AI editing has made it easier than ever to turn an imperfect photograph into a beautiful one—but also harder to decide what that photograph is actually worth. While selecting images for my yearly albums, I hesitated over a Summer Palace photo that Lightroom’s AI tools had transformed from a crowded tourist scene into a calm, elegant landscape. The result looked better, but it no longer showed the moment exactly as I had photographed it. This essay explores where I currently draw the line between useful AI photo editing and changes that begin to weaken a photograph’s authenticity.
The edited image looked almost exactly like the photograph I wished I had taken. Yet somehow, I trusted it less as a record of that moment. It made me wonder whether a photograph altered with generative AI should still belong in my photography portfolio—and how much editing can enhance an image before it begins to replace the reality that gave the photograph its value.
What I Use AI Editing For
The debate over how much post-processing is acceptable did not begin with artificial intelligence. Photographers have always disagreed about where editing ends and manipulation begins. Some prefer images that stay close to the original capture, while others see post-processing as an essential part of the creative process.
Before AI-powered editing became part of my workflow, I mainly used Lightroom for conventional adjustments. As I shared in How to Edit Photos in Lightroom for Beginners: 3 Essential Steps That Make the Biggest Difference, I normally begin by reframing the image through cropping, straightening, or perspective correction. Then I adjust white balance, exposure, contrast, and other tonal settings to shape the mood. Finally, I use masking and removal tools for smaller local refinements.
The edited result can look noticeably different from the straight-out-of-camera file. The colors may feel warmer, the composition may be cleaner, and certain parts of the frame may receive more attention than others. But generally, the photograph still respects the basic reality of the moment captured by the camera.
AI has changed the third step most dramatically.
Its ability to identify people, objects, skies, backgrounds, and separate visual elements is far faster and often more accurate than making selections manually. Tasks that once required careful brushing or complicated masking can now be completed in seconds.
The photograph that triggered this essay was taken at the Summer Palace, one of Beijing’s best-known historical attractions. It was peak lotus-viewing season, and the area was full of visitors. The architecture, lotus ponds, and summer greenery were beautiful, but the people scattered throughout the frame made the image feel busy and visually fragmented.
Using Lightroom’s Distraction Removal tool, I selected the People option. The software detected the visitors, removed them, and reconstructed the areas that had been hidden behind their bodies. The transitions looked surprisingly natural. Unless someone examined the image closely, they might never realize that several people had once been there.
A messy tourist scene became a tranquil landscape.
With the help of AI, I had created a photograph that looked more harmonious, more refined, and perhaps even more emotionally appealing. At the same time, the peaceful scene in the final image had never physically existed in quite that form.
That was when “better” and “more valuable” stopped feeling like the same thing.
Recalibrating the Value of a Photograph
To be clear, I am not writing this as a professional or commercial photographer, nor am I interested in defending photography as a superior art form that must be protected from AI-generated images. I enjoy experimenting with AI image generation myself, and I am genuinely excited by many of its creative possibilities.
What I want to protect is not the status of photography, but the integrity and value of my own photographs.
To think about where the editing boundary should be, I first had to ask a more fundamental question: where does the value of a photograph come from?
Photography has never been a completely neutral copy of reality. The photographer chooses the lens, position, framing, exposure, focal length, and exact fraction of a second to preserve. Editing adds another layer of interpretation. Even without AI, two photographers standing in the same place can produce entirely different versions of what appears to be the same scene.
Still, I think most photographs derive their value primarily from one or more of three things:
- Aesthetic value: providing visual pleasure through technique, composition, color, light, form, or atmosphere.
- Expressive value: communicating emotions, ideas, memories, or a personal way of seeing the world.
- Documentary value: preserving evidence of a real person, place, event, or moment with a meaningful degree of factual accuracy.
A strong photograph does not need to possess all three. A carefully composed landscape may primarily offer beauty. A conceptual portrait may focus on emotional or symbolic expression. A street photograph or news image may matter because it allows viewers to witness something that genuinely occurred.
For images built mainly around the first two forms of value, I think a reasonable amount of AI editing can be acceptable. Photography has always involved interpretation, and creative editing may help the image communicate its intended mood or idea more clearly.
But when a photograph depends primarily on documentary value, generative editing becomes much more problematic. Removing a person, adding an object, replacing a sky, or reconstructing part of a scene does not merely change how reality is presented. It changes the information the image contains.
In an expressive photograph, manipulation may reshape interpretation.
In a documentary photograph, manipulation may reshape evidence.
That distinction matters.
