BBC: AI Is Devaluing Images and Raising the Price of Human-Made Work

BBC put in its headline a thesis the art market has debated for a while: the cheaper a model-generated image becomes, the more a work by a known human author is worth. The piece reached Hacker News, where it collected 60 points and 78 comments. For an art topic in a technical community, that is a notable discussion. A caveat: all we have is the link and the headline, not the full text. So what follows examines the idea itself and the context around it, not BBC quotes or data.

Where the premium comes from

The economics are simple. Generating an image costs a few seconds of GPU time. When supply is practically unlimited, the price falls toward the cost of compute. What stays scarce is what does not scale: the creator's time, biography, reputation and risk of failure.

This has happened before. Industrial production made furniture and tableware cheap, and handmade goods became a premium segment. Photography did not kill painting, but redefined its role. Factory printing did not eliminate the market for original prints. Generative models are repeating this pattern at a far faster pace.

The market is already splitting

The first signals appeared several years ago. In 2018, Christie's sold "Edmond de Belamy," a work created by a generative network, for $432 500. That was a novelty act: buyers paid for the fact that the technology had arrived, not for the quality of the image. In 2022, Jason Allen won the Colorado State Fair competition with a painting made in Midjourney, and the scandal showed how sharply the community reacts to blurred authorship.

The legal backdrop also works in favor of the "human." In the US, works created by AI without substantial human contribution do not receive copyright. The US Court of Appeals for the D.C. Circuit confirmed this position in 2025 in Thaler v. Perlmutter. This matters for commercial use: an image with no rights holder is hard to protect and hard to monetize as an exclusive.

Publishers and guilds are responding with their own labels. In 2025, the Authors Guild launched Human Authored certification for books. The logic is the same as with labeling "organic" food: the buyer pays for a guarantee of origin.

Key facts

• Source: BBC News, discussion on Hacker News — 60 points, 78 comments.
• Thesis: cheaper AI content raises the price of verifiably human authorship.
• Precedent: "Edmond de Belamy," Christie's, 2018, $432 500.
• Law: under Thaler v. Perlmutter, purely machine-made works do not receive copyright in the US.
• Bottleneck: there is no single way to reliably prove that a human made a work.

The proof problem

The main question is how to verify "made by a human." The C2PA standard and Content Credentials record a file's provenance and edit history, but support is uneven and the metadata can be stripped. Process videos, drafts and source layers serve as evidence, but those too can be faked by anyone who wants to.

Hybrids add a separate complication. Many artists use AI for sketches, backgrounds or retouching. Where the line for a "human" work falls, nobody has yet defined. Without a standard, the human-made label quickly turns into a marketing slogan, and the premium dissolves.

What to do in practice

Artists and studios should document their process: keep layers, drafts and timelapses, and, where possible, sign files with Content Credentials. This is cheap insurance and a future argument for a higher price.

Platforms and publishers need transparent rules: what counts as AI-assisted, what counts as fully generated, and how an author's claim is verified. Buyers and clients are better off stating in the contract whether generative models were used, because both the rights to the work and its market value depend on it. And those betting on AI illustration should start calculating the economics without a premium: mass content will be cheap, and they will have to compete on speed and volume, not uniqueness.