AI Can Make a Picture, That Doesn't Make It a Photograph

I still use AI. I'm not out here trying to churn butter by hand in a cabin while yelling at electricity. I use the tools. I test the tools. I've built workflows around the tools when they save time, cut friction, or keep me from doing some repetitive task that makes my soul feel like it got trapped in a printer jam. I'm not precious about it. If a tool works, I use it.

But someone saying I should just prompt an image instead of shooting one is the kind of statement that smells like someone microwaved fish in an office breakroom, wandered back to their desk like nothing happened, and left the rest of us standing there wondering how we all ended up breathing the same air as a person who clearly peaked in a group chat. I've heard the argument enough times now that I can mouth along with it like a bad karaoke track. "Why drag gear into the desert? Why sweat through your shirt? Why fight the wind?" Because the wind is the point, you walnut. The sweat is the point. The part where nothing goes according to plan and you have to actually make a decision with your hands and your brain at the same time — that's the whole job. That's what a prompt box will never touch.

I’ve Been Tempted. Of Course I Have.

I'm not some noble craftsman standing on a hill with a leather camera strap and a moral superiority complex. I like convenience as much as the next over-caffeinated mammal pretending his inbox is under control. I've sat down with generators. I've typed prompts. I've watched something appear in under ten seconds and thought, "Huh. That's actually not deeply terrible," which is exactly the problem, because "not terrible" is the ceiling and I'm moderately convinced it will always be the ceiling. AI can mock up a direction, explore a mood, rough out a concept before anyone starts spending real money. That part is useful. I'm not going to pretend it isn't just so I can cosplay as a purist who still light meters by candlelight.

But the moment anyone confuses that with the actual act of making the thing, we've left the building. What AI generates is a clean, frictionless fantasy where the light behaves, the subject cooperates, the setting has no logistical problems, and nobody forgot the one prop that was apparently central to the entire campaign even though it was only mentioned once in a text thread three weeks ago. A prompt box doesn't know what it's like when the plan meets the location and the location wins for a while. It just conveniently skips straight to the part where everything looks nice without having had to fight for its place at the dinner table.

And I want to be specific about this, because "I use AI" is a vague enough statement to mean almost anything in 2026. Narrative handles my culling and cataloging, and it's genuinely good at it. The AI healing tools in Lightroom get used on almost every edit I touch. Photoshop's version of the same thing shows up when I need it. AI-powered noise reduction and image scaling from Adobe and DxO are solidly in the rotation, and they've gotten scary good. I use Claude to help me look at my own images and pull out keywords and topical themes I might not have thought of. I use AI heavily to study my analytics and figure out what's actually worth publishing. I've built my own tools for managing SEO. None of that is hypothetical. That's my actual Tuesday.

What I don't use it for, almost at all, is generating the image itself. Everything I just listed happens before the shoot or after the shoot. The middle part, the part where I'm standing somewhere with a camera and a plan that's already falling apart, that's still mine. That's the line. And I think it matters that I can tell you exactly where I drew it instead of just gesturing vaguely at the concept of "real photography" and hoping you trust me.

The Mess Is Where the Photograph Gets Its Spine

A photoshoot has weather. It has time. It has people getting tired, awkward, distracted, nervous, or suddenly brilliant when you least expect it. It has the client loosening up three minutes after I thought the moment was dead. It has the light doing something useful for once. It has me actually remembering to pack all my batteries and a memory card, like a raccoon accidentally solving a math problem. It has the location fighting me like a shopping cart with one busted wheel, and me standing there with my expensive little anxiety rectangle, trying to make a real decision before the whole situation evaporates.

None of that is a flaw. That's the raw material. Photographs aren't assembled from visual ingredients the way a prompt box stacks pixels. They're made from pressure, timing, trust, boredom, adjustment, failure, and the specific kind of recovery that only happens when you're physically standing in a situation that doesn't care about your mood board. I've taken bad photos. Plenty of them. Blown the exposure, botched the timing, picked the wrong lens and had to improvise something that almost worked. But even the worst photo I've ever taken has something an AI image doesn't: a person was there, making choices, in a moment that was never going to happen again.

Not every image needs that. Some images are disposable. Some are placeholders. Some are basically visual paper towels, and nobody needs to frame the Bounty. But the work I still care about carries evidence of contact. Somebody stood there. Somebody noticed something. Somebody made a call, and you can feel it in the frame even if you can't explain why.

What I'm Actually Protecting

I'm not protecting photography from AI. Photography doesn't need me to defend it. What I'm protecting is the part of the work that requires me to show up, pay attention, and deal with whatever actually happens instead of whatever I wish would happen. That's the part that makes the work mine. That's the part that earns the frame. If you're still building that craft across disciplines, The Well-Rounded Photographer: 8 Instructors Teach 8 Genres of Photography is worth a look — eight different ways of showing up for the real thing.

AI can make a picture. I've seen it do it. It's fast and it's getting better and I don't care, because speed was never the thing I was chasing. I was chasing the moment where the plan falls apart and something better walks in uninvited, badly timed, gone in four seconds, and I either catch it or I don't.

And honestly, if AI ever gets there — if the machine figures out how to do the part I don't believe it can — I'm not sure whether I'd fully embrace it or just quietly disappear. I don't have a comfortable answer for that one.

I still want the scar tissue. The real, the raw, the experience. For me, that's what makes it fun and worth doing.

Behind-the-scenes images of me on location, by Melissa Hill, used with permission.