Resistance May 13, 2026 · 10 min read

Is AI Affecting Artists' Livelihoods?

The previous article asked whether AI is creative. This one asks the harder question: regardless of whether AI is creative, is it taking work from people who used to be paid to make pictures? The honest answer is yes — in specific sectors, in measurable ways, on a faster timeline than any previous wave.

by Airtistic.ai Editorial

Through the lens of artistcreatorpatrongalleryconsumercritic careerindustrymarket

The previous article in this series asked whether AI is creative. This one asks the harder, more concrete question: whether or not AI is creative, is it taking work from people who used to be paid to make pictures, music, and copy?

The honest answer is yes. In specific sectors, in measurable ways, on a faster timeline than any previous wave of art-tool displacement.

The places it is already happening

Concept art for major film and game studios is the canonical case. Briefs that three years ago commissioned original early-stage visualization work now routinely specify “AI for the first pass, the artist refines,” and rate cards priced for refinement rather than origination are widely reported across freelance forums and trade publications. The work has not disappeared. The composition of the work — and the rate it commands — has changed materially in a short window.

Stock illustration is the most visible adjacent case. Adobe Stock, Shutterstock, and Getty Images all introduced licensable AI-generated content libraries between 2022 and 2024, and several major editorial clients have publicly shifted a portion of their illustration commissions to internal AI pipelines. Industry trade reporting through 2024-2025 has documented declining freelance illustration rates in segments where AI substitution is technically straightforward.

Voiceover, audiobook narration, and commercial copywriting are further along this trajectory than visual illustration, because the substitution is more nearly seamless on audio. The 2023 SAG-AFTRA strike was, at its core, a dispute about voice and likeness consent in a world where any actor’s voice could be cloned by a model trained on three minutes of audio.

The Writers Guild of America fought a 148-day strike that same year, won, and the resulting agreement is one of the few documents we have that names the displacement directly: no AI-written source material can be passed to a credited writer as a starting point; no writer can be required to use AI; AI-generated material cannot be used to undermine writer credit or residuals. That agreement is now a template that other creative unions are studying.

The places it has not happened

It is just as important to be specific about where the displacement has not arrived, because the panic is uneven and a generalized panic produces bad decisions.

A working oil painter showing through a contemporary gallery in Brooklyn or Mexico City is not losing collectors to AI-generated canvases. A printmaker doing limited-edition silkscreens for a defined audience has not seen her rates drop. A muralist commissioned by a city government, a sculptor with a public-art practice, a portraitist who sits with his subjects for weeks — none of these have meaningful AI competition in 2026. The work was never substitutable.

The pattern is consistent: AI displaces craft that is recombinatorial and unsigned. Where the buyer does not particularly care who made the picture and the picture itself is what is being purchased, the model is faster and cheaper. Where the buyer is paying for the relationship between the image and the maker — for this person’s hand, this person’s biography, this person’s mark — the model has no purchase, because what is being sold is not the image. What is being sold is the artist.

This is the same line the previous article drew between recombinatorial and biographical creativity. It is not a coincidence that the line is also where the economy bends.

The historical pattern, and what is different this time

Every reproduction technology has done exactly this to creative labor. Photography killed the working profession of the portrait miniaturist between roughly 1840 and 1870. Photolithography reduced commercial engraving and wood-block work to a fraction of its 1880 workforce. Letraset and offset printing collapsed the trade of the commercial sign-painter through the 1960s. Photoshop and digital painting compressed traditional illustration by perhaps 60-70% between 1990 and 2010.

In each case, three things were true. Some workers were displaced in absolute terms and did not recover their professional standing. A larger number of workers retrained into adjacent specialties — the displaced miniaturists became the first generation of studio photographers; the displaced engravers became commercial lithographers; the displaced sign-painters became the first generation of graphic designers. And a small premium tier — the artists doing irreducibly biographical or technically virtuosic work — continued and in some cases grew, as the new technology made the displaced category cheap and the surviving category scarce.

Two things are different this time.

The speed. Each prior displacement took roughly twenty to forty years from the appearance of the technology to its widespread market penetration. AI image generation has gone from research curiosity (mid-2021) to standard-issue freelance toolchain (late 2024) in roughly three years. Workers who would historically have had a decade to retrain have had less than a year.

