Are AI creativity tools the new snake oil?

Every few weeks, another AI company pops into all the news feeds touting some new creative capability. This one can design full website layouts; this one can build marketing campaigns, generate videos, compose entire songs, write code, or maybe even produce a brand system from a paragraph and a handful of reference images. Each new release or program is presented as a step toward making creative work faster, easier, and less dependent on the people who have traditionally done it.

Why are we so intent on making creative work faster? When did anything meaningfully creative, something that truly moved people, get created by speeding up the process? Can you imagine Beethoven presenting the opening bars of Moonlight Sonata to AI, only to be told, “This is a great start, but the minor keys are doing a lot here. I suggest converting to major”?

Or Jane Austen, anxious to move on to the next chapter, offering ChatGPT the quintessential opening line, “It is a truth universally acknowledged, that a single man in possession of a good fortune, must be in want of a wife,” only to receive the following response:

“The sentence has a strong, distinctive voice, but the opening is very broad and ceremonious. ‘A truth universally acknowledged’ risks sounding inflated unless the exaggeration is clearly intentional. I would suggest something simplified: ‘It is universally acknowledged that a wealthy single man must be looking for a wife.’”

*Note: This was the actual ChatGPT response to that line.

It’s absurd, right? But is that not precisely what these tools are made to do? Are they not cutting out the very human part that moves us and makes our skin tingle in pursuit of what? More polish? Or is it budget?

“Just have AI do it”

Something not dominating the news feeds is the number of talented creative professionals reentering the job market because someone several levels above them decided that AI could absorb enough of their workload to justify a smaller team. In many cases, that decision was not based on evidence that AI could replace the thinking that those people contributed; it was based on the fact that AI could produce something.

A draft appeared, an image appeared, a campaign concept appeared. From the cozy C-level suite at the tippy top, the difference between producing work and understanding the work seems to have become blurred.

This isn’t entirely AI’s fault. It entered an industry that has increasingly rewarded how quickly a team can produce a thing over what that thing actually is. The website launches. The campaign goes live. The identity appears on the packaging. The research, rejected directions, uncomfortable questions, conflicting customer needs, and hundreds of small judgments that should have shaped a better result disappear behind the impossible deadline.

Once creative work is reduced to what can be seen at the end, replacing the people who made it begins to sound fairly reasonable. If creativity is merely the production of a polished image, paragraph, interface, song, or campaign, then AI is already capable of producing a great deal of it. The more useful question is whether producing something that resembles creative work means that creative thinking has taken place.

And it hasn’t.

Creativity can’t be learned.

Not everyone can paint, compose music, write a novel, design a useful interface, direct a film, or take an ordinary business problem and see a solution no one else considered. These abilities can be studied, practiced, and improved. Technique can be taught. A person can learn color theory, sentence structure, musical notation, design systems, research methods, and the mechanics of almost any creative discipline. What cannot be guaranteed is the instinct to do something meaningful with them.

That instinct is talent.

Two people can receive the same education, study the same references, use the same tools, and spend the same number of years practicing, yet one will consistently make connections the other does not see. One will know when to follow the rules and when the rules are flattening the idea. One will recognize that the technically imperfect choice is the one that gives the work its identity.

Talent can be neglected, sharpened, disciplined, and expanded, but it cannot be manufactured by collecting enough examples of other people having it. Which is the only thing AI can do.

AI can only learn from what already exists. It absorbs the products of human talent, identifies patterns within them, and uses those patterns to generate something new in arrangement, if not always in conception. It can imitate the visible evidence of creativity because it has been trained on an extraordinary amount of creative work produced by people who possessed the thing it does not. Calling that process creativity gives the system credit for the talent it borrowed.

A model can study thousands of paintings, songs, campaigns, interfaces, poems, and novels. It can learn which choices commonly appear together and reproduce their recognizable effects. It can become more technically convincing with every generation. What it cannot do is account for the first person who made a choice before there was a pattern to learn.

Someone had to decide that a painting did not need to resemble the visible world. Someone had to write a melody that violated the expectations of the period. Someone had to arrange a page, tell a story, design an object, or frame an idea in a way that initially looked wrong because no accepted version of it existed yet.

AI can’t do that.

AI Can Simulate Empathy. It Can’t Understand Need.

AI can produce language that sounds compassionate. It can recognize emotional cues, mirror a person’s tone, and assemble the phrases commonly associated with reassurance or concern, but that doesn’t mean it understands what another person actually needs in any specific moment.

Someone awaiting a diagnosis may need information presented clearly and without forced optimism. A parent navigating a medical emergency may need the most important action placed at the top of the page, not beneath three paragraphs of comforting language. A customer who has spent an hour trying to resolve a problem probably does not need a chatbot to tell them that their frustration is “completely understandable” before sending them through the same useless steps again.

Empathy in creative work is not simply about sounding kind. It’s about deciding when to reassure, when to explain, when to simplify, when to provide more detail, and when an attempt at warmth will feel dismissive or insincere.

AI can identify the emotional category and generate the expected response. A person can draw from research, observation, memory, and lived experience to recognize when that expected response is wrong.

Brand Consistency Is Not Cosmetic

AI is very good at following a brand system until it’s not. It may use the correct colors, logo, typefaces, and approved terminology while still getting the proportions, pacing, image treatment, leading, or tone slightly wrong. The result often looks close enough to pass an internal review, especially when the reviewer is not deeply familiar with the brand.

Customers are.

In behavioral science, the uncanny valley describes the discomfort people feel when something looks almost human but not quite. A similar effect happens with familiar brands. The closer something gets to looking and sounding right, the more noticeable the small discrepancies become. An obviously different design may register as a redesign. A nearly identical one with subtle inconsistencies can feel counterfeit.

The sense of something being counterfeit triggers more alarms now than it ever did. People are constantly sorting through phishing emails, copied storefronts, fake reviews, impersonated brands, and AI-generated content designed to appear legitimate. Familiarity has become one of the quickest ways consumers decide whether something can be trusted. When a brand they know well suddenly looks almost right, “almost” signals mistrust and can be the quickest way to get them to turn away from the brand for good.

A brand is not just a set of assets and written rules. It is the accumulated effect of hundreds of small decisions made consistently over time. AI can learn those decisions as patterns, but it does not understand why a particular deviation feels harmless in one context and suspicious in another. It forgets. It guesses. It hallucinates.

And in the end, I believe that will prove to cost far more than the budget saved by “just having AI do it.”

So What Is Creativity Worth?

This isn’t an argument against AI. I use it almost every day—I’m using it right now to help me clean up this article. It’s been useful beyond measure for my research, summarizing notes, cleaning up rough drafts, generating code, exploring ideas, and automating work that would otherwise consume hours of my time. It is one of the most useful tools to emerge in decades, and I suspect it will become even more capable.

But “useful” and “creative” are not synonyms. The danger with labeling them that way isn’t that AI will replace creativity. The danger is that we’ll start redefining creativity to fit AI’s limitations.

To me, creativity is the ability to observe people, recognize needs they haven’t articulated, challenge assumptions, make unexpected connections, understand emotion, build trust, and create something that did not exist because one person saw the world differently than everyone else. When we’re talking about AI, we’re talking about something else entirely.

AI will continue to become faster. It will become more convincing. It will automate an extraordinary amount of production, and that’s something we should embrace where it genuinely improves our work.

But the next truly original song, product, company, campaign, scientific breakthrough, novel, or movement will begin exactly where they have always begun: with a person who noticed something everyone else missed.

And decided to do something about it.