The Sameness Machine: How AI Quietly Converges Your Marketing on Everyone Else’s
Your marketing is better than it was a year ago. The emails are sharper. The landing pages are cleaner. Your one-person content function now ships like a team of four. Every individual piece has gone up in quality, and you can feel it.
So why does everything in your category suddenly read like it was written by the same person?
Because, in a sense, it was. And there’s now a peer-reviewed study that explains the mechanism — one every founder scaling marketing with AI should sit with for a minute, because it describes a trap you walk into by doing everything right.
The finding: better individually, samer collectively
In 2024, researchers Anil Doshi and Oliver Hauser published a study in Science Advances with a title that is the whole argument: “Generative AI enhances individual creativity but reduces the collective diversity of novel content.”
They had people write short stories, some with AI help, some without. The AI-assisted writers produced work that was rated more creative and better written — a real, measurable individual lift, biggest for the writers who’d been least confident on their own. That’s the promise, confirmed. AI genuinely makes the individual output better.
Then they measured the stories against each other. The AI-assisted stories were markedly more similar to one another than the unassisted ones. Each writer, pulling from the same model, drifted toward the same ideas, the same structures, the same turns. Every author climbed. The group flattened.
That is the exact shape of what’s happening to marketing right now. Each marketer, prompting the same handful of models, gets individually better and collectively more identical. The tool that lifts your floor also lifts everyone else’s floor, in the same direction, toward the same middle. Doshi and Hauser called it a possible “social dilemma”: what’s rational for each writer to use is, in aggregate, quietly erasing the variety that made any of them stand out.
Why this is a strategy problem, not a content problem
It’s tempting to file this under “AI writing sounds a bit bland” — a tone issue you fix with a better prompt. It isn’t. It’s a positioning problem wearing a content costume.
Differentiation is not a nice-to-have in marketing; it is the marketing. The entire job is to occupy a spot in the buyer’s head that your competitor doesn’t. When a machine is nudging every player in your category toward the statistical center of how your category talks, it is dissolving the one asset you were trying to build. You can be producing more, faster, cheaper — and moving backward on the only dimension that makes a buyer choose you over the next tab.
And it compounds. These models are increasingly trained on the internet’s output — which now includes last year’s AI marketing. The center of gravity pulls tighter each cycle. Left alone, “use AI for everything” is a slow convergence on the median voice of your industry. The median voice does not get discovery calls.
How to use the machine without being averaged by it
The answer is not to swear off the tools — the individual lift is real and refusing it just means shipping less while everyone else ships more. The answer is to be deliberate about what you feed it and where you overrule it. A working rule: use AI to accelerate, never to originate. Four ways to hold that line.
- Bring the raw material AI can’t have. A model can only recombine what’s already public. It has never sat in your customer’s chair, run your last pricing experiment, or watched a deal die for a reason that never made it into a case study. Your first-hand specifics — the number, the story, the thing you actually watched happen — are the one input that isn’t in the training data. Start there and let the tool polish. Start from a blank prompt and you’ll get the average, beautifully formatted.
- Fix the point of view before you draft. Decide the argument, the contrarian take, the exact position you’re staking out — then open the AI. The strategy has to originate with a human who’s willing to be disagreed with, because “willing to be disagreed with” is precisely the edge a consensus machine sands off. If your take is one an LLM would generate unprompted, it’s not a take, it’s the baseline.
- Diversify your prompts and your models, deliberately. If your whole team prompts the same tool the same way, you’ll converge internally too. Vary the inputs, the angles, the sources. Better: make sure the human doing the final pass has range the model lacks — actual experience in the category — so the last edit pushes away from the center, not toward it.
- Audit for sameness on purpose. Once a quarter, put your last five pieces next to five competitors’. Could a reader tell which are yours with the logos off? If not, the machine is winning, and no individual piece being “good” will save you. This is a distinctiveness check, and it belongs on the same shelf as your performance metrics.
The uncomfortable version
Here’s the honest tension, because pretending it away is how you lose. The individual lift is genuinely valuable, especially for lean teams — Doshi and Hauser found the biggest gains went to the least naturally confident writers, which is exactly the founder running marketing by default with no senior marketer in the building. AI is a real equalizer for you. Turning it down is not the smart move.
But an equalizer is, by definition, a sameness engine. The same force that lets you punch above your weight is pulling your weight class toward one indistinguishable blob. Both things are true at once. The teams that win the AI era won’t be the ones that use it most or least — they’ll be the ones deliberate about the exact seam between what the machine accelerates and what only a human can originate.
That seam is a strategy call. It’s the difference between a marketing engine that scales your voice and one that quietly trades it for everyone else’s.
If your marketing has gotten more productive but harder to tell apart from the competition, that’s the convergence, and it’s fixable — but it starts with a point of view no tool will hand you. That’s the work I do with founder-led teams: finding the position that’s actually yours and building the engine around it. Book a free 30-minute fit call and let’s find your seam.
Sources: Anil R. Doshi & Oliver P. Hauser, “Generative AI enhances individual creativity but reduces the collective diversity of novel content,” Science Advances 10, no. 28 (July 2024), DOI 10.1126/sciadv.adn5290.
superwired is a fractional CMO practice helping founder-led and growth-stage companies build marketing that’s wired for the AI era — real strategy and senior leadership, without the full-time cost.
