AI's Missing Link: Constrained Creativity Still Needs the Human Touch

I was recently searching for the perfect domain name for a product I'm working on, a challenge that reinforced what I've been observing about the role of human creativity in the age of AI. I have this love-hate relationship with domain names because they require a specific kind of creativity: distilling identity and meaning into words that must also navigate the messy realities of availability, trademarks, and business context. This time, I decided to use ChatGPT and Claude as sounding boards, knowing that domain availability and trademark constraints limit the use of obvious industry terms.

What I quickly experienced was the superpower of generating names, but no way to encode the constraints that actually mattered. When the .com was taken, should I accept .io or .app, append "App" (e.g., NameApp.com), or try to buy the .com? Those are context calls, not autocomplete.

This process reminded me of what Ralph Hinkley in The Greatest American Hero must have felt: superpowers at my disposal with no manual to guide me. For those unfamiliar, Ralph was a teacher who received a powerful alien suit but immediately lost the instruction manual, spending episodes crashing through buildings while learning his powers through trial and error. AI gave me the superpower of possibility, but not the judgment to navigate real-world constraints.

My search process began with AI prompts that explained my product in detail, providing a theme (such as tree names) to start the brainstorming. With enthusiasm, ChatGPT and Claude would return a list of words and provide insight into how these words could relate to my product. Rarely did an individual word feel right, so I would ask for complementary one-syllable words to pair with. I was thinking along the lines of BlueSky, Facebook, Snapchat, and TikTok. What became clear was that generative AI is brilliant at remixing patterns it has seen before, but struggles to originate ideas that feel genuinely new.

I provided prompt after prompt and tried different themes, requesting over and over again to exclude words like "flow" and "loop" that the AI clearly lifted from other brand names. And when I found an interesting name, I had to manually check the domain registrars and search the USPTO trademark database for availability. It's never as simple as a quick search - when domains are taken, you must decide if they're worth premium prices and which alternative extensions make sense for your business. Trademark searches require more than a keyword check. You must read the filings themselves to spot genuine overlap between business descriptions. Both demand the kind of nuanced judgment at which humans excel. AI could generate endless lists of brandable names, but we kept crashing in the same spot. Every promising domain led to the same dead end: already taken or trademark issues.

The ironic ending? After all this AI brainstorming, my ingenious wife came up with the root word we ultimately chose. But I did use all the prompting techniques I'd developed to evaluate her suggestion. This experience reinforced a broader lesson I've been seeing across AI applications: these tools excel at analysis and iteration, but struggle with constrained creativity, the kind that has to navigate real-world limitations. AI isn't the creator; it's the critic — at least when creativity has to pass real-world constraints like availability, trademarks, and business context. Whether you're naming products, architecting software, or making strategic technology bets, the real skill is learning to guide AI's analytical power while keeping creativity and judgment human.