Voice drift is quiet. Nobody decides to sound generic; it happens one rushed post at a time. This audit catches it in 15 minutes, with a score you can compare month to month, in the spirit of the Iteration pillar.
Run it monthly, or any time content starts feeling "fine" instead of "us." For a deeper monthly version with more dimensions, see the Voice Consistency Scorecard.
Step 1: Gather five pieces (3 minutes)
Pull five recent pieces of real, published content. Mix channels so drift between channels shows up:
- One social post (LinkedIn, X, or Instagram caption)
- One email (newsletter or sales)
- One web page or blog post
- One support or customer reply
- One wildcard (proposal, product description, bio, anything)
Paste them into one doc. Don't cherry-pick your best work. The audit only helps if the sample is honest.
Step 2: Run the ten checks (10 minutes)
Score each piece against all ten. Yes = 1 point, No = 0. Maximum score: 50.
1. The cover-the-logo test. Hide your name and logo. Could a reader still guess it's you? Yes/No
2. The competitor test. Could your closest competitor publish this exact text without anyone noticing? (Yes here scores 0. Distinctive scores the point.)
3. Banned-word scan. Search the piece for your banned words and adjectives. Zero hits = 1 point. (No banned list yet? Use this starter set: leverage, elevate, empower, unlock, seamless, game changer, delve, in today's fast-paced world.)
4. Read-aloud test. Read the first three sentences out loud. Do they sound like something a person on your team would actually say? Yes/No
5. One-breath opener. Is the first sentence under 20 words and does it say something specific? Yes/No
6. Concrete claims. Does every claim name something real (a deliverable, a number, a timeframe, an example)? A piece with even one "world-class" or "next-level" floater scores 0.
7. Consistent temperature. Does this piece sit at roughly the same casual/formal level as the other four? Yes/No
8. On-voice CTA. Does the call to action sound like you, or like a button generator? ("Learn more" scores 0. "See what a Voice Sprint includes" scores 1.)
9. AI tell scan. Any of these present: "In today's digital landscape," "It's important to note," rule-of-three sentence stacks, a closing paragraph that summarizes what you just read? Any hit scores 0.
10. The forward test. Would a happy customer forward or screenshot this? Gut answer, two seconds. Yes/No
Step 3: Score and act (2 minutes)
Add up all 50 possible points.
40 to 50: Tight. Your voice is working. Note which piece scored lowest and which check it failed; that's your one fix this month. Re-run the audit in 30 days to confirm it holds.
25 to 39: Drifting. You have a voice, but it shows up inconsistently. Look at your scores by check, not by piece. If check 7 (temperature) failed across channels, you need per-channel voice rules. If checks 3 and 9 failed, your AI prompts need your banned list and Voice Charter pasted in. Fix the two weakest checks before next month's audit.
Below 25: Generic. Most of this content could belong to anyone. Don't fix the pieces; fix the system upstream. Write your 25-word Voice Charter first, build your Tone & Mood Matrix second, then re-run this audit on your next five pieces. Expect the score to jump 10+ points once every piece starts from the same charter.
Track it
Keep a simple log so the trend is visible:
| Date | Total /50 | Weakest check | Fix committed |
|---|---|---|---|
| _______ | _______ | _______ | _______ |
| _______ | _______ | _______ | _______ |
| _______ | _______ | _______ | _______ |
Two rules for the log. First, one fix per month, actually done, beats five fixes planned. Second, if the same check fails three audits in a row, the problem is your process, not your writers: build the fix into the prompt or template everyone starts from.