This is not a debate about whether AI belongs in creative work. That argument is already over. People are using it. The interesting question is how, and whether they are being honest about it. I asked writers, artists, filmmakers, and designers to describe their actual process. Not the version they perform online. The version that involves doubt, compromise, and decisions about where the human stops and the tool starts. What came back was more varied and more conflicted than any think piece I have read on the subject.

In this post we will:

• Argue that art has always borrowed, and that honesty about influence matters more than pretending otherwise.

• Share how real creatives are using AI as part of their practice, not as a replacement for it.

• Show concrete examples, including one of my own poems being edited by AI.

Art has always been stolen

Michelangelo studied Roman sculpture so obsessively that he buried one of his own pieces in acidic earth to age it, then sold it as an ancient artefact. He was caught. He was also hired on the strength of it, because the forgery proved he could match the masters.

Shakespeare lifted plots wholesale from Holinshed, Plutarch, and Boccaccio. He rarely invented a story. He reinvented how it was told.

The Rolling Stones built their early catalogue on Chicago blues. Keith Richards has said it plainly, in paying homage to Chuck Berry: “I lifted every lick.”

Likewise, Tarantino steals shots and says so. Kehinde Wiley paints Black subjects in Old Master poses. The Beastie Boys built Paul’s Boutique from over 100 samples.

This is what art does. It borrows, transforms, recombines, and responds. The question was never whether artists take from other artists. They always have. The question is whether they truly integrated those influences authentically into their practice and are honest about it.

Where AI complicates this

AI complicates the borrowing because it hides the sources, and people perceive AI-generated art to be inauthentic

When Keith Richards plays a Muddy Waters lick, you can hear the original. When Wiley paints a figure in the pose of a Titian, you can see the reference. When a rapper samples a James Brown break, the liner notes say so.

With AI-generated art, however, the assumption is the product lacks soul. AI-generated art is associated with cheap replicas produced for massive, rapid profit.

When an AI model generates an image trained on millions of artworks, the sources are invisible. No liner notes. No citation. No acknowledgement that the output stands on the shoulders of specific human work.

The copyright debate is real and unresolved. AI companies trained their models on copyrighted work without consent or compensation. That is a systemic problem, and it belongs to the companies, not to individual users.

But there is a separate, more personal question for anyone using AI in their creative practice: are you being honest about what the tool is doing, what it is drawing from, and what remains yours?

This post is a collection of answers to that question. Each contributor describes how they actually use AI in their creative work, what they have learned, and where they draw the line. The thread running through every contribution is the same one that runs from Michelangelo to the Beastie Boys: honesty about influence is what separates borrowing from theft.

How creatives actually use AI in their work

What follows are 11 accounts from writers, artists, filmmakers, and other creatives. Each one describes how they actually use AI in their creative practice. Not the version they perform online. The version that involves doubt, compromise, and decisions about where the human stops and the tool starts.

Slow AI

I write poetry. My work has been published widely, and I have several full collections, including The Poetry of Physics (CRC Press, 2025). I am also the founding editor of an established poetry journal Consilience.

I do not use AI to write poems. I use it to edit them.

Here is a recent poem of mine called ‘Gloss’:

the words come back changed
not wrong but smoothed
the way a thumb wears the face from a coin
each edge that made it singular
filed down to fit the palm
you read your sentence and recognise it
the way you recognise your face in a photograph
you don’t remember being taken

I asked Claude to read this as a critical poetry editor. To tell me where I was coasting. It flagged three places: a metaphor I was extending past its natural landing (“filed down to fit the palm”), a word doing expected work instead of precise work (“smoothed”), and a closing image that risked familiarity. I agreed with the first two. The third I kept.

AI did not improve the poem. It identified where I was settling for good enough. The judgement is mine. The editorial eye is borrowed.

Alyssa Fu Ward, PhD

When I make art, my focus is on the human.

I ask myself — How do I want my viewers to feel when they see my art?

And most importantly — Do my head, heart, and body agree that the finished product represents me completely?

Those questions don’t change just because I bring AI into my process.

