Why Every Brand Is Starting to Sound the Same

If you want to know why every brand sounds the same this year, look at what produced the copy. Most of it now passes through the same three or four models, and those models are built to return the most probable next word. The most probable word is the category average. Feed a market the same averaging machine and it converges, not because anyone chose sameness, but because everyone chose the same default.

The short version: AI writing tools pull every user toward the same center, so brand voices are collapsing into one category dialect. It is measured: generative AI raises the quality of any single piece while lowering the diversity of all pieces together, and the effect holds across 22 different models.

It also strengthens as models train on their own output. And it stays invisible from inside, because each draft looks sharper than what you had, right until every competitor's drafts read the same as yours.

The average is the product

A language model does not write, it predicts. Given your prompt, it returns the highest-probability continuation drawn from everything it was trained on, then preference tuning sands off the edges that annotators found odd. What survives is the middle of the middle: the reduction in output diversity researchers call mode collapse. The tool is working exactly as designed, and its design is to hand you the consensus.

So the "brand voice" a marketing team gets from a model is not their voice. It is the annotator-consensus dialect, wearing their topic. Every team pointing the same tool at the same brief pulls copy toward the same attractor. The output reads as yours. It is not yours. It is the category average, personalised only in its subject matter.

Each writer feels sharper while the category collapses

Here is why nobody notices. The convergence does not show up as worse writing, it shows up as better writing, one desk at a time.

In a controlled study, generative AI made individual writers more capable and their collective output less diverse: stories written with AI help scored higher on their own, yet resembled each other far more than stories written without it. Every writer improves and the set of writers homogenises, both true at once, and only the first one is visible from your chair. Doshi and Hauser put it plainly: generative AI "enhances individual creativity but reduces the collective diversity of novel content."

That is the trap in one sentence. The thing that reads as quality is the convergence. You experience a lift; the market experiences a merge.

Switching models will not fix it

The obvious escape is to change tools, or to prompt harder for a distinct tone. Neither works, because the pull is not a quirk of one vendor. Researchers measured homogenisation across 22 different large language models and found the same collapse, with effect sizes between 1.4 and 2.2. The dialect is shared property.

It also compounds. As the web fills with machine text and models train on that text, the distribution narrows further, a feedback loop that does not reverse on its own. You cannot prompt your way to the exception, because instructing a model to "sound distinctive" moves you to the average of everything already labelled distinctive. This is the same reason a model is a machine for the average: the exception is the one thing it cannot generate for you.

How it feels from inside	What is really happening
"Our AI drafts are sharper than before"	Individual quality up, collective diversity down
"We'll switch to a better model"	The same convergence holds across 22 models
"We'll prompt for a unique tone"	You reach the average of everything tagged unique
"More content means more presence"	More content means more sameness

What the exception costs

If the machine defaults to the average, then distinctiveness is the one thing it cannot default into. It has to be supplied from outside the model, deliberately, by a person who knows something the training data does not. At Ivanooo we put that knowledge in three places:

1. Voice. Language a machine can measure as distant from the category center and hold as recognisably yours. Not a tone setting, a structure. AI can average your voice; it cannot evolve it.
2. Entity. An identity the machine resolves by name, so the distinctive content attaches to you and not to the category. Without it, even your sharpest writing credits the category instead of the brand.
3. Topic Authority. Proof only you own: your data, your methods, your named decisions. The material no other brand's model could have generated, because it was never in anyone's training set.

Firoz Azees built the Ivanooo instrument around that first number, the measured distance from the category average, precisely because it is the property a competitor cannot copy by prompting. The brands that stay distinguishable treat convergence as an operating condition, not a curiosity: they engineer the exception on purpose, then watch it, because the pull to the middle never stops.

The metrics most teams watch will not catch any of this. Mentions and share of voice measure whether the machine knows you exist, not whether it can still tell you apart. The test that matters is simpler. Strip your logo from a page and ask a machine who wrote it; if it cannot say, you have converged. See where your voice sits against your category's average: paste your URL, get the distance measured, with the evidence. No call required.

FAQ

Why does AI-written content all sound similar? Because a language model returns the most probable continuation, and preference tuning narrows it further toward what annotators rated as safe. Every brand using the same models is pulled toward the same statistical center, so the outputs converge on one category dialect.

Won't switching to a different AI model fix the sameness? No. Homogenisation has been measured across 22 large language models with consistent effect sizes. The averaging is a shared property of how these systems are trained, not a flaw in one vendor, so changing tools moves you between near-identical centers.

If AI improves my writing, how can it be a problem? Because individual quality and collective diversity move in opposite directions. Each piece gets better while all pieces get more alike. The improvement is real and visible; the convergence is real and invisible until the whole category reads the same.

Can I just prompt the model to sound distinctive? No. Instructing a model to be distinctive targets the average of everything already labelled distinctive. Distinctiveness has to come from outside the model: a real voice structure, a resolvable identity, and proof only your brand can produce.

How do I know if my brand has already converged? Run the logo test. Take a page of your content, remove every name and logo, and ask an AI who wrote it. If it cannot attribute the writing to you, the machine cannot tell you apart from your category, and neither can the buyer reading its answer.