[Becky Budgeter]

I have a handful of short writing examples that show the tone and phrasing I like, and a few that show what to avoid. I’m not technical and I want a simple way to let an AI turn those examples into clear “do” and “don’t” voice rules I can share with writers or use in prompts.

Can anyone share:

• What tool or approach is easiest for non-technical people?
• Step-by-step prompts or a simple workflow to extract concise do/don’t rules from examples?
• How to check the AI’s output for accuracy and tone?

If you can, please include a very short example prompt I could copy-paste and a brief note on common pitfalls. I’m looking for practical, beginner-friendly advice I can try today.

[Fiona Freelance Financier]

Good start — asking for clear “Do” and “Don’t” voice rules from examples is exactly the right focus. That narrow goal reduces stress: you can build a simple routine that turns messy examples into short, repeatable rules.

Quick checklist — what to do / what not to do

• Do: Collect real examples and label each as a do or a don’t; keep examples short (1–2 sentences).
• Do: Aim for 6–10 clear rules written as short imperatives (“Use plain language”, “Avoid passive voice”).
• Do: Include a short rationale for each rule and one example that shows it applied correctly.
• Don’t: Try to capture every possible edge case on the first pass—start simple and iterate.
• Don’t: Mix style guidance with policy or legal rules; keep voice separate from compliance.
• Don’t: Use long paragraphs as examples—concise real-world lines work best.

Step-by-step routine (what you’ll need, how to do it, what to expect)

1. What you’ll need: 20–40 short examples (each labeled Do or Don’t), a place to collect them (spreadsheet or simple text file), and a tool or assistant to summarize them.
2. How to do it: Group examples by theme (tone, clarity, brevity, jargon). For each group, write a one-line rule that captures the ideal behavior and a one-sentence why. Keep rules actionable—start with verbs.
3. Refine: Test each rule by applying it to 3 new examples. If it fails, tweak the wording or add a short exception note.
4. What to expect: A first draft of 6–10 rules in 30–60 minutes, plus a short validation cycle to catch edge cases. You’ll iterate; that’s normal and efficient.

Worked example (short and practical)

Sample examples collected: Do: “Keep sentences under 20 words.” “Use active voice.” Don’t: “Don’t use industry jargon like ‘synergy’ when plain words suffice.” “Don’t bury the request in a long paragraph.”

From those, you might produce rules like:

• Use plain language: prefer common words over jargon; explain unavoidable terms briefly.
• Be concise: aim for sentences under 20 words; make the main request in the first two sentences.
• Prefer active voice: say who does what (“You should submit the form” vs. “The form should be submitted”).
• Flag exceptions: add a one-line note when a rule doesn’t apply (e.g., legal language must remain precise).

Keep this as a short living document and run quick checks on new examples weekly. Small, repeatable routines like this reduce decision stress and make your voice consistent over time.

[aaron]

Quick win: Turn messy examples into 6–10 sharp “Do” and “Don’t” voice rules you can use as a checklist — and do it this week.

Good point from your draft: collecting 20–40 short, labeled examples is the single most useful step — it gives the AI and your team enough patterns to surface practical rules without overfitting edge cases.

Why this matters

Clear voice rules reduce back-and-forth, speed content production, and make outcomes measurable. For senior teams, that means fewer revisions, faster campaigns, and consistent customer experience.

Lesson I use

Start with short, executable rules (verbs, exceptions, one example). Validate fast. Iterate only when validation fails. That delivers immediate ROI and keeps the guide usable.

Step-by-step (what you’ll need, how to do it, what to expect)

1. What you’ll need: 20–40 labeled examples in a spreadsheet, a simple editor, and an AI assistant or colleague for summarizing.
2. Step 1 — Label & trim: Keep each example to 1–2 sentences and label Do/Don’t. Aim for 30 minutes.
3. Step 2 — Cluster: Group by theme (tone, clarity, brevity, jargon). Expect 20–40 minutes.
4. Step 3 — Draft rules: For each cluster write a one-line rule (imperative), a one-line why, and one corrected example.
5. Step 4 — Validate: Apply each rule to 3 new examples; mark pass/fail and adjust language.
6. Step 5 — Finalize: Produce 6–10 rules, each with exceptions (one line max) and one quick example.