I recently started a 100 day project, where I create a digital illustration every day for 100 days in a row. I draw inspiration from Substack, especially writers of AI, and real life.

I often start with a concept seed, a quote. I feed it into AI with source material and my own vision so it can create a reference image.

I could just stop there. But I don’t.

I hand draw my own creation, bridging ideas the AI evoked for me with my own.

AI is a tool that creates a warp zone where I skip levels 1-3 and start at level 4. Then, I continue through my own levels to create a product I can call my own.

AI helps me expand what is possible for me to do as an artist.

But in the end, the product is me.

Natalie Nicholson

For the last decade, my career has been in Design and Art Direction. People often assume I must hate AI: it’s a threat, it’s taking my job, it’s stealing our work. I’ve even received public backlash for exploring it.

Much of the conversation about AI art centers on theft. Not just theft from other artists, but how we’re being robbed of “craft.” But that was happening long before these tools showed up. Rushed deadlines. Shrinking budgets. Layers upon layers of approvals, and hours spent on concepts cut at the last minute. Now, I use AI to prototype and ideate with clients: to visualize ideas they didn’t yet have words for, and bring clarity to the conversation earlier through tailored imagery.

AI has also exposed what good designers have been doing all along: our job was never to “make images.” It was always about taste, judgment, and creative problem solving. The irony is that the more powerful these tools become at generating, the more important you and your strategy becomes. I love that people see that more clearly now.

Because of how I choose to create with it, AI has only expanded creative possibilities. Recently, I resurfaced a decade-old poem, turned it into an animated short film and invited my Substack community to participate. It was the most rewarding creative project of my career. And a thought kept surfacing: I couldn’t have created it without the help of AI. But AI couldn’t have created it without me, or the inspiration from this community, either.

Mia Kiraki 🎭

I work across a few different mediums (creatively) - painting, creative writing, music. Because my creative background is so varied, my ideas usually arrive as a tangled mess.

Before AI, I would scribble these thoughts on paper, and let them sit for weeks before I had the energy to decipher them (especially in screenwriting!). Now, I use AI as a dumping ground. I can offload a dozen of half-formed concepts at the speed of thought. I never use it to write my stories or give me rough angles. I use it to untangle the knot in my head so I can see the shape of what I’m trying to make.

The spark of the idea and the final execution have to belong to me.

AI Meets Girlboss

AI as an art tool, not an art substitute.

I create my art with AI, and I’m very clear about what that means. AI is a tool. That’s all. Artists have always worked through tools, whether it was charcoal, oil paint, a camera, Photoshop, or now image generation. The ethical questions are real, and they should be taken seriously. But the act of making art still begins with a person: someone who has lived, noticed things, developed taste, collected references, and formed opinions. That part cannot be outsourced.

In my case, Pinterest is a big part of the process. I use it the way other artists use moodboards or sketchbooks: to gather visual cues, sharpen my instincts, and define both what I want and what I absolutely do not want. It helps me translate vague taste into concrete direction. From there, the image generation is not random at all. It is highly constrained, heavily curated, and guided by very specific descriptions, rules, and choices.

The result is not “AI made this for me.” It is closer to: I directed this into existence using a new medium. I wrote more about that process in this post, where I break down how to turn references into something that actually feels like mine.

Dallas Payne

The line I can hold and the one I still can’t find.

I am not an artist and have never been one. I did not arrive at AI with a practice to protect or a medium to master. With my writing, the line is clear and I hold it firmly. Nothing gets published that doesn’t have skin in the game and isn’t mine in a way that actually counts. AI assists, sometimes brilliantly, but I am at the wheel and I know it.

The visual work is where the line, for me, gets complicated. What I have is taste, unused qualifications in interior design, and a fairly precise internal idea of what I want things to feel like. For most of my life, the large gap between an internal idea and something I could actually create was just a fact. Then with AI, suddenly it wasn’t and the gap vanished. AI didn’t just give me a tool, it gave me access to something I had genuinely ruled out as not mine. I cannot decide whether that is good, or whether it raises questions I don’t yet know how to answer. If I can now make things that look and feel like a creative vision, but couldn’t have made any of it without the tool doing the rendering, the motion, the interpretation of “golden hour“ into something that actually looks like the golden hour, what exactly did I contribute? The brief. The iterations. The stubbornness about what felt wrong. A life lived in many different spaces that forms how I view the world.

