The expansion of generative AI has catapulted all industries into a new world: A world watered by the ‘sea of sameness’. Open the last five pieces of content your brand puts into the world. An email, a LinkedIn post, a landing page paragraph, whatever’s closest at hand. Read them back to back, and listen past the typos for a pulse of a rhythm. Could you tell, without checking, that they came from you and not from a template a hundred other companies pulled from the exact same well?

For a lot of brands right now, the answer is no.

What is the Sea of Sameness? And How is it a Metaphor for Generative AI?

That gap has a name: the Sea of Sameness. It’s the point where so much brand content, imagery, and even code gets generated from the same handful of models, trained on the same overlapping slice of the internet, that “polished” and “forgettable” have become the same word.

It’s more than a vibe: it’s measurable, and in at least one case, researchers have proven it’s causal.

Here’s the thesis: when everyone has access to the average of all human data, average becomes the default output. The tools are doing exactly what they were built to do: produce the statistically likely answer, and the statistically likely answer, by definition, cannot be distinctive. That makes distinctiveness the last scarce asset left on the table, well past a creative nice-to-have. Creativity has only ever been created outside the average; only and ever outside of it.

The Three Pillars of Algorithmic Blandness

Sameness is showing up in three places at once, stacked on top of each other.

1. Visual Standardization

Open ten competitor websites in your category right now. Same soft-gradient hero section. Same rounded-corner illustration style. Same stock photo of three people laughing at a laptop none of them are actually looking at.

AI image models are trained on the same datasets, so they converge on the same “modern and clean” aesthetic. Your visual identity now has to stand apart from a machine’s best guess at every competitor’s identity at once.

2. Copy Distinctiveness/Similar Softness

This one has a name and a mechanism now: model collapse. A team of Oxford, Cambridge, and Imperial College London researchers, publishing in Nature, showed that when a model is trained, even indirectly, on its own prior outputs, it progressively loses the “tails” of the original data: the unusual word choices, the odd phrasing, the outlier ideas that made any one piece of writing distinct.

In one of their experiments, the ai model fed on its own past selves,  took a passage about medieval church architecture and, nine generations later, was producing unrelated riffs about rabbit populations. Strange example. Serious implication: marketing copy is sliding down the same slope, just more slowly. Every brand voice trained on the same overlapping pool of internet text, and increasingly on other brands’ AI-generated copy, converges toward the same safe, statistically average center of language.

3. Tactical Over-Indexing

The advice everywhere right now is to produce more: more posts, more emails, more variations, faster.

A natural experiment out of London Business School found the opposite. When a group of small businesses lost AI access entirely, they posted less often and wrote shorter posts, and their audience engagement went up. Position was the lever all along, and volume could never pull it.

Why “Good Enough” Is a Commercial Liability

Here’s where this stops being a creative concern and starts being a line item.

In March 2023, Italy’s data protection authority banned ChatGPT outright for four weeks: a blanket, country-wide block imposed over privacy concerns, unrelated to marketing, which is exactly what makes it useful as a natural experiment. Researchers at London Business School tracked hundreds of independent restaurants in Milan through the ban and compared them to similar restaurants in Paris, Lyon, Madrid, and Valencia, where ChatGPT stayed available the whole time.

During the month Milan lost access, relative to the control cities:

• Similar language across competitors’ content dropped 15%
• Sentence-structure similarity dropped 12%
• Meaning-level and tone similarity each dropped a few points more
• Average customer engagement rose roughly 3.5%, despite businesses posting less often and writing shorter captions

Take away the tool everyone was using to write captions, and the captions stopped sounding like each other’s. Take away the sameness, and customers paid more attention.

The effect was worst exactly where you’d expect: restaurants serving non-local cuisines, run by owners more likely to lean on ChatGPT to write in a language and idiom that wasn’t natively theirs, showed the steepest homogenization pulled back during the ban. The businesses with the least established brand voice to fall back on were the ones AI had flattened the hardest.

What Does This Mean for Brands?

The fix comes down to having enough of a real voice going in that the tool has something worth amplifying instead of averaging.

