State of the Illustration Industry

K-Shaped Divergence and the Migration of Income Structure

A structural analysis based on industry surveys across multiple countries and on practitioner data.

— CreatorAris / ArisFusion Studio, April 2026

Core Conclusions

1. AI's impact on the illustration industry isn't an across-the-board decline—it's a K-shaped divergence. Illustrators with an IP and a fan base have pivoted toward paid knowledge products and fan economies, and their total income has in some cases surpassed the golden age of commissions. Mid- and long-tail illustrators with no accumulated IP, meanwhile, have lost both monetization channels—commissions and courses—at once.

2. The fork variable isn't drawing skill; it's "the ability to be seen." Skill is a necessary but not a sufficient condition. The decisive factor in where you end up is whether you built a personal brand and accumulated an audience before the AI shock arrived.

3. The market is extremely sensitive to "using AI but not admitting it"—and its detection accuracy is improving fast. Consumers aren't resisting the technology of AI itself; they're resisting the feeling of being deceived. This turns each work's "proof of origin" from a moral preference into a commercial asset—and it's the market foundation for Nephele's work on "verifiability of the creative process."

I. The AI Shock: Structural Contraction of the Commission Market

Since generative AI drawing tools went mainstream at scale in 2023, the global illustration industry has experienced a marked decline in income. The data below comes from several independent sources that cross-validate the reality of the trend.

Global Surveys

• The Society of Authors (SoA, UK), Survey on the Impact of Generative AI on Creative Professions (April 2024): the survey covered fiction/non-fiction authors, screenwriters, poets, journalists, illustrators, and translators. The data shows AI's impact on illustrators' work is significant, with most surveyed illustrators reporting reduced income or fewer opportunities.
• Japan's Freelance Association, income survey of manga artists and illustrators (launched October 2025, published January 2026): 24,991 questionnaires, 54.2% illustrators, 15% manga artists. The survey shows a broad-based decline in bargaining power in the commission market.
• World Economic Forum, Future of Jobs Report 2025: AI is displacing white-collar creative roles faster than expected.

Domestic (China) Data

According to Jingzhe Research Institute's survey of the Mihuashi platform:

• Single illustrations priced above ¥10,000 are now rarely seen.
• Chibi-style figures and low-priced simple drawings in the ¥50–100 range dominate the market.

A survey of illustrator income at Cijian Illustration:

A 2023 interview case: a freelance illustrator with 12 years of experience, who after quitting to go freelance in 2020 once earned ¥30,000–40,000 a month, had seen monthly income drop to under ¥10,000 by 2023. By their observation, the concept-art teams at large companies were cut from a headcount of 7–8 down to just 1–2.

How the Shock Propagates

The AI shock propagates "from low to high, from quantity to quality"—it first replaces entry-level execution work, then compresses the bargaining power of mid-level illustrators, and ultimately forces the entire industry to redefine the standard for "irreplaceability."

II. Illustrators Teaching Courses: A Counterintuitive Boom

Unlike the linear nature of commission income (you draw one piece, you get paid for one piece), online courses have a classic non-linear earnings structure: content is produced once and sold many times, with marginal cost approaching zero.

Taking Commissions vs. Teaching Courses: The Income Ceiling

For most freelance illustrators, the revenue from a single mid-sized course run can cover an entire year's commission income—even at the absolute peak of the industry, the price of a single top-tier freelance commission was only around ¥150,000, whereas one course run routinely brings in several hundred thousand. Industrialized salaried roles—art directors at major studios, top overseas cover illustrators—can command annual salaries far above that, but those are the tiny minority at the very top of the pyramid.

So why didn't freelance illustrators of the golden age shift en masse to teaching courses? It comes down to accessibility.

Why There Was No "Course Boom" in the Golden Age

Between 2015 and 2021, illustrators chose commissions over courses not because the income comparison was unfavorable, but because three prerequisites had not yet matured:

Missing infrastructure—WeChat Pay tipping, Bilibili paid courses, knowledge-payment mini-programs, and live interactive teaching tools either didn't exist before 2018 or were just getting started.

Missing personal brand—a large share of golden-age illustrators worked anonymously. Concept art for game companies was credited to the company; outsourcing clients didn't put the illustrator's name on the work. No fans, no personal brand, no social-media influence.

An insufficient demand pool—the narrative that "drawing can be a money-making side hustle" hadn't been manufactured at scale before 2018. What actually made the demand pool explode were two catalysts: the rise of the "side-hustle-as-necessity" and "slasher youth" narratives from 2018–2020, and the spread of the iPad and Procreate, which dramatically lowered the perceived barrier to drawing.

The course boom isn't "the training industry growing because the drawing industry shrank." It's mature platform infrastructure + a demand pool inflated by anxiety narratives + a simultaneous collapse in drawing income—three things that happened to converge in the same window of time.

III. K-Shaped Divergence: Who's Making More Money

The AI shock isn't an evenly distributed blow—it's a wedge that accelerates stratification.

