Creatives and AI: A New Economy of Originality

---

Most of the conversation about AI and creativity focuses on the idea that when everyone uses the same models, the outputs converge and collective novelty declines. The research supports this. Doshi and Hauser (2024) showed it empirically, and anyone who has seen three agencies pitch the same client in the same week can feel it intuitively. But this framing misses a second-order effect that runs in the opposite direction, and it may matter more.

In an AI-mediated creative economy, you do not need to be original in every dimension of your work. You need to be original in one — the idea — and you can borrow competence everywhere else: the tone, the layout, the prototype, the visual language. Generative tools have made the borrowed layers genuinely good, which means the original kernel receives more attention, not less. The result is concentration, not dilution.

And because AI systems absorb and recombine whatever enters circulation, an original idea now compounds faster than it ever could through human diffusion alone. It becomes a resuable infrastructure within months, rather than decades.

The practical implication for designers, strategists, and product people is uncomfortable but clear: originality still pays, but it pays differently. Not through copyright or licensing, but through reputational compounding — influence, audience, consulting demand, the quiet accumulation of being known as someone who introduced something new. The system depends on original ideas continuing to enter circulation. The real risk is that the people capable of producing something original will stop doing it, because they misread the economics.

---

Will we not be original anymore?

The dominant anxiety about AI and creative work runs in one direction: that generative tools flatten everything, that when everyone uses the same models trained on the same data, the outputs converge until nothing feels new. There is evidence for this. Doshi and Hauser (2024), in a study published in Science Advances, found that writers who used GPT-4 for story ideas produced individually stronger work — but their stories became measurably more similar to one another. The collective novelty of the output declined even as individual quality rose. Moon et al. (2024) documented a comparable pattern in ideation tasks: the more participants relied on LLM-generated starting points, the narrower the range of ideas became across the group. A Wharton research panel convened in late 2025 described this as a structural risk — when AI drives early ideation, outputs converge, and the competitive advantage of having ideas at all begins to erode.

There is a second-order effect of AI-driven homogenization: the rising value of the original ideas that do exist, precisely because they now travel further and compound faster than they ever could before.

This is a real phenomenon, and it deserves the attention it has received. But there is a second-order effect of AI-driven homogenization that the current discourse almost entirely ignores, and it runs in the opposite direction: the rising value of the original ideas that do exist, precisely because they now travel further and compound faster than they ever could before.

Consider what actually happens when a designer, a strategist, or a product team has a genuinely original idea — a novel way to frame a problem, a distinctive visual grammar, an unexpected service model. In the pre-AI workflow, turning that idea into finished work required originality at nearly every layer: original copy, original layout, original illustration style, original presentation format. The idea had to survive a long chain of execution, and at each link the team had to either produce something new or accept something generic. The creative energy was distributed across the entire surface area of the deliverable, which meant that the original kernel — the actual insight — often received less attention than it deserved.

Building on originals

Generative AI changes the economics of this chain. The tone of voice expressing the idea no longer needs to be invented from scratch; it can be selected from a palette of proven registers and adapted in minutes. The visual identity carrying the idea no longer demands a bespoke system built over weeks; it can draw on established design languages and be assembled with precision and speed. The presentation format, the copywriting, the interface patterns — all of these layers can now be handled competently, sometimes excellently, by tools that recombine existing creative material. What remains irreducibly demanding is the idea itself: the thing that determines whether the work says something that has not been said in quite that way before.

This is a reallocation, not a loss. The effort that used to be spent on being original across every dimension of execution can now be concentrated on the dimension where originality actually produces the most value — the conceptual one. A team that has one strong original idea and executes it using borrowed competence at every other layer will, in most professional contexts, produce work that is more distinctive than a team that spreads thin originality across every surface. Because the borrowed layers are now genuinely good — generative tools produce competent copy, serviceable layouts, functional prototypes — the original idea sits inside a vehicle that does not undermine it. The contrast between the one original element and the competent-but-familiar everything else may even sharpen the idea's visibility. In a sea of polished sameness, the thing that is actually new stands out more, not less.

This produces a compound effect that is worth examining. When an original idea circulates (through a product, a campaign, a published framework, a public talk) it enters the pool of material that AI systems draw from. The models that power generative tools are, at bottom, recombination engines: they absorb patterns, styles, and structures from their training data and reassemble them in response to prompts.

