Two marketers at competing brands, each using ChatGPT, ask the same question about positioning their respective brands. They get nearly identical answers. The answers are well-written. They're plausible. They're also indistinguishable from each other in any meaningful way. Then both marketers ship campaigns based on the answers. The campaigns look like each other. The category gets blander. Nobody quite figured out why.
This is happening at scale right now. Millions of marketers using the same tools, asking similar questions, getting similar outputs, then implementing those outputs. The aggregate effect is a flattening of marketing across every category that AI has touched. The work isn't getting worse, exactly. It's getting more similar.
THE MECHANISM
Large language models work by predicting the most statistically likely next word. Given any prompt, the output is shaped by what the model has seen most often after similar prompts in its training data. The training data is essentially the open internet. So the output is essentially the consensus of the open internet.
If your competitor uses the same model, they get the same consensus. The model isn't biased toward their brand or yours. It's biased toward the average answer. The average answer is what comes out, regardless of which prompt-engineer is asking the question.
Tweaking the prompt moves you around inside the consensus. It doesn't get you outside of it. You can ask for "a bolder version" or "a more contrarian take." The output gets a more bolder-shaped or more contrarian-shaped texture. The underlying generation is still pulling from the average. You're navigating a slightly different region of the same distribution.
WHY THIS MATTERS FOR DIFFERENTIATION
The whole job of a brand is to be different. Not different in some superficial way. Different in a way the customer notices and remembers and prefers. The traditional way brands got different was through a combination of original strategy, distinctive voice, and choices that competitors hadn't made.
If everyone is using the same tool for strategy, voice, and choice-making, the inputs to differentiation collapse. Everyone arrives at the same strategic position, the same voice, the same set of choices. The brand layer flattens. Customers are left choosing between options that look interchangeable, which means they choose on price.
This is already happening in some categories. Direct-to-consumer brands all use the same playbook because they all use the same AI for the playbook. SaaS landing pages have converged on a narrow set of structures because everyone is asking the same model how to write a landing page. The work is technically fine. It's also forgettable, because everyone else's work is the same.
THE WAY OUT
Use AI for the layers where averaging is fine. Operational work. Drafting. Quick research. The 80% of marketing work where the consensus answer is the right answer.
For the layers that require differentiation (strategy, positioning, voice, the creative call), don't use the same general-purpose tool everyone else is using. Use a tool with a different corpus, or use no AI at all and do the work yourself. The differentiation has to come from somewhere outside the consensus, or it doesn't come.
The brands that will pull ahead in the next few years are the ones that recognize this distinction. They'll use AI heavily for execution and almost not at all for strategy. The brands that use AI for everything will keep producing competent work that looks like everyone else's work. Competent and indistinguishable is the worst place a brand can be.