There’s a problem hiding in plain sight across the advertising industry right now. It’s not about data privacy. It’s not about cookie deprecation. And it’s not about rising CPMs. It’s this: Generative AI is making your brand sound like a polite robot. And polite robots don’t sell anything.
The Conventional Wisdom Is Half Right
Most conversations about AI in ad copy focus on three things: speed, volume, and cost. And look, those are real benefits. Writing fifty headlines in thirty seconds is genuinely impressive. Testing more variations without bleeding your budget matters. But here’s what nobody wants to admit: all of that efficiency is coming at a cost you can’t see on your dashboards.
The real risk isn’t bad copy. It’s flawlessly correct copy that’s strategically useless. Let me show you what I mean.
The Silent Killer: Median Copy
Large Language Models are trained on the entire internet. They know what’s safe. They know what’s agreeable. They know what’s median. When you prompt ChatGPT to write a LinkedIn ad for a B2B product, here’s what you get:
“Unlock your team’s full potential with our innovative platform. Streamline workflows and drive measurable results.”
Grammatically perfect? Yes. Factually accurate? Probably. Universally inoffensive? Absolutely. Worth anyone’s attention? No.
Now compare that to the ads that actually work:
- “Think Different.” Fragmented. Bad grammar. Unforgettable.
- “Just Do It.” Imperative. Barely a sentence. Iconic.
- “Because You’re Worth It.” Slightly presumptuous. Emotionally charged. Culturally defining.
Generative AI will never write “Think Different.” It will write “We encourage you to explore innovative approaches.” The first is a cultural artifact. The second is a data point. This is the paradox: the more you optimize for algorithmic approval, the less you resonate with actual humans.
The Hidden Cost Nobody’s Talking About
Here’s what keeps me up at night. Every brand is now using the same LLMs, trained on the same data, prompted with similar briefs. The result? Everyone is converging toward a single, bland, “best-practice” tone.
When everyone sounds like a competent consultant, nobody stands out. Open your Facebook feed right now. How many ads can you identify by brand before you see the logo? How many have a voice so distinct that the text alone tells you who it is?
If your answer is “not many,” you’re experiencing the homogenization problem in real time. The algorithm rewards clarity and directness. But the human brain craves novelty and surprise. By optimizing for the algorithm, we’re training our audiences to ignore us.
Precision Kills Persuasion
Here’s the strategic truth most agencies won’t tell you: great advertising doesn’t sound like a committee. It sounds like a person. And people are messy. They break rules. They use syntax that doesn’t parse. They say things that are slightly weird.
That weirdness? That’s your edge. When you strip it out in favor of AI-perfected precision, you’re not improving your copy. You’re neutering it.
The most effective ads throughout history all share one thing: they sound like someone wrote them for another human being, not for an algorithm.
How to Use AI Without Losing Your Voice
I’m not saying abandon Generative AI. That would be naive. But you need a different approach. Here’s the framework we use:
1. Prompt for Constraints, Not Copy
Stop asking the AI to “write a compelling headline.” That’s a recipe for mediocrity. Instead, give it constraints that force it out of its comfort zone:
- “Write this ad as if it were a note passed in a high school class.”
- “Write a headline that breaks a grammatical rule. Explain why.”
- “Generate ten headlines. Then swap the adjective in headline three with a verb. Tell me which version is more surprising.”
This forces the AI to produce tension, not just text.
2. Feed It Real Transcripts, Not Brand Guidelines
Most brand voice guides are useless for AI. “Friendly” and “professional” and “disruptive” are adjectives, not instructions. Instead, feed the AI transcripts of your actual customers talking to your sales team. Raw. Unfiltered. Ungrammatical. That language-with its hesitations, metaphors, and slang-is the only antidote to AI’s sterility.
3. Use AI for Problems, Not Solutions
Here’s the most important shift: don’t use Generative AI to write the final copy. Use it to uncover emotional triggers.
Ask the AI: “List five deep-seated anxieties a CEO has before a Q3 review.” It will produce a great list. Then you take that list and turn it into a single, sharp sentence using a specific noun or metaphor the AI would never choose. The AI handles the research. You handle the resonance.
The One-Second Test
Here’s a simple way to check if your AI-generated copy is working. Read it out loud. Does it sound like a person talking to another person? Or does it sound like a brochure written by an algorithm that has never felt a feeling?
Most AI copy passes the logic test but fails the feeling test. If your copy doesn’t make someone feel something in the first second, it doesn’t matter how much it optimizes for click-through rate. You’re winning a short-term metric at the expense of long-term brand equity.
What This Means for You
The brand that wins in the Generative AI era will not be the one that produces the most copy. It will be the one that uses AI to protect and amplify its unique voice against the tide of algorithmic mediocrity.
Your customers don’t want to talk to a polite robot. They want to talk to a brand that sounds like it gives a damn. Stop asking AI to write your ads. Start asking AI to find the friction in your customer’s journey. The best ad copy doesn’t fit the algorithm perfectly. It fits the human brain’s craving for a signal in a sea of noise.
Your job-as a strategist, a brand leader, or an agency partner-is to make sure that signal isn’t sanitized. The paradox of precision versus persuasion is real. Navigate it well, and you win. Ignore it, and you sound like everyone else.
