Naming a brand with AI
Use AI to widen the list of names, then let fixed criteria and proper availability checks decide what survives.
Ask a model for brand names and it will give you fifty in a few seconds. Most will sound plausible. Many will already belong to someone else. Speed is not the difficult part of naming; choosing and clearing a name is.
AI is useful here as a wide net. It does not replace the brief, the shortlist discussion, or a trade mark search. This is a way to use it without mistaking volume for progress.
Write the criteria before the first prompt
A name is judged against a job. Decide what the job is while you still have no favourites. Who has to say the name aloud, and in which languages? Does it need to work as a web address, a social handle, or a word on a small label? Should it describe the product, suggest a feeling, or stand apart as an invented word?
Write down what is ruled out as well: words competitors already use, anything that is hard to spell after hearing it once, and anything that ties the business to a product it may outgrow.
Ask for routes, not a list
A single request for “fifty names” tends to return fifty versions of the same idea. Ask for distinct routes instead, and name them. For a fictional example, take a small-batch hot sauce made in Cape Town. Useful routes might be:
- Descriptive: names that say what it is or where it comes from.
- Suggestive: names that hint at heat, patience or craft without stating them.
- Invented: new words built from sounds that fit the product.
- Borrowed: real words from another field, such as geology or music.
Generate each route separately, with the criteria attached. Then ask for more of whichever route produced something interesting. You are steering a search, not accepting a result.
Cut the list quickly and by rule
Go from a few hundred candidates to about twenty using the criteria alone. Say each one aloud. Type it as a web address. Read it next to the competitors’ names. Remove anything that fails a rule, even if someone likes it.
Models are also prone to returning names that already exist, because existing brand names are part of what they learned from. A name that feels instantly right may feel that way because you have seen it before.
Check availability with real sources
Do not ask the model whether a name is available. It cannot see a trade mark register, and it will often answer confidently anyway. Check the sources yourself.
- Search engines and app stores: look for businesses using the name, or a near spelling, in your category.
- Trade mark registers: search free public tools such as WIPO’s Global Brand Database and TMview, plus the register for each country you plan to trade in.
- Company registers: check the registered company names where you will incorporate.
- Domains and handles: see what is taken, and by whom.
These searches are a first screen. They can rule a name out; they cannot clear it. Similar-sounding marks, related categories and unregistered use all matter, and they need a trade mark professional before you commit money to a name.
Check meaning in the languages you will trade in
A model can flag obvious problems in other languages, and that is a reasonable first pass. It is not a substitute for a native speaker, who will also notice slang, tone and unfortunate associations. Ask a person for every market that matters.
Test the survivors in context
By now there may be three to five names left. Put each one where it will live: on the label, in an email subject line, in a sentence a customer would say to a friend. A name that looks fine in a list can feel wrong on a bottle.
This is a good use of AI image and layout tools. Quick mock-ups make the comparison concrete, as long as everyone understands they are tests and not finished design.
Keep the decision record
Record the criteria, the routes explored, the names removed and why, and the checks carried out with their dates. When someone asks six months later why a favourite was dropped, the answer is written down. If a conflict appears, you can show what was searched and when.
Let the model widen the list. Let the criteria and the checks narrow it.
The value of AI in naming is range: more routes explored in an afternoon than a team would cover in a week. The decision still belongs to people who know the business, and the clearance belongs to someone qualified to give it.
Sources and further reading
Search tools checked on 10 November 2026: WIPO Global Brand Database, TMview and the UK trade mark search. The hot sauce example is fictional. This article is general guidance, not legal advice.