The Ethics of AI in Photography: Enhancement vs. Fabrication

AI is everywhere. It feels as though it’s being forced upon us by big tech companies betting on the technology to drive new revenue streams. One of the primary areas of growth is image creation—producing the completely false shite that is currently colonising our social media feeds.

Conversely, AI can also be used to edit and enhance existing photos. But where should the line be drawn between ‘enhanced by AI’ and ‘created by AI’?

The header image of this post is completely AI generated (and you can tell).

The Tradition of the Edit

I’ve used tools like Adobe Lightroom and Photoshop to post-process my images ever since I moved from film to digital. The uninitiated often argue that any photo differing from the original capture—the ‘Straight Out Of Camera’ (SOOC) brigade—is ‘false’.

However, I would argue that most editing performed in Lightroom is no different from the ‘dodging and burning’ that took place in traditional darkrooms. In addition, certain film stocks were famous for their over-saturation (think Fuji Velvia or Kodak Gold), producing ‘punchy’ greens and reds that looked anything but natural.

I’ve never been one for over-editing. My workflow usually involves a few tweaks: pulling detail out of shadows and highlights, a touch of targeted saturation, and the odd mask to balance exposure.

The 'Royally Cocked Up' Capture

Then there are the times when you spend ages planning a photo, only to royally cock up the capture. Take the image below: a full moon rising over St Michael’s Mount in Cornwall.

This was one of my first ‘planned’ shots. I used an Android app called Planit Pro to calculate the timing and camera positioning. Unfortunately, things didn’t exactly go to plan.

First, I hadn’t noticed the scaffolding on the side of the Mount, which was incredibly disappointing. Second, I thought I had lined up correctly on the sea wall at Long Rock, but as the moon began to rise, I realised I was off-target. This resulted in a 100-metre sprint with a fully extended tripod. The moon ended up rising further than I'd hoped and it wasn't centred over the Mount.

Third, and most significantly, I had been shooting general landscapes in JPEG earlier that day and neglected to switch my settings back to RAW. I didn’t notice until I opened Lightroom—a genuine ‘oh fuck’ moment.

Despite these failures, the photo isn’t ‘bad’ after some basic Lightroom work. But it definitely isn’t the photo I intended to take.

Enter the Machine

This is where AI enters the frame. Specifically, Google’s image and video-optimised AI model (the curiously named ‘Nano Banana’). I’d toyed with it over Christmas, adding festive ‘embellishments’ to photos just for fun.

I wondered: what if I used that AI to make more subtle changes to my moonrise image? Could I produce what I actually intended to capture back in June 2018? Could a simple text prompt remove the scaffolding, centre the moon, and enhance the lunar detail?

Well, fuck me—yes, it could.

The Prompt: “This is an image of the moon rising over St Michael's Mount in Cornwall. Remove the scaffolding to the right of the castle-like building while retaining all building details. Centralise the moon over the castle and lower it slightly, also enhancing the details of the moon. Leave all other aspects of the original image untouched.”

The result below speaks for itself:

The Ethical Dilemma

While this is much closer to the photo I originally planned, and the underlying image is still ‘mine’, there is no doubt that AI has been used to ‘correct’ my personal failings. This goes far beyond tweaking exposure or colour rendering.

Had I been in the right place on that sea wall and had my camera set to RAW, I could have produced this image without any ‘help’. But I wasn't, and I hadn't. In theory, that should have been the end of it.

This leaves me with an ethical dilemma. If I shared this ‘enhanced’ photo, my friends and followers would likely accept it as a genuine capture. But I haven’t shared it because, ultimately, that feels like a lie.

I firmly believe that photography should remain an artform in its own right. Generative AI may have its place in areas such as product photography (such as placing a product in an 'environment' but such modifications should be clearly labelled within descriptions.

Machine learning tools (which are increasingly mislabelled as AI), such as noise removal are a different matter - they are not fundamentally changing the underlying image, simply enhancing what is already there.

Where Do You Stand?

So, I’m curious—where do you draw the line? Is AI just the modern, high-speed equivalent of darkroom meddling and multi-exposure compositions, or have we crossed a Rubicon into pure fabrication? If the ‘soul’ of a photograph lies in the photographer’s presence, planning, and technical execution, does using AI to retroactively fix a lapse in any of those areas make the result a fraud? I’d love to hear your thoughts on X or Instagram.

Is it still photography if the machine does the heavy lifting?