Part 10 of the “UX × AI” series.

Ten articles ago, I told you why I was writing this series.

I said that I had watched UX absorb wave after wave of disruption over 25 years and that in every wave, the industry split into uncritical excitement and defensive fear, with almost nothing honest in the middle. I said that gap was what UX × AI would fill. I said this series would think clearly, acknowledge genuine uncertainty, and be honest enough to say what it does not know alongside what it does.

I want to open this final article by asking whether I have kept that promise—and by closing the series with the question I believe is the most consequential one we have not yet fully answered.

Across nine articles, we reframed the fear into a working relationship—AI as an intern, not a replacement. We recognized that prompting is a brief-writing skill designers already possess. We protected the one thing that cannot be delegated—genuine empathy with real human beings. We built a disciplined, 30-day path to genuine workflow fluency. We dismantled the myth that more data produces more insight. We reclaimed authorship—designing with AI rather than for it. We learned to evaluate tools with the rigor our careers deserve. We turned the lens around and audited the UX of AI products themselves, finding them wanting. We looked at the frontier where AI stops waiting for instructions and starts acting on its own.

Each of those eight prior articles eventually arrives at the same unresolved question, sitting quietly beneath them all. If AI generated the wireframe, who designed it? If AI drafted the copy, who wrote it? If AI proposed the user flow that the team refined and shipped, whose work is in the product?

This article answers that question as honestly as the law, the ethics, and the profession currently allow.

The question the industry has been avoiding

I want to be direct about why this question matters, beyond the philosophical interest of it.

Authorship determines credit. Credit determines career advancement, portfolio value, and professional reputation—the currency that every working designer depends on. Authorship determines legal ownership. Ownership determines who can license a design, who can prevent its use by a competitor, and who bears liability when a design causes harm. Authorship determines accountability. And accountability, as we established across this entire series, is the one thing that AI categorically cannot hold—which means that somewhere in every AI-assisted design process, a human being is the author, whether the organization has clearly designated who that is or not.

The industry has been treating this as a question to be resolved later, informally, case by case—while design teams ship AI-assisted work, designers list AI-assisted projects in their portfolios, agencies bill clients for AI-accelerated work at rates set before AI was part of the process, and disputes quietly accumulate about who deserves credit when a design succeeds and who bears responsibility when it fails.

Later has arrived. The legal system has already begun answering this question, whether the design industry has caught up or not.

“AI cannot be an author. Only human beings can hold copyright authorship. AI-assisted works can be copyrighted—if the human contribution is sufficiently creative and controls the expressive elements of the final work. Prompts alone are not enough.” — U.S. Copyright Office, Copyright and Artificial Intelligence Report (2025–2026)

What the law actually says right now

Let me walk through what is genuinely settled, because the design community needs this clarity, and most of the conversation happening in design forums and social media is operating on assumptions that are no longer accurate as of 2026.

The foundational case is Thaler v. Perlmutter, in which an inventor attempted to register copyright for a work generated entirely by an AI system, listing the AI itself as the author. The case worked its way through the United States court system for several years. The full D.C. Circuit denied an en banc rehearing in May 2025, and the U.S. Supreme Court denied certiorari in March 2026—leaving the ruling intact and effectively settled. The court’s reasoning is worth understanding precisely: copyright law’s use of terms like “person,” “widow,” “children,” and “heirs” in its duration and ownership provisions presupposes a human being with a finite lifespan. There is no room in this legal framework to read “authorship” as encompassing a machine. The court also rejected the argument that the work could be owned through a work-for-hire arrangement, reasoning that the work-for-hire doctrine assigns ownership of a copyrightable work to an employer—but it presupposes that a human authored the work in the first place. A work with no human author has nothing to assign.

The practical consequence of this ruling, now firmly established, is this: work generated entirely by AI, without meaningful human creative control over its expressive elements, is not copyrightable. It exists in a legal vacuum—not owned by the AI, not owned by the company that built the AI, and not owned by the person who prompted it, unless that person exercised meaningful creative control beyond simply entering a prompt.

