AI Didn't Make Your Marketing Generic. It Exposed That It Always Was

Adobe reported a 57% engagement lift from an AI-selected creative treatment in its MAX campaigns. Around the same time, a 2024 study in Science Advances gave 293 writers generative AI ideas and found their stories were rated more creative, and grew measurably more similar to one another. Both results are real. Read side by side, they explain something uncomfortable about AI and marketing.

The researchers called their finding "individual creativity at the risk of losing collective novelty." That was fiction, not marketing, so hold the comparison loosely. But the mechanism transfers exactly. Every asset gets better. The body of work gets blander. When every team runs the same systems trained on the same dominant patterns, average quality rises while recognizable difference disappears. The result is a fluent, credible, interchangeable center.

The 57% measured the wrong thing. It is a real result, and it tells you which treatment earned more immediate engagement. It does not tell you whether the campaign built a more defensible position, improved how buyers understand your company, or produced an idea a competitor would struggle to copy. Platforms reward creativity that is easy to process. Select the most immediately responsive version enough times and you will optimize away the precision that made the idea worth running.

Production was never the moat. For years the work looked original partly because making it required time, budget, and specialized talent. Scarcity created the impression of value. That barrier has collapsed. What is left is the uncomfortable test: does your organization actually see something different about this market? In healthtech, nearly everyone can claim to improve efficiency and support better outcomes. Those claims are usually true and reveal almost nothing.

If you already run AI across your content: do these three things this week

1. Write the sentence a competitor would refuse to sign. One sentence on what you believe about this market that your closest rival could not put their name to. If you cannot write it, no model can write it for you. It will produce a more finished version of the same category claim.
2. Audit your last ten assets for flattening. Read them back to back. If the arguments are interchangeable and only the topics differ, your quality went up and your distinction went down. That is the tradeoff in the study, showing up in your own library.
3. Separate the optimization metric from the position metric. Engagement tells you what got processed easily. Track comprehension and credibility separately, or the short-term number will quietly govern strategy.

If you are still holding AI at arm's length: do these three instead

1. Stop treating restraint as strategy. Not using the tools does not make your positioning sharper. It just makes the same generic claims slower and more expensive to produce.
2. Feed it your sharpest claim, not your boilerplate. Give a model a real thesis and it will stress the logic, find the weak spots, and help the idea travel. Give it your about page and it will polish the sameness.
3. Put a human veto on the last 10%. The World Health Organization has warned that generative systems encourage automation bias, where users stop catching errors they would otherwise catch. In healthcare, a simplified claim is not a style problem. Name the person who owns accuracy and precision before anything ships.

The reframe

AI is not creating the sameness. It is removing the cost that used to hide it.

Judgment is the scarce capability now: deciding which ideas deserve investment, which parts of a position must survive optimization, and when a better metric is eroding long-term authority.

Every serious team will have these tools within a year. Having something specific to say will not become ordinary.

And if you don't want to do it alone, let us know.

‍