30 AI art statistics on artists and collectors
The most defensible AI-art statistics come from two 2023–2024 surveys with clearly described populations. DACS surveyed 1,000 artists and artists’ representatives; Hiscox and ArtTactic surveyed 210 art collectors and 243 art enthusiasts. Their results show strong artist concern about training, consent and compensation, while collector and enthusiast attitudes toward buying AI art differed sharply.
These figures are historical snapshots, not a measurement of the 2026 public or the current AI-image market. Each statistic below keeps its source population attached so it can be cited without turning a subgroup result into a universal claim.
Artists and representatives: 95% in the DACS survey felt they should be asked, 93% credited and 94% compensated when their work is used to train AI.
Training uncertainty: 63% did not know whether their work had been used for AI training; 22% identified that it had.
Skills gap: 96% reported no AI training or education, while 31% saw missing skills or training as a barrier.
Collector adoption: 2% of Hiscox’s collector sample had bought AI-generated art, compared with 28% of its broader art-enthusiast sample.
Transparency: 82% of collectors and 76% of enthusiasts wanted to distinguish AI-generated from human-created art.
Methodology and citation limits
DACS’s January 2024 report covers a SurveyMonkey survey distributed through DACS channels from September 8 to October 16, 2023. Exactly 1,000 artists and artists’ representatives responded, and 352 left comments. DACS is a UK visual-artists’ rights organization; this is not described as a random sample of all artists.
Hiscox’s Art and AI Report 2024 covers two ArtTactic surveys conducted from April through June 2024: 210 active art collectors from ArtTactic’s collector sample and a purchased sample of 243 U.S. and European art enthusiasts ages 18–65 who reported interest in art or a prior art purchase.
Do not combine the denominators. Artists, artist representatives, active collectors and broadly defined enthusiasts answer different questions.
Do not call stated intent a purchase forecast. “Might consider” and “believes demand will rise” are opinions, not observed future transactions.
Do not treat old model names as the market. The surveys reflect tools, media coverage and policy debates present during their field dates.
Do not infer causality. Neither study tests whether AI caused attitudes, purchases, job loss or market movement.
Use the original report. Link the report, field date, population and question whenever you reuse a number.
15 statistics from artists and artist representatives
# | DACS survey result | Respondents |
|---|---|---|
1 | Concerned their work was used to train AI models | 74% |
2 | Identified that their work had been used for AI training | 22% |
3 | Did not know whether their work had been used for training | 63% |
4 | Of those surveyed, said they had not given permission for AI training on their work | 96% |
5 | Felt they should be asked when their work is used for AI training | 95% |
6 | Felt they should be credited | 93% |
7 | Felt they should be compensated | 94% |
8 | Would join a consent-based collective licensing mechanism | 84% |
9 | Saw legal and ethical concerns as a barrier to using AI | 64% |
10 | Were concerned about their artistic style being mimicked | 69% |
11 | Agreed or strongly agreed AI could replace jobs and opportunities | 77% |
12 | Agreed or strongly agreed AI could create new opportunities | 31% |
13 | Had received no AI training or education | 96% |
14 | Saw lack of skills or training as a barrier | 31% |
15 | Felt the UK should introduce safeguards and regulation for safer AI | 89% |
The DACS results contain concern and conditional openness at the same time. Seventy-seven percent agreed AI could replace jobs and opportunities, while 31% agreed it could create new opportunities. Eighty-four percent said they would join the proposed licensing mechanism when use occurred with consent.
15 statistics from art collectors and enthusiasts
# | Hiscox and ArtTactic survey result | Respondents |
|---|---|---|
16 | Art collectors who had bought AI-generated art | 2% |
17 | Art collectors who might consider buying it | 29% |
18 | New collectors who had bought it | 7% |
19 | New collectors who might consider buying it | 39% |
20 | Art enthusiasts who had bought it | 28% |
21 | Art enthusiasts who might consider buying it | 52% |
22 | Collectors excited about the evolution of AI | 33% |
23 | Art enthusiasts excited about it | 71% |
24 | New collectors excited about it | 54% |
25 | Collectors concerned about it | 41% |
26 | Art enthusiasts concerned about it | 16% |
27 | Collectors who wanted to distinguish AI-generated from human-created art | 82% |
28 | Art enthusiasts who wanted that distinction | 76% |
29 | Collectors concerned about authenticity and originality | 61% |
30 | Collectors concerned about lack of emotional connection | 60% |
The Hiscox report separates active collectors from a purchased sample of art enthusiasts. That distinction explains why 2% and 28% can both be correct: they describe different groups. It does not prove that enthusiasts will become investment collectors or that either percentage has persisted since 2024.
What the statistics say about trust and provenance
Across the two studies, transparency is the most consistent operational theme. DACS respondents wanted consent, credit and compensation around training; Hiscox respondents wanted to distinguish AI-generated art from human-created work. For creators and buyers, useful provenance can include the source asset, tool or model, prompt, significant human edits, date and permissions.
In the United States, the Copyright Office’s January 2025 copyrightability report concluded that generative-AI output can be protected only where a human author determined sufficient expressive elements. Human-authored material, creative arrangement or modifications can qualify; merely providing prompts does not. This is legal guidance for U.S. copyright registration, not a statistic and not universal legal advice.
How creators can use this research
Document authorship: save source files, prompts, selections, edits and approvals instead of relying on the exported image alone.
Choose an original brief: specify subject, composition, light, materials and intended use without asking to imitate a living artist.
Check the output: review logos, people, text, product details, cultural references and accidental resemblance before commercial use.
Preserve disclosure: label AI involvement where the platform, client, contract or jurisdiction requires it.
Measure the actual job: compare revision time, approval rate, usable outputs and campaign results rather than total generations.
Run a documented AI-art test
Create one original visual from a defined brief. Save the prompt and edits, check rights and provenance, then compare the result with your existing workflow.
Create AI ArtFrequently asked questions
In DACS’s 2023 survey of 1,000 artists and artists’ representatives, 74% were concerned about their work being used to train AI models. The result should be cited with that population and field date.
In the Hiscox and ArtTactic 2024 survey, 2% of 210 active art collectors said they had bought AI-generated art. In its separate 243-person art-enthusiast sample, 28% said they had. The groups are not interchangeable.
In the DACS survey, 84% said they would join a consent-based collective licensing mechanism to be paid when their work is used by AI. That is support for the described mechanism among the survey respondents, not a global artist estimate.
The U.S. Copyright Office says protection depends on sufficient human-authored expressive elements. AI assistance does not disqualify an otherwise human-authored work, but prompting alone does not make the machine-determined output copyrightable.
Start with the AI art generator for a bounded original brief, browse AI art styles for visual vocabulary, or compare the best AI illustration tools by workflow when production requirements drive the choice.