This is the second critical clarification, and it is the one most relevant to working designers. The U.S. Copyright Office has stated explicitly that merely entering prompts into a generative AI tool does not make the user an author of the output. The landmark case here is Zarya of the Dawn, a graphic novel that was initially granted copyright registration in 2022, including its AI-generated illustrations. In early 2023, the Copyright Office revoked the registration for those specific illustrations—while retaining protection for the text and the arrangement of elements, which were created by the human author—because the text prompts and subsequent modifications the human made to the AI’s output were not sufficient to qualify the images themselves as a work of human authorship.

This is the distinction the entire industry needs to internalize. Prompting an AI to generate a design is not, by itself, an act of authorship in the legal sense. Authorship requires meaningful human creative control over the actual expressive content of the final work—selection, arrangement, modification, and the kind of substantial creative decision-making that goes well beyond describing what you want and accepting what you receive.

The critical distinction: Curation vs. creative control

This legal distinction maps with striking precision onto the authorship argument I made in Part 6—designing with AI, not for it. It turns out that the difference I described there, between the designer who maintains genuine creative agency and the designer who has quietly ceded it to a tool’s defaults, is not just a professional and craft distinction. It is, increasingly, a legal one.

The Copyright Office’s guidance establishes that AI-assisted works may qualify for protection only if the human’s contribution is substantial, demonstrable, and independently creative—not merely the selection of one AI output over another, and not merely surface-level edits to an output that remains, in its essential expressive character, the product of the AI’s generation rather than the human’s design judgment.

This means that the designer who treats AI output as a hypothesis to evaluate and substantially reshape according to specific user understanding—the practice I advocated throughout Part 6—is also the designer who is building a legally defensible claim to authorship of the resulting work. And the designer who is accepting AI output with minimal modification because it is good enough and the deadline is close—the practice I warned against—is also the designer whose claim to ownership of that work is, under current law, genuinely fragile.

This is not a small thing for working professionals to understand. A designer who has built a portfolio of work substantially generated by AI, with only minimal human modification, may discover that they do not hold enforceable rights to that work—that a client or employer could use it freely, that a competitor could reproduce a substantially similar design without infringing anything, and that the designer’s claim to have authored it, in any legally meaningful sense, is weaker than they assumed.

“For a work to qualify for protection, creative human involvement must be substantial, demonstrable, and independently copyrightable. The line between trivial modifications and meaningful human authorship is where most of the unresolved ambiguity in AI-assisted creative work currently sits.” — U.S. Copyright Office, 2025 Report on Copyrightability

What this means for the working designer, practically

I want to translate this legal landscape into specific guidance, because abstraction is not useful to a designer who needs to make decisions about real client work this month.

  • Document your creative decisions, not just your prompts. If meaningful human creative control is the standard that determines authorship, then the designer who wants a defensible claim to their AI-assisted work needs to document the creative decisions they made—what they rejected from the AI’s output and why, what they substantially reshaped, and what specific design judgment they applied that the AI’s default output did not reflect. This documentation is not bureaucratic overhead. It is the evidence that distinguishes your authorship from mere curation, both legally and professionally.
  • Be honest in your portfolio and your client communication about the nature of AI involvement. As the design community develops norms around AI-assisted work, the designers who are transparent about the role AI played—and specific about the creative judgment they applied beyond it—are building a more sustainable professional reputation than those who either overclaim full authorship of heavily AI-generated work or hide AI involvement entirely. Clients and employers are increasingly sophisticated about this distinction, and the trust cost of being caught misrepresenting it is significant.
  • Understand that your contracts need to address this explicitly. In light of the ambiguity around how copyright law applies to AI-generated content, businesses are increasingly relying on contracts to allocate risk and define ownership explicitly, rather than relying on default copyright protections that may not exist for AI-generated elements. If you are a freelance designer or running a design agency, your client contracts need explicit language about which elements of a deliverable are AI-assisted, who owns the resulting work product regardless of its copyrightability, and how liability is allocated if an AI-generated element later turns out to infringe on existing copyrighted work—a live and unresolved legal risk that the same body of case law is still working through.
  • Recognize that the absence of copyright protection is itself a business risk you need to manage. If a design element in your product was substantially AI-generated with minimal human creative modification, it may not be protected by copyright at all—meaning a competitor could potentially reproduce it without legal consequence. For organizations building genuine competitive advantage through distinctive design, this is a material consideration in deciding where AI assistance is appropriate and where the investment in substantial human creative authorship is necessary to maintain a defensible, ownable design asset.

