Why Generic AI Content Fails
"Write an Instagram post for a restaurant" produces generic mush. Anyone using ChatGPT recognizes it: fluent, but soulless. Not your client. Not their voice.
The problem isn't ChatGPT. The problem is you haven't taught it how your client sounds.
The 7 Steps to Get Brand Voice into AI
Step 1: Collect 10 Original Client Texts
Posts, emails, "about us" pages, FAQ texts. Not AI-generated — original work. The more diverse, the better the AI can write in all contexts later.
Step 2: Analysis Prompt
Analyze these 10 texts from [brand name]. Describe in detail: 1. Tone (formal/informal, warm/businesslike) 2. Vocabulary (simple/jargon, NL/English loanwords) 3. Sentence length (short and punchy / long and explanatory) 4. Emotional charge (rational / emotional / mix) 5. Typical sentence openings 6. Typical closing phrases (CTA style) 7. Personal pronouns (je / u / wij / ik) 8. Use of humor/irony Output as a structured profile with examples from the source texts.Step 3: Save the Profile
The output from step 2 is your brand voice document. Store it in a central system (Senly's brand kits per client, or Notion). Not in your head or inbox — the entire team must be able to access it.
Step 4: Create a Writing Test
Now ask AI to write a new post using the brand voice document as input:
Write an Instagram post about [topic]. Use this brand voice profile: [paste profile from step 2]. Length 100 words, with CTA.Compare the output with the original texts. Mismatch? Refine the profile (step 5).
Step 5: Iterative Refinement
The first output is rarely perfect. Ask AI:
What about this post doesn't sound like [brand name]? Improve and explain which adjustments you made.AI's own feedback is often surprisingly good. Add the improvements to the brand voice document.
Step 6: Document "Don'ts"
It's just as important to document what does NOT fit:
- Words the client never uses ("excellent", "amazing")
- Taboo topics (mentioning competitors, politics)
- Stylistic devices that don't fit (too many emojis, all caps for emphasis)
Step 7: Make it Reusable
Save the brand voice + don'ts as a reusable prompt library. Senly has built-in brand kits per client for this — automatically loaded with every caption generation so the team doesn't have to do prompt engineering anymore.
How to Know if You're Succeeding
Test: Have 3 colleagues guess blindly which posts were AI-generated and which were written by the client themselves. If they can't predict the difference — success.
Test 2: Send 5 AI posts to the client without context. Do you get approval without revisions? The brand voice is correct.
What You Save With This
A typical agency spends 30-45 minutes per client per post on briefing + writing + revisions. With a strong brand voice profile: 8-12 minutes. With 15 clients and 4 posts/month per client = 35-50 hours per month saved. That's half an FTE.
Ready to scale your agency?
Senly is the all-in-one platform for marketing agencies.
Start free trial →