Although, is any of that really enough to call the visual work mine? I simply don’t know. What I know is that I keep going, iterating until visual images and animations feel right. The tool doesn’t care about the difference, I do.

Caitlin McColl 🇨🇦

I use AI mainly for images. I have a specific world in all my images – a cozy village. The woman (Elara) is an avatar of me. There’s a core colour palette too, and a specific item in each image. All those things combined into each image makes them, in my mind, unique and mine. I write about it more in an article by AI Meets Girlboss here.

I’ve also started to use AI to transform some of my static images into animated ones. I’ve created some ‘magazine covers’ featuring Elara and have a template where I simply change what Elara is doing and wearing, and add the different titles, but everything else remains the same. When I started animating them with AI (which gives them Harry Potter moving pictures vibes) that made them even more unique, and is a really fun way to be creative. After all, who else is doing that?!

Finally, I’ve used Suno, the music platform to create my podcast theme song. The exciting thing about AI is finding ways to be creative in ways that are uniquely you, that no one else is doing, or could do, in your specific, human way.

Story Beyond Play (Avery K. Tingle)

AI was a game-changer for me. Where it used to take entire months just to outline a project, now I can move from conception to final draft in about three months. I don’t have AI do my writing because it cheapens the experience. Now I have something I can bounce ideas off of at two in the morning, something that can handle all manner of editing my drafts, and most importantly, push the final draft to its best form even when I want it to be done.

More importantly, AI taught me to be a better communicator; the more clearly I could frame my requests, the better the output became. This made real-life communication much simpler and less awkward for me, because I felt like I knew exactly how to state my point without being confrontational.

My favorite feature is AI telling me which stories aren’t worth pursuing. Instead of running down every idea, Sage (my AI) will sometimes stop me gently because it either doesn’t fit the brand, or maybe it’s a story I’m not experienced enough to tell yet. AI goes far beyond organizing my notes and schedules. It was the missing piece of my life’s puzzle.

AI Cinema By Elettra Fiumi

I’m Florentine. Michelangelo feels like an ancestor. He, like all artists, first learned perspective, proportion, how to copy Roman sculpture so precisely he could sell a forgery as ancient. But he also stood in front of a block of marble and knew where to move his hand to pull the David out of it. The copying was training. The David was his.

I spent fifteen years pointing cameras at the real world making documentaries and video journalism. That training is how I direct AI. But directing is only half of it. The other half is thinking out loud.

I have too many ideas running at once. Films, courses, platforms, my father’s architecture archive. I dictate with Wispr Flow in a stream of consciousness, and Claude helps me see the structure inside the mess, find the link between two things I couldn’t connect on my own. With my recent film, Alma Robot, I needed to show a flashback that was visceral. My ideas were messy and Claude helped me articulate the visual language to get there. I knew the sound should carry the texture of Patagonian nature, but I had no musical vocabulary to describe what I was hearing in my head. Claude gave me the acoustic language. The vision was always mine.

That’s the circle. You arrive with everything you know. You use the tool to articulate what you can’t yet say. Then you break what it gives you and make it yours.

Lag Phase

Department of Gentle Futures is a piece of political theatre I built with AI. You’re dropped into a world where you file a vision for a softer future with a fictional government department before an efficiency auditor can optimize it out of existence.

The project draws from retrofuturism, disability justice, care ethics, punk rock, and my personal and professional experience with the social services system. Some of those influences I brought consciously, and others were revealed to me throughout the process. At the end of the build, I asked the AI what influences hadn’t been named (brb adding some Afrofuturist speculative fiction to my tbr list).

What made this mine was treating interaction design as dramaturgy. When you’re building worlds, worldview becomes the design material. It’s more useful than technical expertise and more durable than any individual prompt or AI model. I ran the scenario again and again, watching for moments where the world came together or fell apart, until it felt real.

When it felt complete, it almost felt too strange. Like I was letting people into a little piece of my world. That’s how I knew it was mine and time to share.

