In the past year, we've noticed something strange in every other naming session: the client shows up with a document full of ChatGPT-generated options. We read through them carefully. And, invariably, they all sound like... the same three or four words recycled into different combinations. Nova, Lyra, Zenith, something-ify, something-io. Not because the AI is bad, but because it's trained on the exact same patterns every other AI sees, so it produces the exact same patterns. The result: an entire generation of brands that sound like they came out of the same algorithm. Because, in a way, they did.

That doesn't mean AI is useless for naming, it's a decent tool for generating a quick volume of raw options. The problem starts when it's the only tool, and when a name gets picked before there's a clear answer to a much more important question: what exactly does this name need to say about the brand?

A name doesn't describe the brand, it positions it

The most common mistake we see clients make is believing a name has to "explain" what the company does. That's how label-names are born: FastCargo, SmartClean, TotalCare. The problem isn't that they're ugly, it's that they're impossible to find online (search "fast cargo" and you'll get ten thousand identical results), impossible to protect legally (you can't trademark a common descriptive word), and, worse, they say nothing about the brand's personality.

A good name doesn't describe what you do. It positions how you do it and why it matters. There are, broadly, four directions a name can come from:

• Descriptive, states plainly what you do (rarely recommended today, for the reasons above)

• Invented, a new word, a morpheme mashup, a term borrowed from a lesser known language (this is also where the AI trap lives: invented badly, it sounds generic; invented well, it can become the next Google)

• Experiential, ties to the experience of using the product or service

• Evocative, doesn't describe literally, it suggests metaphorically; it draws on culture, myth, story, association

The big brands you remember without effort are almost always evocative. Not because they sound pretty, but because they tell a story your brain retains without having to memorize it.

Why every AI-generated name sounds the same

A language model, when asked for a brand name, does exactly what it does with any other task: it searches for the statistical average pattern in the data it was trained on. The result is a kind of "average of all names", pleasant, pronounceable, completely risk-free, and precisely because of that, completely without fingerprint. It's a phenomenon the naming industry already knew about before AI existed: an "invented" name can be technically unique, a combination of letters that never existed before, but if ten thousand other companies independently arrived at the same type of combination (the same soft syllables, the same -ly or -io endings), structural uniqueness stops meaning anything. It's like a snowflake in a blizzard: each flake is, technically, unique. In a blizzard, that stops mattering.

The difference between a good invented name and a mediocre one isn't how "new" the word sounds, it's how many layers of meaning it can carry. An algorithm optimizes for pronounceability and for avoiding obvious collisions. It doesn't optimize for story, because story doesn't come from a statistical pattern, it comes from the specific context of the brand: where the company comes from, who it speaks to, what it wants to mean ten years from now.

How we did it for MyKoolio

When we built the MyKoolio brand, the company didn't exist yet, we created the name, the domain and the philosophy from scratch, before the product had an identity of its own. We didn't start from a list of automatically generated suggestions. We started from positioning: an education platform for children, which had to sound playful without sounding childish, memorable for a parent who hears the name once, and flexible enough to grow alongside the product. Only after the positioning was clear did we test naming directions, and only the directions that supported that exact positioning stayed in the conversation.

The same logic shaped the name ROCO Chicago: the organization needed a name that could work simultaneously as a community anchor for the Romanian diaspora and as a name easy to pronounce and remember for an American audience, a very specific positioning, impossible to land on by chance or by mass automated generation.

The process that actually works

1. Map the competition before you write a single name. Write down every name in your category. Group them visually, descriptive, invented, experiential, evocative. You'll immediately see where the crowding is (probably the "generic invented" zone) and where the open ground is.

2. Lock the positioning before generation. What do you want the audience to feel about the brand, not what the product does, what the person using it feels? Every candidate name needs to be tested against that answer, not against one person's personal taste on the team.

3. Generate a lot, alone, then discuss in a small group. The best naming ideas don't come out of brainstorming sessions with ten people and a whiteboard, they come from focused, individual work, followed by a short, sharp discussion with 2-3 people who have real decision-making authority. The more people involved, the more certain you are to land on the safest, and most boring, name on the list.

4. Check the trademark before you get emotionally attached. A preliminary trademark office search and a basic Google search will save you months of delay later. Do this early, not at the end.

5. Don't let the .com domain decide the name. A compound domain (try-thename.com, yourname.io) is a perfectly acceptable compromise for a strong name. A perfect domain for a weak name isn't a win.

6. Test the name out loud, not just on paper. Have a colleague call you and answer by saying the company name. How does it sound? Can it be written easily on a business card, but also spoken naturally on the phone, in traffic, in a pitch?

The bottom line

A good name doesn't need to be "safe." It needs to be yours, impossible to mistake for anyone else's and impossible for an algorithm that knows nothing about your brand's specific story to generate. AI can be a decent tool for volume, but the final call still needs human context: who you are, what you want your brand to mean ten years from now, and why someone should remember exactly this name and not your competitor's.