An idea that is genuinely original, once it has been published and distributed, becomes raw material for millions of subsequent AI-assisted outputs. Its influence propagates not because people consciously credit it, but because the pattern becomes part of the substrate. The Swiss typographic grid, Dieter Rams's design principles, the jobs-to-be-done framework — these are all examples of original ideas that now function as infrastructure. They were invented once, and their influence has compounded across decades because other people could use them without having to reinvent them.

AI accelerates this compounding. A distinctive visual language, once published, can be referenced, adapted, and recombined by generative tools within weeks rather than years. A novel strategic framework, once articulated clearly enough to be indexed and retrieved, begins appearing in AI-assisted analyses almost immediately. The cycle from original idea to widespread recombination, which used to take years or decades of human diffusion, now happens in months. This means that the return on producing an original idea has, in a sense, increased: the idea's surface area of influence expands faster and reaches further, even as the formal mechanisms for crediting and compensating that influence remain mostly unchanged.

Originality, rewarded

This is where the reward question becomes interesting. Traditional copyright was designed for a world in which creative value resided primarily in the specific expression — the particular arrangement of words, images, or sounds. It protects the artefact, not the idea. But in an AI-mediated creative economy, the most durable value increasingly resides in the idea itself — the conceptual contribution that remains recognisable even after it has been recombined, adapted, and absorbed into the substrate of generated outputs. Copyright is largely silent on this kind of influence.

The market, however, is not. Social media has produced an alternative reward system — imperfect, noisy, often inequitable, but real — that does compensate originality at the idea level. When a designer publishes a distinctive approach and it circulates widely, the reward is not a licensing fee; it is reputational capital that converts into speaking invitations, consulting engagements, hiring interest, and audience growth. The creator economy, for all its structural flaws, has made visible a mechanism that copyright theory always struggled to formalise: the attribution of influence. Platforms like LinkedIn, Substack, and even TikTok function as rough-and-ready systems for tracking who introduced an idea into circulation, and the social proof that accumulates around early originators is a form of compensation that operates entirely outside intellectual property law.

This is not an argument that the system is fair. It rewards visibility and timing as much as it rewards depth, and it systematically favours those who can narrate their ideas fluently over those who develop them quietly. But it is an argument that originality is being rewarded, and that the reward pathways are multiplying rather than contracting. The economic logic is straightforward: in a world where execution quality is increasingly commoditised — where anyone with access to generative tools can produce competent copy, serviceable design, and functional code — the scarce input is the one thing the tools cannot reliably generate from within themselves, which is an idea that is genuinely new.

The practical implication

For professionals in design, strategy, and product work, the practical implication is counterintuitive but important. The pressure is not to be original in every aspect of the work. The pressure is to be original in one aspect — the one that matters — and to be comfortable borrowing everywhere else. This requires a kind of creative confidence that the current discourse around AI and originality tends to undermine: the confidence to say that the idea is yours, even if the tone is borrowed, the layout is adapted, and the prototype was assembled with AI assistance. The originality of the work resides in the decision about what to make and why, not in the handcraft of every surface it touches.

The compound dynamic also creates a quiet obligation. If the system depends on original ideas entering circulation in order to replenish the material that AI tools recombine, then the professionals who produce those ideas are performing a kind of public service — seeding the commons with new patterns that everyone else can use. The risk is not that AI strangles originality. The risk is that the people capable of producing original ideas stop doing so because they perceive, incorrectly, that originality no longer pays. It pays differently now. It pays through influence, through compounding, through the slow accumulation of reputational weight that accrues to people who are known for having introduced something new. Understanding this shift — and acting on it — may be the most important strategic adjustment that creative professionals can make in the next several years.

---

Sources cited:

• Doshi, A.R. and Hauser, O.P. (2024), "Generative AI enhances individual creativity but reduces the collective diversity of novel content," Science Advances, 10(28).
• Moon, K., Green, A.E. and Kushlev, K. (2024), "Homogenizing Effect of a Large Language Model on Creative Diversity," preprint.
• Ashkinaze, J. et al. (2025), "How AI Ideas Affect the Creativity, Diversity, and Evolution of Human Ideas," arXiv:2401.13481.
• Hosanagar, K., Doshi, A., Hauser, O. and Boussioux, L. (2025), "How AI Shapes Creativity: Expanding Potential or Narrowing Possibilities?", Wharton Human-AI Research Panel.
• Gaenssle, S. (2026), "Beyond Likes: Social Media Remuneration and Financing of Digital Cultural Production," in Cultural Funding and Finance, Springer.
• NYU JIPEL (2025), "Beyond Incentives: Copyright in the Age of Algorithmic Production."