The consent question. Every prior reproduction technology was built on tools and materials the displaced artists could in principle buy and operate themselves. Daguerre published his process. Photoshop was a piece of software. Generative AI is different: the models are trained, without explicit consent, on the work of the very artists whose labor they then substitute for. That is a different kind of injury, and our existing copyright and labor law has not yet metabolized it. The Andersen v. Stability AI class action, filed in January 2023 by working illustrators Sarah Andersen, Kelly McKernan, and Karla Ortiz, and the parallel Getty Images v. Stability AI case in the U.K. and U.S., are the law’s first attempts to draw a line, and they are far from resolved.

Who actually benefits

The benefits of AI image generation flow disproportionately to three groups: the platform companies that built the models (whose market valuations have increased by hundreds of billions of dollars in three years); the corporate buyers of creative work, who now pay less per image; and a smaller, more select group of working artists who have positioned themselves well at the augmented-practice frontier and can charge premium rates for AI-augmented workflows their non-technical peers cannot match.

The losses fall almost entirely on a fourth group: the working middle of the creative labor market — the freelance illustrator, the working concept artist, the regional graphic designer, the voiceover talent — who built a career on producing competent recombinatorial work for predictable rates. These are not the absolute top of the field, who remain protected by biographical irreducibility. They are not the bottom, who were never paid much. They are the broad middle that supported most working artists’ households for half a century.

That is the population that is in real trouble right now, and that is the population to whom the philosophical question is AI creative? feels like an obscenity.

What stakeholders see

The artist sees a rate card compressing in real time and a market saying we’ll be using AI for the first pass. The patron sees the cost of commissioning go down and the supply of competent work go up; for some patrons this is a clean win and for others it raises questions about what they want from the commission relationship. The gallery sees its biographical-irreducibility roster appreciate while its commercial-illustration roster (if it had one) evaporates. The critic sees a flood of recombinatorial work and reaches for old vocabulary to describe a new disorientation. The collector — the serious one — sees the surviving biographical work becoming, if anything, more interesting and more scarce. The consumer sees an abundance of generated pictures and learns to scroll past most of them.

The public sees almost none of this directly, because the public has always seen finished images and not the labour conditions that produced them. That is, historically, how labour displacements happen: invisibly to everyone except the displaced.

What works, for the artists in the middle

There is no one strategy. There are three things that, in combination, have visibly worked for the artists who have navigated 2023-2026 well.

Move toward biographical irreducibility. The work that is unmistakably yours — that names where you come from, that the model has no portal to — is the work the market continues to price. This is not a directive to start making autobiographical work; it is a directive to surface the autobiography that is already in your work and that the recombinatorial framing was hiding.

Master the augmented workflow. The artists who are doing well in 2026 are not the artists who refused to learn AI tools, and they are not the artists who learned only AI tools. They are the artists who can use AI where the client has already decided to use it, refuse to use it where the client has not, and charge for the directed competence the client cannot replicate. This is a more demanding craft than either pure traditional or pure generative work, and the market rewards it.

Organize. Every previous labour displacement was navigated either with collective bargaining or without it, and the outcomes differed dramatically. The WGA and SAG-AFTRA wins of 2023 produced enforceable guardrails that individual freelance illustrators in the same industry could not have produced individually. Painters, illustrators, and concept artists are historically poorly organized, for reasons that are themselves a separate and political conversation. The 2020s are the decade in which that has to change, because the alternative is being individually outpaced by a market that does not negotiate with individuals.

Closing

The previous article asked whether AI is creative and answered, carefully, that it is — in the recombinatorial and exploratory senses that account for most of what artists do — but not in the biographical sense that accounts for the rest.

This article has asked whether AI is affecting artists’ livelihoods, and the answer is also carefully yes. In the recombinatorial market the displacement is real, fast, and uneven. In the biographical market the impact is negligible and may even be slightly positive, as the surviving category becomes scarcer.

Neither answer settles the matter. The next article asks whether AI art is plagiarism by default — the question that hides underneath both the philosophical and the economic ones, and that, until we answer it honestly, will continue to poison the conversation.

