Are you pro-AI or against AI?

Would you be happy for AI to help your camera keep a bird in focus, but feel quite differently about software adding a bird that was never there? It’s perfectly reasonable to feel both ways. Before we decide whether we’re for or against AI, it’s worth asking what we’re actually using it for.

That thought stayed with me after the Royal Photographic Society’s first International AI Forum on 17 September 2026, where I had the pleasure of taking part as an invited speaker. I was grateful for the chance to share recent developments, and listening to fellow photographers turned out to be just as valuable.

Some were keen to see how AI might make their workflow easier or open up new ways of creating artwork. Others would rather not use it, but still wanted to understand what was going on. I really appreciated that curiosity. We don’t have to agree about a technology to learn from each other.

More than image generation

The conversation also left me with the impression that, for some photographers, ‘AI’ immediately means generated pictures. From there, it’s easy to conclude that AI has nothing to do with photography. In fact, artificial intelligence covers far more than image generation, and understanding the difference helps us make choices with more confidence.

Much of the AI in photographic tools works by learning patterns from training examples. Developers use collections of data to teach a system to recognise features and relationships, such as the shapes and positions of eyes, faces or animal bodies. The trained system then applies what it has learnt to new images.⁠¹ How systems are trained, and how much data is involved, varies from one to another.

AI already in your camera

You may well have this technology in your camera already. Sony’s Alpha 7R V, for example, uses an AI processing unit and deep learning to recognise subjects, including people, animals, vehicles and insects.⁠² Recognising the subject helps the autofocus decide where to focus. It doesn’t generate a replacement subject in the photograph.

Some cameras also use AI for tracking and for assessing the light. Sony documents AI subject tracking in its Alpha 7 V,⁠³ along with deep learning for automatic white balance. Both help the camera respond to the scene in front of it.

That said, automatic doesn’t necessarily mean AI, and different cameras use different technologies, so it’s worth checking what yours actually does. Even so, someone who avoids AI editing may already be using AI when taking the picture, without realising it.

Autofocus is a useful comparison. Does letting the camera focus diminish our craft, or does it help us catch a fleeting expression or a bird in flight? Manual focus still has its place, of course. Whichever we choose, viewpoint, composition and timing still matter. Personally, I find that help with a technical task leaves more attention for the creative decisions.

Assisting or generating?

Some editing tools work in a similar way. Adobe Lightroom’s AI masks can pick out a subject or the sky so that we can adjust just that area.⁠⁴ Selecting the existing sky to darken it doesn’t create a new sky. It’s a clear example of AI helping with an editing task without generating replacement scene content.

Generative AI is different, because it can create new visual content. If I ask it to add a lion to my forest photograph, that lion was never recorded by my camera. A picture made entirely by image generation isn’t a camera photograph, and a photograph with generated additions needs to be described honestly.

Things get less obvious when something is taken out. A generative tool might remove people and fill the gap with convincing background details. The result may blend in beautifully with the surrounding landscape or architecture, but that doesn't tell us what was really behind those people. Believable detail isn't necessarily captured detail, and that applies to traditional tools too. The Clone Stamp and Content-Aware Fill typically copy or sample pixels from elsewhere in the same photograph, so the patch may look right without showing what was actually there. What differs is where the new pixels come from, which is why the method matters.

Authenticity, authorship and the rules

We also shouldn’t judge authenticity simply by whether any pixels have changed. Everyday exposure and colour adjustments change pixel values too. The more useful questions are where the content in the picture came from, what the processing has altered and whether the result meets the standards of its intended use.

This is where photographic authorship comes in. If a competition requires every visual element to have been photographed by the entrant, prompting a convincing lion simply can’t meet that requirement. However much creative intention and careful editing go into it, the generated animal doesn’t become something the entrant photographed.

In its guidance for entrants to exhibitions with FIAP patronage,⁠⁵ APS’s Salon Entrant Essentials⁠⁶ draws a line between permitted AI enhancement, which works only from the pixel data in the original photograph, and prohibited synthetic generation, which brings in outside content or visual elements that weren’t there originally. Even so, it’s important to check the rules for each competition and category. The 2026 BirdLife Australia Photography Awards,⁠⁷ for example, allow minimal AI noise reduction and sharpening but prohibit generative AI editing. We shouldn’t assume that every competition treats every tool the same way.

Look past the label

Software labels don’t always make things easier. In its announcement of 27 August 2026, Adobe described Photoshop’s AI Assisted Editor, in beta, alongside the familiar Pro Editor. Adobe lists Generative Fill, Generative Expand and Generative Upscale among the AI Assisted Editor’s tools. It also says Prompt to Edit works in the Pro Editor and produces a generative layer.⁠⁸

I think the names could confuse photographers. ‘AI Assisted’ can sound like help with existing photographic content, yet this interface includes tools that generate content. And the ‘Pro Editor’ has generative features too. Adobe’s documentation⁠⁹ explains how the two editors relate, but the names alone can’t tell us whether a particular edit is suitable for a photography competition.

Language adds another wrinkle. A generative tool can ‘assist’ someone’s workflow. ‘Assistance’ describes the role a tool plays, while ‘generation’ describes what it actually does. The two aren’t mutually exclusive. Looking past the label helps us ask the real question: does this feature recognise existing content, help us adjust it or create new visual content?

Keeping the conversation open

I came away from the forum encouraged by how willing people were to learn. None of us has to embrace every new tool, and it's perfectly reasonable to set boundaries for our own work. Even so, keeping up with developments matters, especially if we enter competitions. Many rules depend on what a tool does to the picture, so we need to know what our cameras and software are actually doing. 'I didn't know it used AI' is unlikely to be much of a defence. In salons with FIAP patronage, for example, entrants accept that they can be sanctioned if they don't comply, with penalties ranging from short-term to lifetime bans. So even if we never use AI ourselves, understanding it is still our business.