Not Real, but True: What Photojournalism Must Protect in the AI Era

Right about 10 years ago, I wrote an article that went viral and sparked a lot of debate. At the time, AI was just starting to show the tip of its beak, and image manipulation was already in full force thanks to a variety of very potent software like Adobe’s Photoshop. The origin of the article was shaped by the ever-changing rules of acceptance for the various photojournalism competitions, trying to keep up with a fast-evolving technology. The piece questioned whether rigid rules could ever truly guarantee photographic integrity. The intent was noble, obviously, in trying to keep authenticity and veracity as the number one requirement for any photo to be accepted in the competition.

At the time, my position—which still remains—was that there is no setting boundaries on an ever-shifting landscape, as technology will continue to stay a step ahead of whatever rule we think permanently sets things straight. Today, as we see, it has become impossible to distinguish a real image from one totally created via synthetic methods. And while not yet in photojournalism, some have even ended up winning “photography” contests.

My conclusion at the time—which also still holds—was that it wasn’t the tools that mattered in judging the authoritative and authentic quality of a photojournalistic image, it was the intent. The photographer’s intent.

All photographs are a lie

My path to this conclusion is that all images, all photographs, are a lie. Or rather, a reconstruction of reality purposefully modified to fit one’s perspective. Beyond smells, movement, and noise—automatically deleted—photographers do not decide what to capture, but rather what to edit out. Not only via their framing, but through their choice of lens and camera format, the distance from the subject, and even the decision on what to cover. A photojournalist’s work is a testimony. It is not a scientific reproduction. It is biased, opinionated, and heavily infused with one’s own culture, gender, and race, among many other influences.

Thus, saying one should not use the dodge or burn tool on an image already taken because it will alter reality is like insisting a sci-fi filmmaker shoot only what exists—too late. The damage to reality has long already been done.

The photojournalism world needs to come to terms with the fact that photojournalism is not the truth, nor is it an exact depiction of reality. It’s an interpretation, a testimony. And like all testimonies, it is full of flaws—subjectivity flaws. Thus, it really doesn’t matter what “tools” are being used, as long as they sincerely try to convey the emotional, factual, and contextual essence of the moment.

It’s all about intent. If the intent is to testify with as much accuracy as possible, using dodge or burn tools, cropping, highlighting, and even removing objects is absolutely fine. Whether there was an ugly cable on the top right side of the image doesn’t take away from the information in the rest of the image. When we give an oral testimony of an event, we never depict it in the same way and constantly add or delete information from it. And it doesn’t matter, because the core of the testimony—the information it conveys—remains the same. And before cameras, this is how we built magnificent civilizations.

What is real ?

In the age of AI, where we can now reproduce any image to the quality of a “real” image, it is even more important to acknowledge—and define—the difference. What is the difference, then, between a photographic image, if both are realistic?

One is done by capturing light-bouncing real objects at a certain place and time. The other is a replica—a construct of that event—with no physical, material relationship to it. The fundamental difference is both temporal and physical. One is taken in the same moment and location as the event depicted. That’s the “reality” part. The other is completely unrelated—before, after, even during, it is still not in perfect sync with that fraction of a second.

A photograph is not only anchored in time and space—it’s the result of witnessing. Even when automated, it emerges from a moment that actually occurred. The AI image is imagined, not encountered.

And thus, photojournalism, instead of playing the role of scientific documentation—which was never its origin—should highlight and build on its uniqueness: witnessing. And allow for whatever image manipulation (from camera selection to lens, settings, positioning, post-production tools, and editing) as long as it contributes to the effort.

The intent to inform

AI-generated images that aim to represent real events can serve journalism—just as sketches, infographics, or reenactments have in the past—so long as they are transparently presented and motivated by a sincere intent to inform, not to deceive. It is not the medium that determines truthfulness, but the motive behind it. Transparency—through labeling, metadata, or visual cues—is essential to preserve trust.

What should not be allowed—what should not be crossed—is the intent to inform deceitfully. That is what we judge in a photojournalist’s image: its fidelity to truth-telling.

And how do we judge that? Where are the lines, the boundaries?

We judge it by defining provenance—how the image was made, by whom, and why—and by reinjecting the human element. In the same way that, in a trial, we know who the witness is (policeman, neighbor, lover, etc.) and how that impacts their testimony, it is the same for photographs.

So, two requirements are needed to properly judge a photojournalistic image: its provenance and a transparent list of what was done to the image from capture to submission to the jury. Because trust is a relationship – a human relationship- and understanding intent is its foundation.

Photojournalism was never about freezing reality—it was about choosing where to stand, and why. In an age where images can be conjured from nothing and perfection is simulated with ease, what still sets a photograph apart is the presence behind it: the intent, the context, the accountability. Not unedited, but transparent. Not neutral, but human. Loudly human.