It’s not a simple “more distinct always wins” would be too simple of a story. The relationship between similarity and engagement curves: moderate similarity to your category (the conventions that make you legible as a restaurant, a law firm, a SaaS company) helps, up to a point. Past that point, more sameness costs you. The businesses getting hurt are the ones that sound exactly as each other.

Genericness carries a cost either way you cut it. Kantar’s own brand research, drawn from tens of thousands of brands, has found a consistent relationship between how unique a brand is perceived to be and how much more a customer will pay for it. Sameness is a discount you’re handing your competitors for free.

Visibility is cheap now. Everyone has it. So buyers choose the brand that’s distinctive enough to trust.

The Anti-Generic Playbook

This whole piece is an argument for putting something in front of AI worth protecting.

1. Radical Clarity and Opinion

A real point of view is one of the only things a model can’t generate on your behalf, because it has to be told exactly what to believe before it can echo it back. If your position wouldn’t make at least one segment of the market slightly uncomfortable, what you have is a preference nobody would ever fight you on.

2. Human Touchpoints

The Milan data makes an uncomfortable point here: the restaurants that leaned hardest on AI, and got homogenized hardest for it, were the ones with the least established voice of their own to begin with. A strong brand voice holds up. What AI erases is the absence of one.

The fix is making sure there’s something specific enough on the input side: proprietary language, a real point of view, a documented voice that fails a “does this sound like us” test when it’s wrong, and a visual system distinct enough that it wouldn’t survive being run through the same generator your competitors are using. Then the tool has something worth amplifying instead of averaging.

Some of this can’t be solved on the input side at all, and that’s the point. A model can write your caption and generate your hero image, but it can’t design the box something arrives in, run the room at a live event, or decide how your team sounds on the phone. Those are physical, sensory moments, and they’re becoming more valuable precisely because they sit outside anything a prompt can touch. Unlike a clever headline, they can’t be replicated by a competitor’s model in an afternoon. That’s what turns a nice touch into a moat.

3. Brand as the System Prompt

Here’s the idea worth sitting with. A system prompt governs everything a model produces before a single word gets generated. Your brand strategy has to work exactly the same way, governing every freelancer, every tool, every AI system that touches your content, instead of getting bolted on after the fact to clean up whatever came out.

This cuts both ways. Harvard Business Review research on AI-mediated customer decisions found that companies now have to actively manage how AI tools describe them to prospective customers, on top of how they use AI themselves. One boutique hotel in the study started monitoring what AI assistants were telling searchers about the property and rewrote its public content specifically to correct what wasn’t landing accurately or distinctly.

What this means for brands: the system prompt you write now governs more than what you put out. It’s also competing with whatever a customer’s AI assistant has already decided to say about you before they ever reach your site.

Direction Over Speed

The sameness problem was here before AI. What AI did was hand every competitor the same accelerant at the same time. AI scales execution. It always will. Strategy is what decides which direction that execution moves in, and that part was never going to come from a model.

The researchers studying this at the model level found something that should make the urgency land harder: as AI-generated content fills more of the internet, future models increasingly train on that same synthetic output, compounding the drift toward sameness with every new generation. The pressure toward average is a cost that keeps growing.

There’s a name for where that trend leads if nobody pushes back: the dead internet theory. The idea that most of what’s online has become bots talking to bots, with no one steering. It hasn’t come true, and the reason why matters more than the theory itself: what still breaks through, what still goes viral, is still overwhelmingly the material a machine wouldn’t have created on its own. The sharp opinion. The unexpected reframe. The observation nobody else saw and made.

If there’s a reason for optimism here, it belongs to the thing this whole piece has been arguing for: distinctiveness is surviving the flood, and it’s the only thing still cutting through it.

Go back to those five pieces of content from the start of this blog. If you couldn’t tell they were yours, that’s a sameness problem, and it’s usually a lot easier to fix once you can see it for what it is.

Ready to Find What Makes You Distinctive?

That’s the conversation worth having before your next campaign brief goes out. Book a Brand Consultation Call with Bluebird, and find out what’s distinctive about your brand before a model averages it away.