The Upper Arm (the K's Upward Line)

Illustrators with an IP and a fan base have, in the AI era, actually gained more monetizable assets:

• The narratives of "anxiety courses," "transition methodologies," and "how I cope with AI" are themselves commodities.
• AI has also helped them lower the cost of producing content.
• Total income may double thanks to the course business.

The Lower Arm (the K's Downward Line)

For illustrators lacking an IP foundation, AI directly replaced their most central exchange value—execution-level drawing ability. As commission income collapsed, they had no ticket to the course market either (being "seen" and being "trusted").

The Fork Variable Isn't Skill—It's IP

IV. Monetizing the Narrative of Tragedy

There's a hidden narrative-arbitrage mechanism inside the K-shaped divergence:

The real pain of the industry's downturn is borne mainly by mid- and long-tail practitioners. But the public emotion that pain generates (anxiety, sympathy, a sense of crisis) is converted by top-tier practitioners into fuel for course sales.

The top illustrator says publicly: "The industry is hard; I've taken a hit too."

The first half of that statement is fact (commissions really have dropped). But their total income may have doubled thanks to the course business.

This framing has a dual effect:

• It sustains the "illustrators are suffering" narrative (making course buyers feel they need to save themselves).
• And it sustains the trust that "this illustrator is impressive and therefore worth learning from" (because they're still holding the line).

The narrative of tragedy has itself become a monetizable commercial asset.

This isn't a moral judgment. It's a market mechanism you need to see clearly—when you pay for "the industry is hard" content, what you're buying may not be a solution; it may be the anxiety itself.

V. The Structure of the Course Buyer Market

The narrative that "you can make a living learning to draw" is increasingly untenable, yet the course market remains active. The reason: the population buying courses has undergone a structural shift.

Of the four buyer types, only the "information-lagging entrants" make payment decisions positively correlated with the industry's real health. The other three are either unrelated or countercyclical.

This explains why the drawing industry is shrinking while the course market appears to be growing—the two serve different demand pools.

VI. Consumer Backlash: The Overlooked Second Chain of Transmission

The mainstream AI-crisis analysis describes a single economic chain of transmission: AI layoffs → consumers have no money → the market shrinks. That chain holds, but it isn't enough.

There's a second chain of transmission, and it doesn't have to wait for a wave of unemployment—it's happening right now.

Case One: Alchemy Stars Valentine's Day Greeting Art (Losing Face)

A work at the ceiling of gacha-game art. The Valentine's Day greeting art was caught by players showing obvious traces of AI drawing—not a conspiracy-theory accusation, but textbook AI artifacts visible to the naked eye. The community erupted on the spot, and the illustrator was thrust into the line of fire.

Case Two: Clair Obscur: Expedition 33 IGA Award Rescinded (Losing the Award)

On December 18, 2025, the Indie Game Awards (IGA) gave its "Game of the Year" and "Best Debut" awards to French studio Sandfall Interactive's Clair Obscur: Expedition 33. Two days later, the organizers announced both awards were being rescinded—at nomination time, the developer had declared no generative AI was used, but an earlier interview resurfaced, and pre-launch placeholder-texture details dug up by players, confirmed the presence of AI material. Sandfall explained these were "temporary placeholder textures, replaced within 5 days of launch," but the organizers upheld the decision to rescind.

Case Three: Postal: Bullet Paradise (Losing the Studio)

On December 3, 2025, Running With Scissors, the IP holder of the POSTAL series, announced a trailer for the new title Postal: Bullet Paradise. Within 24 hours, players had dug out traces of AI material in the trailer and resurfaced the publisher's own past public statements opposing the use of AI. The project was canceled that same day. A few days later, the development studio Goonswarm Games announced its closure, leaving 9 developers out of work. The studio initially denied it, then admitted the promotional trailer material had used generative AI.

The Statistical Pattern

Quantic Foundry's global player survey covering October–December 2025 (published December 18, 2025):

• 85% of players hold a negative attitude toward the use of generative AI in games.
• Of those, 62% are "very negative."
• Only 7.6% expressed approval.

This isn't the sentiment of a small minority—it's an overwhelming consumer will.

The Two Chains Side by Side

The mainstream chain is "no money, so they don't buy." The chain I see is "they saw through it, so they don't buy."

Consumers aren't blind. They can smell it. And they already are.

VII. The Impact of Piracy on the Course Economy

The course market faces a serious piracy problem. According to Rights Knight's Content Industry Copyright Report:

• On platforms like Xianyu, Baidu Tieba, Zhuanzhuan, and Taobao, infringement leads for paid content each number in the tens of thousands.
• Pinduoduo's copyright-governance rate is only 88.8%, the lowest among comparable e-commerce platforms.
• Online courses originally priced at thousands or even tens of thousands of yuan can be bought on these platforms for ¥5–10.

But piracy hasn't destroyed the course business model. The reason: the core selling point of top illustrators' courses isn't the recorded content itself, but three things piracy can't provide:

1. Real-time interaction and homework feedback (live courses)
2. A sense of community belonging (joining a group to talk with the teacher and fellow students)
3. Identity affirmation conferred by the illustrator's personal IP ("I'm a student of Teacher So-and-so")

What piracy mainly siphons off are price-sensitive users, who were unlikely to become full-price paying customers in the first place.

