually means," never a superlative), a plain "Hi," open, a first paragraph that states the news with no windup, middle paragraphs that show the reasoning with names/numbers/places, an acknowledged tradeoff if one exists, and a sign-off that matches the email's mood rather than a generic "Best,".

Web copy

No dedicated sample exists for this format, so lean on the guide's "Successful vs. Unsuccessful example" as the model: plain statement of what changed, the specific mechanism behind it, the honest current state including what isn't finished yet, no line that "says nothing verifiable." Match length and structure to what the page needs (a product blurb reads more like a caption; an About/story page reads more like an email body) but the underlying moves — specificity, acknowledged tradeoffs, no inflation — stay constant regardless of page type.

Menu descriptions

One sentence. This is the tightest format Groundswell uses, so every word has to earn its place — usually the ingredients plus what one of them actually does, using the "practical about health" pattern (what it does in the body, in plain terms) rather than an adjective ("refreshing," "delicious"). No banned words fit in one sentence anyway, but check regardless. Example shape, built from the Repair Red material in the guide: "Tart cherry, beet, ginger, and collagen — tart cherry and ginger both settle a hard workout, beet keeps blood flow moving." (Illustrative only; do not reuse Repair Red copy verbatim across outputs — write fresh copy for whatever product or topic is actually being asked for.)

Customer service replies

Read references/customer-service-examples.md for tone calibration — this isn't a fifth format with its own structural rules, it's the same voice addressed to one person instead of an audience. What shifts: acknowledge the specific situation before anything else, offer a concrete remedy (a credit, an amount, a next step) rather than a vague apology, and when the news is bad (a complaint, a "no"), be honest that a tradeoff is a tradeoff rather than either over-apologizing or getting defensive.

Drafting workflow

1. Identify the format (social / email / web copy / menu / customer service reply) — ask if it's not stated.
2. Check whether the input has enough real specifics to write from. If not, ask for them rather than inventing.
3. Draft using the relevant format pattern above and the matching reference file.
4. Run the Hard rules check against the draft.
5. Present the copy. If it's for email or web copy and a subject line / headline is expected, include one in the matching style.

Reviewing a draft

The goal is to hand back something the writer can act on quickly: what's off-brand, why, and what to say instead — not a silent full rewrite that hides how much changed or forces them to diff it themselves.

Review workflow

1. Identify the format the draft is in (or ask if genuinely ambiguous) so you're checking it against the right structural expectations — a social caption and an email are held to different shapes.
2. Read references/brand-voice-guide.md and, if the format has one, its matching example file (social/email/customer-service) so the comparison is grounded in real examples, not just the compressed Voice rules above.
3. Go through the draft against the Hard rules checklist first (banned words, em dashes, exclamation points, unsupported claims) — these are unambiguous and quick to flag.
4. Then check against the Voice rules — vaguer, more judgment-based: is there a real specific behind every claim, is an unfinished/imperfect part being smoothed over, does anything read as performative or self-congratulatory rather than grounded.
5. Report findings as a list: quote the offending line, name the rule it breaks, and give a specific fix (not just "make this more on-brand"). If the draft is clean, say so plainly rather than inventing nitpicks to seem thorough.
6. Only produce a full rewritten version if asked for one, or if the violations are pervasive enough that a line-by-line list would be less useful than seeing it fixed. Default to the flag-and-fix list.