Joshua Sherman

I write and record songs under the name First Radio On. I’ve been in bands and played shows up and down the West Coast. Today, I use AI as a thinking partner for lyrics, music theory, harmonic analysis, arrangement, and song structure. I’ve never used a line AI wrote. But I’ve never finished a song without arguing about the song with AI first.

Two songs show what that actually looks like.

“Low HRV Alien” is about feeling like a different species in an optimization-obsessed world. I brought a half-finished draft. AI flagged things that it thought didn’t work. I pushed back. We went rounds on a single word in the opening line—”Cortisol conversations fill my mind”—because “fill” was dead. AI offered options. I picked “crush,” then later changed it to “crowd.” Neither was AI’s suggestion. Then came the chorus. AI said “syncopated” and “alien” were doing the same work and I should drop one. I said “we’re alien” doesn’t work grammatically. We went back and forth until I found the answer neither of us had proposed: drop the full title into the chorus as a lyric—”I’m a low HRV alien”—and save “syncopated” for the emotional resolution at the end. The song’s second verse and bridge, which mirror and invert the first (”I brace against the frame” becomes “You pull me to your side”; “I read a book that claimed an answer” becomes “I found that you were the answer”), I wrote entirely on my own.

“That’s a Long Time” was different. I brought the song nearly finished. It is a love song set against cosmic time, grounded in the walk from bathroom to bed, night after night. AI’s role was mostly to listen and confirm: yes, “wear a path through the floor” makes sense. Yes, “pink noise” makes sense. The biggest moment came from a typo. I’d written “too long for my mind” and meant to write “to long for your mind.” AI pointed out both readings were strong but framed it as a typo and that I should choose one. I kept both: “Too long for my mind / To long for your mind.” That wordplay came from my own mistake.

Those are two very different modes. One is a debate and discussion. The other is a sounding board. But the constant across both is that AI never writes. AI reacts, and I decide whether the reaction is useful. Half the time it isn’t. I’ve told the AI when it’s wrong, when it’s misreading the moment, when its feedback has become useless. That friction is the point. A good collaborator isn’t someone who agrees with you. It’s someone whose wrongness helps you find what’s right.

Beyond individual songs, I use AI to extract principles from our arguments. I’ve developed a songwriting checklist that encodes what is important to me. These are rules about when declarative lines open doors versus close them, why lyrics that read well on paper can fail when sung. I apply them without AI.

Calling AI a rhyming dictionary undersells it. But calling it a co-writer oversells it. I use it as a sparring partner. Its job is to be wrong in useful ways. Its best contribution is helping me see clearly enough to find the song that is inside me.

What this adds up to

Art has always borrowed. Michelangelo forged Roman sculpture. Shakespeare pillaged plots. The Stones lifted licks. Hip hop built a genre from other people’s records.

The difference between influence and theft was never about whether you borrowed. It was about whether you said so. And how you engaged in the craft of making the art itself.

Every contributor in this post uses AI differently. Some use it early. Some use it late. Some use it for planning and refuse it for execution. Some use it to challenge their own assumptions and then throw the output away. All of them are honest about what the tool is doing in their process.

That honesty is the line. Not a legal line. Not a copyright line. A creative one.

Behind that honesty lies something deeper, something that turns AI-generated art from a cheap, gimmicky product into a unique contribution to the world that can stand on its own.

That something is care. Every contributor in this article is not just using AI. They are practising with it. They are paying attention to what the tool does well and where it flattens. They are engaged with craft, not convenience.

And they are sharing what they learn. Not to perform transparency, but because creative practice has always been communal. Artists learn from each other by watching each other work. This post is an open studio, not a disclosure form.

If you can name what you borrowed, what you made, and what you cared about in the making, the work stands.

If you have a creative practice that involves AI, share it in the comments. Not the polished version. The real one. What do you borrow? What do you make? Where does AI fit, and what have you told it to stay out of?

This post exists because of Alyssa Fu Ward, PhD, who co-edited it and brought most of the contributors together. Thank you, Alyssa, as a community we appreciate everything you bring.

Go slow.