In 2022, Boris Eldagsen’s AI-generated image “The Electrician” won the Sony World Photography Awards. He entered it deliberately — to test whether the distinction between photography and AI art still existed. The jury did not detect the difference until he refused the award and explained.

The incident became a Rorschach test. Photographers saw catastrophe. AI advocates saw proof of equivalence. Neither response addressed the harder question: what, exactly, is being lost — or gained — when machines generate images indistinguishable from photographs?

Defining the terms (before they collapse)

Photography — from the Greek photos (light) and graphein (to draw) — literally “drawing with light.” A camera records photons reflected from physical reality. Even heavily processed digital photography originates in a moment of light hitting a sensor.

AI-generated imagery — neural networks trained on millions of existing images produce new images from text prompts or reference inputs. No photons are involved. No moment occurred. No light was drawn.

The output can look identical. The process could not be more different.

What AI does better than photography

Impossible scenes — historical figures in contemporary settings, architectural fantasies, surreal compositions that would require impossible budgets to photograph.

Speed of iteration — hundreds of visual concepts in minutes. For mood boards, pre-visualization, and creative exploration, this is genuinely revolutionary.

Democratization of visual creation — people without cameras, studios, or technical training can produce images that communicate visually. This is not trivial. Visual literacy should not require equipment.

Style synthesis — combining aesthetic traditions that no single photographer could merge. “In the style of Vermeer, but cyberpunk, photographed at golden hour” produces something new from something old.

What photography does that AI cannot

Indexical truth — a photograph, at its core, certifies that something was in front of a lens at a specific time. This evidentiary quality — however manipulated — grounds photography in reality. AI images certify nothing except the existence of a training dataset.

The decisive moment — Cartier-Bresson’s famous concept: the photographer recognizes and captures a convergence of elements that exists for fractions of a second. AI generates moments that never happened. The difference is not aesthetic. It is ontological.

Relationship with subject — portrait photography requires negotiation, trust, vulnerability. The subject and photographer co-create the image. AI generates faces without persons behind them — a fact with profound ethical implications for representation and consent.

Physical presence — standing in a place, adjusting to its light, waiting for conditions to align. Photography is a embodied practice. AI generation is a sedentary one. The body of the photographer is part of the image, even when invisible.

Imperfection as signature — grain, motion blur, lens flare, the specific failure modes of optical systems. These are not bugs in photography. They are the medium’s handwriting. AI can simulate them but the simulation lacks the history.

The competition problem

When AI images enter photography competitions, the issue is categorical, not qualitative. It is like entering a novel in a poetry contest because both use words. The skill sets, processes, and relationships to reality are fundamentally different.

Proposed solutions gaining traction:

None of these resolve the deeper cultural question: if viewers cannot tell the difference, does the difference matter?

The case that it matters enormously

Journalism — a news photograph certifies that an event occurred. AI-generated “photographs” of events that did not happen are misinformation by design. The distinction between photography and AI art is a distinction between truth and fabrication in the context most vulnerable to manipulation.

Historical record — photographs document the 20th century. AI images document nothing except the prompts that generated them. Future historians will need tools to distinguish between images of events and images of imaginations.

Labor and craft — photographers who spent decades developing eye, technique, and relationship with light compete against systems that synthesize the appearance of that development in seconds. The economic impact on working photographers is real and largely unaddressed.

Consent and representation — AI can generate images of people who do not exist but look real, and images of real people in situations they never experienced. Both scenarios create ethical crises that photography’s relationship with reality at least makes visible.

The case for coexistence

AI-generated imagery is a new medium, not a replacement for an old one. Watercolor did not eliminate oil painting. Photography did not eliminate painting. Video did not eliminate photography.

AI art will find its domains:

Photography will retain its domains:

The danger is not AI art. The danger is category collapse — the lazy assumption that because outputs look similar, processes are equivalent.

What photographers should do now

  1. Disclose your process — transparency builds trust
  2. Embrace what AI cannot do — presence, relationship, moment, truth
  3. Use AI as pre-visualization — not as final output
  4. Support provenance standards — C2PA, IPTC metadata, ethical disclosure
  5. Make work that requires a body in a place at a time — the irreducible core of photography

The image that started the debate

Eldagsen’s “The Electrician” was beautiful. It was also a lie — not because it deceived the eye, but because it claimed to be something it was not. The lie was not in the pixels. It was in the category.

Photography and AI art may look identical on a screen. They are not identical in the world. One records what was. The other imagines what was not.

That distinction is worth defending — not for nostalgia, not for gatekeeping, but because some things in human culture depend on the difference between seeing and dreaming.

We should know which we are looking at.

Spectrum is edited by Yuki Tanaka. Related: AI Tools for Creatives · Film Photography Revival