The ethical question beneath the legal one

The legal framework, however, does not fully resolve the ethical and professional questions that this article’s title raises—and I want to address those directly, because they matter independently of what courts have ruled.

There is a real ethical issue regarding the credit and recognition that designers receive for work that involved substantial AI assistance. A junior designer who used AI extensively to produce a polished deliverable, under time pressure, with a senior designer providing the creative direction and final judgment—who is the author of that work, in the professional sense that determines how credit is allocated in a team, in a portfolio, and in a performance review? The legal answer, focused narrowly on copyrightability, does not fully address this professional and organizational question, which design teams need to resolve through their own explicit norms rather than waiting for the law to settle it for them.

There is a second ethical issue that this series has touched on throughout, and that belongs centrally in this final article: the question of what was used to train the AI systems that designers now rely on daily. The fair use status of using copyrighted creative work—design portfolios, illustration, photography, the visual output of decades of working designers—as training data for the generative AI systems that now compete with and assist those same designers remains genuinely unsettled in law. Several courts have split on this question in 2025, and the Copyright Office’s own analysis has explicitly declined to treat AI training as categorically fair use while also declining to rule it out. Appellate courts will likely need to resolve this question, and the resolution will have direct consequences for the design profession—because if courts ultimately rule that training was not fair use, it raises serious questions about the legitimacy of design outputs generated by models trained on unlicensed creative work, including, quite possibly, work produced by designers reading this series.

I do not think the design community can responsibly treat this as someone else’s legal battle. The tools we are integrating into our daily practice were built, in significant part, from the uncompensated and unconsented contribution of decades of design and creative work—quite possibly including our own. This is an uncomfortable truth to sit with, and this series has been committed to sitting with uncomfortable truths rather than avoiding them. I do not have a clean resolution to offer. I have an obligation to name it clearly, as part of the honest reckoning this entire series has tried to model.

Applying LucyUX to the authorship question

The LucyUX framework—Listen, Understand, Conceptualize, Yield—provides a useful closing structure for how individual designers and design organizations should approach the authorship question in their own practice.

  • Listen: To the specific creative decisions you are making throughout an AI-assisted design process and to the difference between the moments you are exercising genuine judgment and the moments you are accepting AI output because it is adequate and time is short. This listening—to your own practice, with honesty—is the foundation of being able to answer, for yourself and for anyone who asks, what you actually authored in a given piece of work.
  • Understand: Build an accurate, specific understanding of the current legal and ethical landscape, not the version of it that circulated in design forums eighteen months ago and has since been superseded by settled case law. Understand what your specific jurisdiction’s copyright office has ruled, what your client’s contracts currently say about AI-generated elements, and where the genuine ambiguity remains versus where the law has already settled.
  • Conceptualize: Design your own practice’s explicit norms around AI-assisted work, rather than allowing them to emerge accidentally through a thousand small unexamined decisions. Decide, deliberately, what level of human creative control you require before you consider a piece of work genuinely authored. Decide how you will document that control. Decide how you will communicate AI involvement to clients, employers, and the design community.
  • Yield: Measured in a body of work whose authorship you can defend, honestly and specifically, to a client, a court, or your own professional conscience. This is the yield that matters most as this series closes: not efficiency, not speed, not volume of output—but a practice you can stand behind, fully, because you know precisely what you brought to it that the AI could not.

Why I wrote this series

I want to close not with another framework, but with the thing I told you I would close with when I began.

I wrote this series because I believe the design community deserves better than the two options the broader conversation about AI has offered it—uncritical enthusiasm or defensive fear. Across ten articles, I have tried to give you a third option: clear thinking, grounded in evidence, honest about uncertainty, and committed to the conviction that has anchored my own 25 years in this profession.