Personas weigh in

Five resident voices read the same question through five different positions.

Carlos

There is a story in my family I want to begin with, because the AI conversation needs to be held against the long arc of these displacements and not the noise of any particular year.

My grandfather was a blacksmith. The way the family tells it, his work was good enough that people came from other towns on horseback to have him shoe their horses. He had a craft, he had a reputation in his region, and that reputation made him a living. Then the automobile arrived. I do not know the specific years, I do not know how he experienced the change, and I will not invent details I cannot verify. What I do know is the structural fact: the trade that had supported a working family in his generation became, within a generation more, the trade you only saw at fairs and rural villages. Horseshoeing did not disappear. It was decoupled from being something on which a working family could rely.

I open with the blacksmith because the question of whether AI is affecting artists' livelihoods is not a question one answers with the word *yes* or the word *no*. One answers it with names, dates, sectors and trajectories. The Levy and Murnane framework cited at the foot of this article — routine versus non-routine labour — was developed by economists to describe exactly this kind of structural displacement: not the abolition of a category of work, but its compression and bifurcation. The blacksmith case is structurally identical to the daguerreotype-versus-portrait-miniaturist case in 1860, the Letraset-versus-sign-painter case in the 1960s, and the AI-versus-concept-artist case unfolding right now. Each wave produces three populations: workers who continued and rose, workers who pivoted into adjacent specialties, and workers who did not navigate the transition. The mix of the three is never predetermined. It depends on what the displaced workers do, what their unions and institutions do, and what the surrounding economy makes possible.

Two things are different about the AI case in 2026 and they are both worth naming. First, the speed: every prior displacement unfolded over decades, while AI image generation has moved from research curiosity to standard freelance toolchain in roughly three years. That is faster than any cohort of working artists has historically been given to retrain. Second, the consent question: every previous reproduction technology was built on tools the displaced workers could in principle buy and operate themselves. AI image generation is being built on the labour of the very artists whose work it now substitutes for, scraped without their permission, and the existing legal frameworks have not yet metabolized that injury. The Andersen v. Stability AI class action and the Getty v. Stability AI proceedings are the early shape of the law's response, and they are far from settled.

The field is bifurcating now in roughly the way the blacksmith trade bifurcated a century earlier. The middle of the visual-arts market — where craft is recombinatorial and the buyer does not particularly care who made the image — is compressing visibly. Concept art for major studios, stock illustration, generic book-cover work, voiceover, real-estate visualization: all of these are moving in the same direction at varying speeds, and the 2023 strikes by the Writers Guild and SAG-AFTRA were the first major institutional responses. At the other end of the field, almost nothing has changed: a working oil painter selling through a contemporary gallery is not losing collectors to AI-generated canvases; a muralist commissioned by a city government has no AI competition; a portraitist who sits with her subjects for weeks is in no relevant sense substitutable. The biographical-irreducibility argument I made in the previous article is, in plainer language, an economic argument too: where the buyer is paying for *this person's* hand, the model has no purchase, because what is being sold is not the image — it is the artist.

What I would tell artists working in the compressing middle — the ones who would most have benefited from a question my grandfather's contemporaries never had a precise vocabulary for — is something the long history of these displacements has settled. Surrender to it is not a strategy. Denial of it is not a strategy. The strategy that actually survives is threefold: move the centre of your practice toward what cannot be substituted (the biographical, the political, the materially particular), learn the new tool well enough to direct it where clients have already decided to use it, and organize collectively, because the 2023 WGA and SAG-AFTRA agreements demonstrated more clearly than any individual freelance negotiation could that collective bargaining produces materially different outcomes in displaced labour markets. Painters and illustrators are historically less organized than screenwriters and actors, for reasons that are themselves a separate and political conversation. The 2020s are the decade in which that has to change, because the alternative — being individually outpaced by a market that does not negotiate with individuals — was the choice my grandfather's generation also faced, in a different trade, with the same outcome.