The ones truly hit hard by piracy are instead the pre-recorded courses of small and mid-sized training organizations and individual illustrators—products that lack community service and IP backing, where the content itself is the entire value; once it's recorded and circulated, it loses its competitiveness.

This further intensifies the K-shaped divergence: top-tier courses with an IP moat can resist piracy's erosion, while tail-end courses with no IP protection have almost no defense against it.

VIII. Conclusion: Why This Report Matters

Three core conclusions:

1. The AI shock didn't create an across-the-board decline in the illustration industry—it created a K-shaped divergence. Illustrators with an IP and a fan base pivoted toward paid knowledge products such as courses, and their total income actually surpassed the golden age of commissions. Mid- and long-tail illustrators without an IP foundation lost both monetization channels—commission income and course income—at once.

2. The essence of the course boom is neither "a reluctant move forced by a bad industry" nor "a more lucrative new opportunity." It's the result of three independent variables—mature platform infrastructure, anxiety-driven expansion of the demand pool, and the collapse of commission income—converging in the same window of time.

3. The fork variable in the K-shaped divergence isn't drawing skill—it's "the ability to be seen." Skill is a necessary but not a sufficient condition. The decisive factor in whether you can maintain or even raise your income is whether you built a personal brand and accumulated an audience before the shock arrived.

What Does This Have to Do with Nephele?

After reading the report above, one conclusion surfaces on its own: being "seen" only solves half the problem of the divergence.

The other half is—once you've been seen, how do you prove that the image people saw was drawn by you?

This is the real dividing line between top-tier and mid-tier illustrators. A top illustrator's moat was never the technique; it's the kind of recognition that needs no defense—"this image could only have been drawn by Teacher So-and-so." That recognition holds automatically through years of accumulated fans, and no one bothers to question it. Mid-tier illustrators can't sustain that recognition—the only thing they can rely on is evidence that can be independently verified by a third party.

That dividing line is being accelerated by the data in Section VI: consumer sensitivity to AI material is rising fast, and that same sensitivity hangs over the head of every illustrator who didn't use AI—it isn't aimed only at the studios that did. Once "is this AI-generated?" becomes the first filter a work is judged through, simply saying "I didn't use it" is no longer enough to convince anyone. The pricing logic for "being able to draw" has changed; whether you can prove "this image came from your hand" has become the new fork variable.

What Nephele does is make that recognition hold at the level of every single image—no longer something only top illustrators can sustain, but a way for any single image to carry a piece of creative evidence that can be independently verified. The argument behind this (why the old aura died, why the new aura needs infrastructure) is written up in Design Philosophy; the specifics of how it's achieved (the engineering boundaries of digital provenance, invisible watermarking, and rights forensics) are written up in Security & Provenance.

It's not marketing copy—it's the only logically sound conclusion left standing after reading the seven sections above.

Appendix: Primary Data Sources

1. The Society of Authors (SoA), Survey on the Impact of Generative AI on Creative Professions, April 2024.
2. Japan's Freelance Association, income survey report of manga artists and illustrators, launched October 2025, published January 2026. 24,991 questionnaires.
3. World Economic Forum, Future of Jobs Report 2025, January 2025.
4. Sina Finance, "From a ¥700K Salary to a Few Thousand! The First Humans Replaced by AI Have Arrived," January 2026. Based on global labor-market data from 2023–2025.
5. Woshipm (Everyone Is a Product Manager), Survey Report on the Survival Conditions of Chinese Internet Designers Under the AI Wave, April 2025.
6. The Paper / Sina Tech, "AI Heat Investigation: Turmoil in the Illustration Industry," interview series, April 2023.
7. Jingzhe Research Institute / 21jingji.com, "How Did Learning Illustration from Scratch Become a Giant Intelligence Tax?", February 2024. Includes the Mihuashi platform survey and interviews with several illustrators.
8. Sina Tech, "Earning ¥175M in Three Years—Selling Courses Is the Easiest Way to Make Money in Chinese AI," April 2025. Includes the Li Yizhou case and Douyin course-sales data.
9. Rights Knight / Whale Copyright, Content Industry Copyright Report (2019–2020).
10. Citrini Research, 2028 Global Intelligence Crisis, February 2026.
11. Quantic Foundry, Gamers Are Overwhelmingly Negative About Gen AI in Video Games, December 18, 2025. Global player survey, sample covering October–December 2025.
12. Coverage of the IGA 2025 rescission of Clair Obscur: Expedition 33's awards by outlets including AV Club / Kotaku / VGC / Insider Gaming, December 2025.
13. Coverage of the cancellation of Postal: Bullet Paradise and the closure of Goonswarm Games by outlets including GameSpot / VGC / Push Square / Decrypt, December 2025.

— CreatorAris

This report will be updated as new data comes in. If you have industry data to contribute, email [email protected].