A major photography contest just put a stake in the ground on what counts as a real photograph in the AI era. According to The Verge AI, the rules now ban any synthetic or generated imagery, disqualify generative fill in post-production, and even rule out specific commercial AI upscaling tools by name. This is one of the clearest institutional answers yet to a question the industry has been dancing around for two years: where does photography end and AI generation begin?
The new rules, in plain terms
The Verge AI reports that every entry must be captured with a camera. Smartphone shots are allowed, but only in standard shooting mode. That means HDR, portrait mode, creative lighting effects, and panorama mode are out.
AI-powered enhancement tools can still be used, but with a strict condition: they can’t significantly change the image, add new information, or remove information that the camera captured. Two products get called out by name as automatic disqualifiers:
- Adobe Super Resolution
- Topaz Photo AI
Both rely on generative models to invent pixels when enlarging or sharpening. That’s the line the contest is drawing. If the tool hallucinates detail, it’s not photography anymore.
Why this matters now
What stands out here is the specificity. Most institutions have issued vague “no AI” statements that fall apart the moment you ask about denoising, sharpening, or computational photography on a phone. The contest is doing something harder. It’s separating AI tools that interpret existing pixels from AI tools that generate new ones.
That distinction is going to ripple outward. Newsrooms, stock agencies, legal evidence standards, and insurance claim processors all need a working definition of “unaltered photograph.” Until now, nobody had a clean one. This ruleset gives them a template.
It also puts pressure on Adobe, Topaz, and every camera maker shipping AI features as defaults. When Apple’s portrait mode and Samsung’s moon-enhancement features get filed under “not a real photo,” the marketing story for computational photography gets complicated fast.
The broader industry tension
This is the same fight playing out across creative industries. Publishers wrestling with AI-written articles. Music labels suing over training data. Stock photo libraries quietly flooding with synthetic content. Every field is being forced to define its own boundary, and nobody’s answer is clean.
Photography just happens to be further along because the manipulation problem predates AI by decades. Photoshop forced the first reckoning. Generative AI is forcing the second, and it’s a much harder one because the line between “enhancement” and “generation” is now a spectrum, not a switch.
Practical takeaways
For working photographers and anyone submitting to juried competitions:
- Read the AI clauses on every contest, grant, and publication agreement. The rules are diverging fast.
- Keep raw files. Provenance is becoming the new authentication standard.
- Be cautious with default phone modes. “Standard” shooting is being treated as the new clean baseline.
- If you use AI denoise or sharpening, document it. “No new information added” is becoming the test that matters.
For product teams building AI photo tools, the message is sharper. Generative upscaling is now a disqualifying feature in serious photography contexts, not a premium one. Expect more institutions to follow this lead, and expect a market split between tools that respect captured information and tools that invent it.
The contest’s answer to “what is a photo?” won’t be the last word. But it’s the most precise one the industry has produced so far. Full details at the original source.