The short answer
- Everyone now drafts with the same handful of models, trained on the same corpus. The default output is the average.
- Average is invisible: to buyers, and to the AI engines that now decide who gets recommended.
- The fix is not less AI. It is a voice system the AI has to follow.
The average is the default
Ask any large language model for a homepage headline and you get the same five ideas: seamless, empower, unlock, elevate, tailored. That is not a bug. A model predicts the most likely next word, and the most likely word is, by definition, the one everyone has already used.
When every competitor drafts with the same tools and accepts the first answer, a category quietly converges. Websites, emails and social posts start to read like they were written by one very polite person.
Why it costs you more than style
Sounding generic used to be a branding problem. Now it is a distribution problem too. AI search engines summarise a category and cite a few sources. They favour material that is specific, quotable and clearly attributable. Text that says what everyone says gives an engine nothing to quote.
Buyers do the same thing faster. If three proposals read the same, price becomes the only difference.
“Average is invisible.”
Build a voice the machine has to follow
- Write the rules down. Ten words you always use, ten you never use, and three sentences that sound unmistakably like you.
- Feed it proof, not adjectives. Real numbers, named clients, specific places. Specifics are the hardest thing for a model to invent and the easiest for a reader to trust.
- Make a house prompt. One shared system prompt with your voice rules, used by everyone on the team, versioned like code.
- Keep a human editor. AI drafts, a person decides. The last 10% of editing is where a brand lives.
What we do differently
Every engagement starts with a brand voice system: rules, examples, banned phrases and a prompt library that plugs into the tools your team already uses. It makes AI faster for you without making you sound like everyone else.