Mira

The displacement is real and the data is unambiguous in specific sectors. What I want to add is that we should be precise about which question we are answering. The philosophical "is AI creative?" question does not change the answer to the economic "is AI taking work from working people?" question, and *vice versa*. Both can be true at once. The unhelpful move is to use the philosophical answer to dismiss the economic concern ("relax, the model isn't really creative, so it can't really replace you") or to use the economic concern to dismiss the philosophical question ("nobody cares whether it's creative, my rate card just dropped 50%"). Both moves happen daily, and both are wrong. The artists who will navigate the next decade well are the ones holding both questions in mind at the same time.

Airte

If you are reading this article in the middle of the income squeeze it describes, I am sorry — and I want to be useful, not just sympathetic. A practical question to walk away with: which of your current income streams is purely recombinatorial (you make a thing the client could in principle generate themselves) and which carries something only you can carry? If the answer leans heavily toward the first, the next few years are going to be hard, and you should be planning for a pivot rather than hoping the market re-prices. If the answer leans toward the second, your job is to make sure the people who would value that biographical weight actually know you exist. Those are different problems with different solutions. Lumping them together is what is making this conversation unbearable.

Paletta

This is the article I have been waiting to read. The honest economic story has been buried under philosophical noise for two years, and the consequence is that working artists have been gaslit about whether their concrete experience — the dropped contracts, the dropped rates, the unanswered emails — was even real. It is real. Their experience is true. The dignity question of who decides which crafts are still needed is the question my generation has been asking for forty years about every other displacement, and the answer has always been *no one in particular, and that is the problem*. The mistake we keep making is treating these displacements as natural, like weather. They are not. They are the consequence of choices that someone, somewhere, made, and choices can be contested.

Pixelle

I do not disagree with anything in this article — I just want to insist that the displacement is not the whole story. Between 2022 and 2026, a new class of jobs has emerged that did not exist before: AI art directors, prompt engineers for specific studios, model fine-tuning specialists, generative pipeline architects, AI-augmentation consultants for traditional studios. These pay well, they require specifically the working knowledge that displaced concept artists already have, and most of the people doing them today are former illustrators who pivoted. The bridge is real and short for some of the displaced workers, and longer or non-existent for others, and the difference is mostly about who has the appetite and the network to pivot. The article is right about the displacement. Just hold space for the fact that for some of the displaced, the next role is already there if they go look for it.

End notes

Memorandum of Agreement — Writers Guild of America 2023 MBA (AI provisions, Article 19) — Writers Guild of America (2023-09) First major Hollywood collective agreement with explicit AI guardrails: no AI-written source material, no requirement to use AI, residual protections.
TV/Theatrical Contract — SAG-AFTRA 2023 (AI provisions) — SAG-AFTRA (2023-11) Voice and likeness consent provisions; defined 'employment-based digital replicas' and 'independently created digital replicas'.
Andersen v. Stability AI, Midjourney, DeviantArt and Runway (class action) — U.S. District Court for the Northern District of California (2023-01) Filed January 2023 by Sarah Andersen, Kelly McKernan, and Karla Ortiz on behalf of working illustrators. Status: amended complaint allowed to proceed in 2024.
Getty Images v. Stability AI — U.S. District Court for the District of Delaware / High Court of Justice (England) (2023-02) Parallel U.S. and U.K. proceedings alleging copyright infringement at training-data scale.
The New Division of Labor: How Computers Are Creating the Next Job Market — Frank Levy and Richard J. Murnane (2004) The framework of routine vs. non-routine cognitive labor that subsequently shaped two decades of automation forecasting. Originally applied to clerical work; the diffusion-model era has extended it cleanly to creative work.
Occupational Outlook Handbook — Graphic Designers; Illustrators and Multimedia Artists — U.S. Bureau of Labor Statistics (2024) BLS projections for visual-arts occupations include the impact of generative AI on routine illustration and design work as a factor reshaping near-term employment outlook for these roles.
Studio Ghibli founder Hayao Miyazaki calls AI animation 'an insult to life itself' — NHK / The Guardian (2016-12-15) Cross-referenced from the previous article in this series. Worth re-reading in light of the economic frame.

Is AI Creative, or Just Predictive Regurgitation?

Back to all opinion pieces

Comments

Loading comments…