That conviction is this. Design has always been, at its core, a deeply human discipline—an act of genuine attention to another person’s experience, translated through skill and judgment into something that serves them. AI is the most powerful tool that has entered this discipline in my career. It is not a replacement for the discipline itself. Every article in this series has been an attempt to show, from a different angle, exactly where that line sits—what AI can do, what it cannot, and what we lose if we forget the difference.

In Part 1, I told you AI is your intern, not your replacement—fast, tireless, well-read, and entirely dependent on your judgment. In Part 2, I told you that you already have the skill prompting requires, because brief-writing has always been design’s quiet superpower. In Part 3, I told you to stop calling pattern-matching empathy because the word means something AI cannot do, and protecting that meaning protects the people design is supposed to serve. In Part 4, I gave you thirty days to discover all of this for yourself, in your own practice, rather than taking my word for it. In Part 5, I told you that scale is not understanding and that the most dangerous data is the data that feels too authoritative to question. In Part 6, I told you to stay the author of your work—to design with AI, never for it. In Part 7, I gave you the questions to ask before you commit your team’s time and budget to a tool that has not earned it. In Part 8, I turned the same critical eye on AI products themselves and found the industry falling short of standards our own profession established three decades ago. In Part 9, I showed you the frontier where the interface stops waiting for permission and the design discipline that frontier urgently requires. And in this final part, I have tried to answer, as honestly as the evidence allows, the question of who we are when AI does half the work—and what we owe each other, our clients, and the profession in answering it honestly.

If you have read this series from the beginning, thank you for the time and trust that represents. I hope it gave you something more useful than reassurance or alarm—I hope it gave you a way of thinking that will keep serving you long after this specific moment in AI’s development has been replaced by the next one.

The tools will keep changing. They always have, across my entire career. What does not change is the thing this series has tried to protect in every article: the designer who knows their user, exercises genuine judgment, takes real responsibility for what they ship, and never mistakes a powerful tool for a substitute for their own care.

That designer is not going anywhere. I am glad to have spent these ten articles with you, thinking about how to make sure of it.

Thank you for reading “UX × AI.” It has been one of the most meaningful things I have written.

The complete “UX × AI” series

  1. Part 1: “AI Is Your New Intern, Not Your Replacement.” The foundational reframe: AI is fast, tireless, and well-read—and completely dependent on your judgment, context, and accountability.
  2. Part 2: “The Prompt Is the New Brief.” Prompting is not a new technical skill. It is brief-writing, a discipline designers have practiced their entire careers, applied to a new surface.
  3. Part 3: “Stop Calling It Empathy, AI Does Not Feel Anything.” The most dangerous language in the AI conversation, and why protecting the real meaning of empathy protects the users’ design is meant to serve.
  4. Part 4: “Your First 30 Days With AI in Your Design Workflow.” A week-by-week, evidence-based guide to building genuine AI fluency in your actual practice, not generic AI literacy borrowed from someone else’s workflow.
  5. Part 5: “More Data Is Not More Insight.” Why scale amplifies confidence without amplifying understanding, and why the most dangerous data is the data that feels too authoritative to question.
  6. Part 6: “Designing With AI, Not For It.” Every tool has a bias. AI is deeper and more consequential than any before it. How to stay the author of your work rather than its editor.
  7. Part 7: “How to Evaluate an AI Tool Before It Evaluates Your Career.” A no-hype checklist: the seven questions vendors will not volunteer, and the evaluation discipline that protects your time, budget, and professional credibility.
  8. Part 8: “The UX of AI Itself Is Broken.” Turning the lens around: an audit of how AI products are designed today and why the industry is failing the usability standards it established decades ago.
  9. Part 9: “Agentic UX, When the Interface Stops Waiting for You.” The biggest interface paradigm shift since the graphical user interface, and the design discipline it urgently requires, before trust collapses at scale.
  10. Part 10: “Who Owns the Design When AI Made Half of It?” The unresolved question beneath every prior part, answered as honestly as current law, ethics, and the profession allow.

References & further reading