Let me tell you what I keep seeing happen. A founder opens Claude or ChatGPT, describes the business in a sentence, and asks for a list of names. Ten seconds later they’ve got twenty options, and one of them is perfect. Punchy, memorable, exactly right. By the end of the week it’s on a domain, a logo, a slide deck.
Nobody’s checked yet whether it’s actually a name they’re allowed to own.
I don’t say that to make anyone feel silly for doing it. It’s a sensible way to get past a blank page. But whether you can own a name has nothing to do with how good it sounds or where it came from, and everything to do with a part of trade mark law that AI tools are, by their very nature, bad at respecting.
Why the name that sounds best is often the one you can’t keep
Here’s the part that surprises most founders. To register a trade mark in Australia, a name has to be distinctive. Section 41 of the Trade Marks Act 1995 lets an application be rejected if the mark isn’t capable of acting as a badge of origin, something that tells a customer this specific business made this, rather than simply describing the product or praising it. Call a cleaning business “Spotless” or a coffee brand “Bold” and you’ve hit that wall immediately: the word is doing the job of describing or complimenting what you sell, not identifying who made it.
Now think about how a language model actually works. It’s built to produce the most likely, most natural-sounding answer to whatever you ask it, which in practice means it reaches for words that already sound like the category: warm, competent, on-brand. Those are exactly the descriptive and laudatory words the Act treats as too generic for one business to own. The model is chasing “this sounds like a real brand.” The law is chasing “this is different enough that only one business gets to have it.” I’ve come to think of those as two systems optimising for opposite things, and most founders don’t find out they’re pulling against each other until an application comes back rejected.
And even if it clears that hurdle, someone might already have it
Say your name is distinctive. That’s only half the test. Section 44 can still block an application if it’s deceptively similar to an existing mark for related goods or services, and deceptively similar is a lower bar than most people assume, nowhere near identical, close enough that a customer could mix the two up.
Australia has already lived through a famous version of this problem, long before AI entered the picture. When Burger King expanded here, it found the name already registered by an unrelated takeaway shop, which is the entire reason the chain still trades as Hungry Jack’s in this country and nowhere else. That happened to a global company with a full legal department behind it. It happens far more easily to a founder who liked the third suggestion on a list and never looked any further.
This is where asking the AI to check for you falls apart. A chatbot has no live connection to the Australian Trade Marks Register, no way to weigh deceptive similarity the way an actual examiner does, and a habit of answering confidently whether or not it actually knows. If you’ve asked one whether a name “looks available” and taken that as clearance, what you’ve really got is a guess with good manners.
The reassuring part is you don’t need a lawyer for the first pass. IP Australia gives you a full search of the register for free, plus a separate tool called TM Checker that flags likely problems against the office’s own examination history in about a minute, no account needed. Running your shortlist through both costs nothing and takes less time than the conversation you had with the chatbot in the first place.
Why it’s getting worse, not better
Here’s what nags at me about this moment specifically. An enormous number of founders are prompting the same handful of models, trained on largely the same data, for the same kind of name. That means more businesses are converging on near-identical suggestions at the same time, in the same industries, and the register is filling up with lookalikes faster than it used to. Trade mark rights in Australia broadly go to whoever files first. Being early with something genuinely distinctive matters more now than it did two years ago, precisely because thousands of other founders may have been handed a very similar shortlist this week.
A founder my team and I got to help with exactly this
Syed Mosawi is a registered Australian trade marks attorney who’d worked across boutique and larger IP firms, and inside a Queensland cleantech startup, before he came to us. What he kept noticing was the same gap I see constantly: growing businesses without any dedicated IP expertise, valuable names and commercial assets quietly missed or left unprotected, founders spending time and money on a trade mark application without understanding the legal risk before they filed.
His original vision covered a much broader slice of intellectual property management than what he eventually built. My team and I worked through that with him, refining what the solution actually needed to be and guiding him through a series of design phases until the product had a genuine, focused shape. What came out the other side was Joi, built through his company Inventico: a professional conflict search, a legally grounded registrability report, and attorney-supported filing, all through one transparent process, so a founder can work out whether to file, amend or rethink a name before committing to an application. Inventico picked up support through the UQ Ventures accelerator along the way, including $15,000 in funding, and Joi launched properly this year.
What I like about that story is what it actually proves. AI is well suited to the mechanical side of this process, drafting, structuring, moving faster through a filing. The judgment call underneath it, whether a name is distinctive and whether it’s already spoken for, still needs a real check behind it. Syed built a business on exactly that distinction, and it’s the same one I’d want every founder reading this to hold onto.
What I’d actually do
Use AI for what it’s brilliant at: volume. Ask for thirty or more names, not five, and don’t let the tool talk you out of the strange ones early. Then apply the filter it can’t apply for you. Cut anything that describes what you do or praises it, if a competitor would want the same word, it’s not yours to keep. Favour the invented words and the ordinary words used somewhere they don’t obviously belong, those are your strongest candidates under section 41. Run the shortlist through IP Australia’s free tools before you fall in love with any of them. Once you’ve picked one, get an actual clearance search and registration done properly, a clean TM Checker result is a useful first filter, not a certificate.
That last step is worth doing through someone who does it every day. As a Hyper community member, you get a free 30-minute consultation and a discount on ongoing plans through our partners at Lawpath, who run their own trade mark registration service with real trade mark attorneys reviewing every application before it’s lodged. It turns “I think this name is fine” into an actual answer, before the logo, the domain and the pitch deck are all built around it.
None of this is legal advice tailored to your situation, and if a name matters to your business, it deserves a proper check rather than my general take on it. But if there’s one thing I’d want you to walk away with, it’s this: AI is a brilliant brainstorming partner for a business name and a hopeless judge of whether you’re allowed to keep it. That second part still comes down to a real person and a real search.
If you want to talk through your naming or brand position more broadly, book a free strategy session with my team.
Sasha Reid is the co-founder and CEO of Hyper, an Australian startup studio. Since 2015, he has helped more than 1,528 founders turn early ideas into viable products and businesses. Founders supported by Hyper have raised $482 million in capital.
Sasha writes about validating app ideas, defining an MVP, app development costs, AI-built products, early traction and preparing startups for investment.
