World Press Photo has tightened up the rules for its 2026 contest, and the list of what's now out is longer than you might expect. No synthetic images. No generative fill. No AI enlarging either, so Adobe Super Resolution and Topaz Photo AI are both named and both ruled out. Then there's the bit that surprised me more: phone pictures are only eligible if they were shot in standard mode, so HDR, portrait mode, creative lighting effects and panorama mode all disqualify an entry, and multiple exposures and stitched panoramas are gone as well, however they were made.
It's worth sitting with that for a minute, because a fair bit of ordinary, honest photography falls outside those lines.
The rule is about trust, not technique
To be fair to World Press Photo, none of this is really about punishing anybody for blending frames. It's a photojournalism contest, and it's trying to hold a line on one question: does this image record something that actually happened, in front of a lens, at a moment in time. Nothing added, rearranged, reversed, distorted or removed from within the frame. Colour work and cropping are fine, sensor dust can go, but anything that changes what's in the picture is out, and any influence over the scene has to be declared, so a posed portrait or a reenactment gets said out loud in the caption, as do any directions given to the person in front of the camera.
That's a sensible place to draw it for news photography, where the whole point of the image is that it's evidence. I understand exactly why they've put it there.
Worth saying as well, because it gets missed, the rules aren't anti-AI across the board. Denoise is fine. So are automatic adjustments on levels, colour and contrast, along with object selection for local adjustments, all on the condition that they don't add or remove anything the camera captured. You can use AI to pick out a sky and work on it, but you can't hand in a shot your phone quietly built out of several exposures, which tells you the line isn't really about how clever the software is, it's about whether anything got added along the way.
But hardly any of us are shooting under those rules
Here's the bit that's been nagging at me. Outside of photojournalism, computed images aren't the exception; they're the default. Plenty of landscape and seascape photographers stack focus or blend exposures as a matter of course, and most stopped thinking of it as anything unusual years ago. Every phone in every pocket is doing HDR without ever being asked. Portrait mode is now how most people on earth see a photograph of a face, and none of that is invented out of nothing, it's still light hitting a sensor, but it is an image assembled from more information than one exposure could ever hold.
So when a contest draws its line at the single frame, it's protecting something real for its own corner of the industry, but it also shows up how far the rest of us have drifted without ever really noticing we'd crossed anything.
Then there's what AI is doing to all of it
The reason World Press Photo has had to spell this out in so much detail is the same reason it matters to the rest of us. Generated images are now indistinguishable from photographs at a glance, and they're being made at a scale that's hard to picture. The Gemini app passed a billion people a month on the 11th August, and Google reckon it's turning out over 150 million images a day. That's one app.
Google also had to pull the AI image feature out of Google Earth a day after it launched, because people were immediately generating photorealistic satellite imagery that broke its own policies. One day.
Against that, a photography contest banning AI outright isn't an overreaction; it's one of the very few places drawing any line at all, at a point where a convincing photograph of a real place, taken by nobody, is a perfectly normal thing to come across.
Where I land on it
I don't think World Press Photo is wrong to ban AI images from a photojournalism contest, because that's precisely what the contest is for. However, I don't think the stricter stuff about HDR and portrait mode and stitched panoramas is really about AI at all. It's a reminder that "unmanipulated photograph" stopped being a simple category a long time before generative AI turned up, and most of us made our peace with that a good while ago without giving it much thought.
So the more honest question probably isn't whether a picture is AI or not; it's what any of us are actually claiming happened in front of the camera when we hand somebody an image and call it a photograph.
I haven't got a tidy answer to that, and I'm not sure the industry has either.
🚨 Here’s a link to the World Press Photo Competition Rules: www.worldpressphoto.org/contest/entry-rules