Brands sound the same after adopting AI not because AI homogenizes writing, but because most teams prompt with generic instructions (“write a LinkedIn post about X”) instead of feeding the model their brand’s specific opinions, vocabulary, and things it refuses to say. The fix isn’t switching tools — it’s building a prompt system that encodes what makes your brand disagree with the category consensus, not just how it sounds.
The Complaint Everyone’s Making
Scroll any marketing feed right now and you’ll see the same take: AI is flattening every brand voice into the same confident, slightly upbeat, faintly corporate register. It’s become the go-to explanation for why timelines feel duller than they did three years ago.
It’s also the wrong diagnosis.
What’s Actually Happening
AI models don’t have a default “boring” setting. They have a default average setting — they predict the most statistically likely next sentence unless told otherwise. When a marketer types “write a caption about our new product launch,” the model has nothing to differentiate on. So it reaches for the industry’s most common patterns, because that’s what “launch caption” means in the absence of any other signal.
The output isn’t generic because the AI lacks personality. It’s generic because the prompt didn’t supply one.
The Prompt Most Teams Are Using (And Why It Fails)
Most brand voice prompts look like this:
- “Our tone is friendly but professional.”
- “We’re bold and innovative.”
- “Keep it conversational.”
These are adjectives, not instructions. Every competitor in your category has written nearly identical guidelines — “friendly but professional” describes roughly 80% of B2B brands on earth. Feeding that into a model gives it nothing to differentiate against, so it defaults to the category average. You haven’t told it who you are. You’ve told it who everyone is.
What Actually Produces a Distinct Voice
Distinct brand voice isn’t a tone description — it’s a set of opinions and refusals. The brands that still sound like themselves after heavy AI use share one habit: they’ve told the model what they believe that competitors don’t, and what they’ll never say even if it’s on-brand-adjacent.
A stronger prompt structure looks like:
- A stated opinion. Not “we’re innovative” — something like “we believe most productivity advice makes people worse at their jobs.” Give the model a stance to write from.
- A rejection list. Specific words, phrases, or framings the brand never uses — “supercharge,” “game-changer,” rhetorical questions as hooks. This does more to create distinctiveness than any tone description, because it removes the model’s safest, most generic defaults.
- A real example of disagreement. One past piece of content where the brand said something contrary to category norms. Models pattern-match far better from an example than from an adjective.
The Actual Test
Before publishing AI-assisted content, ask one question: could a direct competitor have posted this unchanged? If yes, the prompt gave the model nothing brand-specific to work with — the tool isn’t at fault.
The Takeaway
AI didn’t remove brand personality from marketing. It removed the requirement to have one, because generic instructions now produce fluent, publishable, forgettable copy by default. Brands that still sound distinct aren’t using better AI — they’re feeding it sharper opinions and clearer refusals than “friendly but professional” ever gave anyone.