A Photograph Can Hold Two Kinds of Truth
The boundary is still not completely black and white, because photographs can preserve more than one kind of truth.
The first is factual truth: what was physically present in front of the camera at the moment the shutter was pressed.
The second is experiential truth: what the place felt like to the photographer, what drew their attention, and what they were emotionally trying to preserve.
My original Summer Palace photograph was more factually accurate. The site was crowded. People were moving through the landscape, and the popularity of lotus season was part of the real scene.
The AI-edited version, however, may have been closer to my experiential truth. When I looked across the water, I was not emotionally focused on the tourists. I noticed the lotus leaves, the traditional architecture, the summer light, and the quiet elegance I have always associated with classical Chinese gardens.
Removing the visitors made the image less faithful to the physical scene, but perhaps more faithful to the scene I believed I was seeing.
Neither kind of truth automatically cancels the other. The problem begins when they are confused—or when an image presents an emotional interpretation as if it were an untouched factual record.
Imagine someone photographing a landscape and using AI to add a rainbow that never appeared. The rainbow may create beauty, but most people would probably hesitate to classify the result as straightforward landscape photography if the alteration were not disclosed.
Now imagine a candid photograph of a father and son saying goodbye. Their expressions and body language carry genuine emotion, but an unintentionally funny road sign in the background distracts from the moment. Would removing the sign destroy the photograph’s integrity, or simply help viewers focus on its emotional center?
The answer may depend on what the image is meant to be.
If it is presented as photojournalism or documentary street photography, removing the sign changes the recorded scene and would be difficult to justify. If it is a personal family image or an expressive portrait, the same edit may feel relatively harmless.
The tool is identical. The promise made to the viewer is different.
The Real Boundary Is Context and Trust
Because of this, I no longer think the most useful question is simply, “Was AI used?”
AI now exists inside tools for masking, denoising, sharpening, selecting subjects, removing distractions, and many other ordinary editing tasks. Treating every use of AI as equally problematic would not tell us very much.
Instead, I think the more useful questions are:
What is this photograph claiming to be? Did the edit only change color, tone, framing, or emphasis, or did it change the actual content of the scene? Would knowing about the alteration affect how a viewer understands the image? And does the context create an expectation of factual accuracy?
A photograph submitted to a documentary competition should follow a much stricter standard than an artistic image created for a personal project. A commercial photograph used to show a hotel room or property should not remove defects that customers need to see. A family portrait, travel illustration, social media post, or fantasy-inspired image may allow far more creative freedom.
Transparency also matters. The ethical problem is often not simply that an image has been changed, but that viewers are encouraged to believe it has not.
Calling an image “AI-assisted,” “digitally altered,” or “a creative composite” does not necessarily reduce its artistic value. It simply helps the audience understand what kind of image they are looking at.
In fact, as AI-generated visuals become more realistic, this kind of honesty may become part of the value of the work itself.
Where I Draw My Line—for Now
Returning to my Summer Palace photograph, I would still share the AI-edited version on social media. I would use it as an illustration in an article, especially one that openly discusses how it was edited. Visually, it communicates the calm and classical atmosphere of the Summer Palace much better than the crowded original.
But I ultimately decided not to include it in my photography portfolio.
For me, a portfolio is not only a collection of attractive images. It also represents what I noticed, framed, and captured through the camera. Removing several people changed too much of the actual scene for the photograph to feel fully representative of that process.
Someone else may draw the line differently, and I do not think that automatically makes their answer wrong.
My current position is not that AI editing is good or bad. It is that the acceptable boundary depends on the primary value of the photograph, the degree to which its content has been altered, the context in which it appears, and the expectations of the people viewing it.
If the main purpose is beauty or personal expression, AI can be a legitimate creative tool—as long as the result is not misleadingly presented.
If the main purpose is documentation, adding or removing meaningful content should usually stop, because factual trust is the foundation on which the photograph stands.
If the image exists in a gray area, transparency is probably the safest boundary.
And if you are entering a photography competition, the practical answer is even simpler: read the rules. If the organizers allow a particular form of AI editing, it is allowed. If they prohibit it, it is not.
But if you are editing a photograph simply to send to someone you love, perhaps the only rule that matters is whether the image brings both of you joy.
Ultimately, I do not think the future value of photography will come from technical perfection. AI can manufacture perfection more quickly and cheaply every year.
What it cannot manufacture is the fact that a real person stood somewhere, noticed something, chose to preserve it, and attached part of their own memory to the image.
Perhaps that is the line I care about most: not whether a photograph is completely untouched, but whether it remains honest about what it is.
You must be logged in to post a comment.