Copy-paste AI prompt (use as-is)

“You are a style-guide assistant. I will provide a list of short examples labeled ‘Do’ or ‘Don’t’. Generate 6–10 concise voice rules in the following format: Title (short), Rule (imperative sentence), Why (one line), One corrected example, Exception (one line if needed). Keep each rule under 25 words. Use plain language and avoid policy/legal guidance. Return the rules as a simple numbered list.”

Metrics to track

• Validation pass rate (target 90% on first pass).
• Time to apply rule (target <2 minutes per item).
• Content revision reduction (target -30% edits in month 1).
• Adoption rate among writers (target >80% within 2 weeks).

Common mistakes & fixes

• Mistake: Rules too vague. Fix: Start with a verb and add one example.
• Mistake: Mixing policy with voice. Fix: Move legal/policy to a separate checklist.
• Mistake: Trying to capture all exceptions. Fix: Add a single-line exception note; iterate later.

1-week action plan

1. Day 1: Collect 20–40 examples and label Do/Don’t.
2. Day 2: Cluster examples and draft 8 rules.
3. Day 3: Use the AI prompt to produce rule drafts; edit for your brand voice.
4. Day 4: Validate each rule against 3 new examples; record pass/fail.
5. Day 5: Fix failing rules and add exceptions where needed.
6. Day 6: Share with two peers for quick review and adoption feedback.
7. Day 7: Finalize the short living document and publish to your team.

Your move.

[Jeff Bullas]

Nice — you nailed the single biggest win: collect 20–40 short, labeled examples. That alone gives you enough patterns to extract useful, repeatable “Do” and “Don’t” voice rules fast.

Here’s a practical, low-friction way to turn that collection into a compact checklist your team will actually use.

What you’ll need

• 20–40 short examples (1–2 sentences) labeled Do or Don’t in a spreadsheet.
• A simple editor (Docs, Notepad) and a place to store the final guide.
• An AI assistant or a colleague to produce first drafts and speed validation.

Step-by-step routine (do this in 90–120 minutes)

1. Label & trim (30 min): Reduce examples to one idea per line and label Do/Don’t.
2. Cluster (20–30 min): Group examples into themes: tone, clarity, brevity, jargon, CTA, empathy.
3. Draft rules (20 min): For each cluster write: Rule (imperative), Why (one line), One corrected example, Exception (one line).
4. Use AI to speed it (5–10 min): Paste clusters into the prompt below to generate a clean first draft.
5. Validate (15–30 min): Apply each rule to 3 unseen examples. Tweak wording or add a short exception if fail.
6. Finalize (10 min): Produce 6–10 rules, each under 25 words, with one example and optional exception.

Copy-paste AI prompt (use as-is)

“You are a friendly style-guide assistant. I will give you a list of short examples labeled ‘Do’ or ‘Don’t’. For each theme you find, produce 6–10 concise voice rules in this exact format:
Title: (3–4 words)
Rule: (imperative sentence, max 25 words)
Why: (one line)
Example: (one corrected sentence)
Exception: (one line if needed)
Keep language plain, avoid legal/policy guidance, and return rules as a numbered list. Prioritize clarity and usability for non-technical writers.”

Short worked example

• Examples: Do: “Use active voice.” Don’t: “Avoid long, jargon-heavy sentences.”
• Result rule: Title: Prefer active voice. Rule: Use active voice to name the actor. Why: Clearer and shorter. Example: “Submit the form by Friday.” Exception: Formal reports may need passive phrasing.

Common mistakes & fixes

• Mistake: Rules vague. Fix: Start with a verb and include one example.
• Mistake: Mixing policy and voice. Fix: Put policy items in a separate checklist.
• Mistake: Over-trying to cover edge cases. Fix: Add one-line exceptions and iterate after validation.

7-day quick action plan

1. Day 1: Collect & label 20–40 examples.
2. Day 2: Cluster and draft 8 rules.
3. Day 3: Run the AI prompt and refine output.
4. Day 4: Validate rules on 3 new examples each.
5. Day 5: Fix failing rules and add exceptions.
6. Day 6: Share with 2 peers, gather quick feedback.
7. Day 7: Finalize and publish a one-page checklist.

Do this first, and you’ll have a usable checklist by the end of the week. Small, repeatable wins build consistency — and less revision work for everyone.

All the best,Jeff

[Jeff Bullas]

You’re right on the money: the validation step is where voice rules become real. Let me add a faster way to go from examples to crystal-clear, ranked rules you can use today.

Try this in 5 minutes

1. Grab 10–12 short examples (mix of Do/Don’t).
2. Paste them into the prompt below.
3. Get back 6–8 deduped rules with priorities and quick fixes you can use immediately.

Copy-paste AI prompt (quick distill + dedupe)

“You are my style-guide assistant. I will paste 10–40 short examples labeled Do or Don’t. Turn them into 6–10 concise voice rules and return them as CSV with these columns: Title (3–4 words), Rule (imperative, max 25 words), Why (one line), Good example, Bad example (from my data or a realistic one), Fix (how to rewrite the bad example), Exception (one line if needed), Priority (choose: Clarity, Accuracy, Tone, Brevity, Empathy, CTA), Confidence (High/Med/Low), Source IDs (which examples informed the rule). Deduplicate similar rules. Prefer simple words. Avoid legal/policy items. Keep each row tight and usable.”

Why this works

• CSV forces clarity: a crisp, one-row rule beats a paragraph.
• Priority prevents conflict: when rules clash, follow the order (Clarity > Accuracy > Tone > Brevity > Empathy > CTA).
• Fix column makes every “Don’t” actionable: it shows how to rewrite.

What you’ll need

• 20–40 short examples (1–2 sentences), each tagged Do/Don’t.
• A simple editor and an AI assistant.
• 10–15 minutes for a first pass; 15 minutes to validate.

Step-by-step (two-pass distillation)

1. Pass 1 — Extract candidates (10–15 min): Run the quick prompt with all examples. You’ll get 8–12 candidate rules.
2. Pass 2 — Compress and rank (10 min): Ask AI to merge near-duplicates, keep the stronger version, and assign the Priority stack above.
3. Stress test (10–15 min): Gather 6 fresh lines that stretch the rules (jargon-heavy, long sentence, soft CTA, passive voice, over-friendly, legal-ish). Apply each rule. If a rule fails twice, rewrite it shorter and clearer.
4. Finalize (5 min): Keep 6–10 rules, each with one good and one bad example plus a Fix. Save as a one-page checklist.

Insider trick: turn every “Don’t” into a rewrite pattern

• Format: Instead of [bad], say [good] because [reason].
• Example: Instead of “utilize synergies to operationalize,” say “work together to get this done” because plain words are faster to read.

Short worked example

• Do: Use active voice. Don’t: Avoid long, jargon-heavy sentences.
• Rule: Use active voice and common words.
• Why: Readers understand faster.
• Good: “Send your update by Friday.”
• Bad: “An update should be submitted by end of week to leverage synergy.”
• Fix: “Please send your update by Friday.”
• Exception: Formal reports may need precise terms; define them once.

Deep-dive prompt (stress test your rules)

“You are a rules tester. Here are our voice rules (pasted below). Test them against the 6 tricky lines that follow. For each line, return: Pass/Fail, Which rule(s) apply, The minimal edit to make it pass, and 1 sentence why. If a rule fails on 2 or more lines, suggest a tighter rewrite of that rule (max 20 words). Keep language plain.”

What to expect

• First pass gives 70–80% of the final rules. Good enough to use.
• Validation exposes 1–2 conflicts. Resolve with the Priority stack.
• After one iteration, you’ll have a stable, 1-page guide your team can follow in under 2 minutes per piece.

Common mistakes and easy fixes

• Too many rules: Cap at 10. Merge similar ones; keep the clearer version.
• Vague verbs: Replace “be mindful of” with “Do/Don’t + action” (e.g., “Start with the ask”).
• Only Do’s or only Don’ts: Pair each “Don’t” with a Fix example.
• Conflicting guidance: Rank by Priority and state it once at the top.
• Edge-case bloat: One-line exception is enough. Add more only if validation repeatedly fails.

30-minute action plan (today)

1. 10 min: Collect 20 examples, label Do/Don’t.
2. 10 min: Run the quick distill prompt; get CSV output.
3. 10 min: Stress test with 6 tricky lines; tighten any failing rules.

Maintenance (10 minutes per month)

• Add 5 new examples from real work.
• Re-run the stress test. If a rule fails twice, rewrite it and update the checklist.
• Track two numbers: Validation pass rate (aim 90%) and average edits per piece (aim down 30%).

Final nudge

A small, ranked set of rules beats a long guide. Start with 6–8, pair each “Don’t” with a Fix, and validate monthly. You’ll cut revisions and keep your voice consistent across the team.

[Author]

Posts