Where AI Belongs in Photography (And Where It Doesn't)

The Generated Sunset

A client asked me to mock up a campaign concept — Mediterranean coastline, golden hour, product in foreground. I generated three options in fifteen minutes. The compositions were solid. The light was warm. The mood was exactly right.

Then I looked more closely. The shadows fell in two different directions. The water reflected a sky that didn’t match the one above it. The golden hour light had no source — it just existed, evenly, everywhere.

A non-photographer would have approved all three. The images looked good. They just weren’t real.

The Split

The photography world has split into two camps: people who think AI will replace photographers, and people who think AI is a gimmick that serious photographers don’t need.

Both are wrong.

AI is a tool. Like any tool, it’s excellent at some tasks and terrible at others. The interesting question isn’t whether AI will replace photography. It’s which parts of the workflow benefit from AI, and which parts need a camera and a human eye.

Where It Works

Concept visualization. You’re pitching a campaign and need to show the client what you’re thinking. AI generates mood boards and stylistic explorations in minutes. This used to require hours searching stock libraries or sketching by hand. For pre-production ideation, AI saves real time without meaningful trade-offs.

Batch editing. Background removal across hundreds of product images. Format conversion for different platforms. Sky replacement for a real estate portfolio. These are mechanical tasks. AI handles them faster and often better than manual work.

Content at scale. Social media needs volume. Blog posts need illustrations. Presentation decks need visuals. For content where the concept matters more than the authenticity of the specific image, generation is faster and cheaper than a photoshoot.

Style exploration. Before committing to a shoot direction, AI can show what different approaches might look like. Different color palettes, different compositions, different moods. This refines the brief before spending money on production.

Where It Doesn’t

Documentation. If you need to show that something happened, that a building exists, that a product looks a certain way — you need a camera. AI images are not evidence. They’re illustrations.

Authenticity. People can tell. Maybe not consciously, but generated images lack the imperfections that make real photography feel real. The slightly off composition. The unexpected light. The human element that a camera captures and AI approximates.

Brand consistency. AI-generated images have a look — and it’s not your look. Maintaining consistent brand photography requires someone who understands your brand, your spaces, your people. AI can supplement this. It can’t replace it.

Anything with legal weight. Product photography for e-commerce has compliance requirements. Real estate photography has disclosure obligations. Event documentation may have contractual implications. AI images don’t meet these standards.

How I Think About It

Pre-production: AI. Concept visualization, mood boarding, brief refinement. This is where AI adds the most value per hour.

Production: camera. The shoot benefits from human judgment, real-world conditions, and the quality that comes from actual light hitting actual surfaces.

Post-production: selective. Background removal, batch processing, format conversion — AI. Color grading, creative editing, final selections — human judgment.

Content repurposing: AI for generating variations and platform adaptations. The core photography stays real. The derivatives can be augmented.

The Four Questions

When deciding between AI and a camera, I ask four things:

Does this need to be real? If it represents something that exists or happened, photograph it. If it represents an idea, AI is fine.

Will someone scrutinize this? Product close-ups, architectural documentation, portraits — these get examined carefully. Artifacts matter.

Does this need to match existing content? If it’s part of a broader visual system with existing photography, inconsistency will show.

What’s the shelf life? Campaign visuals used for weeks can be generated. Brand photography used for years should be real.

The Skill That Matters

Knowing how to prompt doesn’t replace knowing how to see. But knowing how to see — and then applying AI tools with that foundation — is genuinely powerful.

The photographers who will do well aren’t the ones who refuse AI, and they aren’t the ones who embrace it without judgment. They’re the ones who can evaluate AI output with a trained eye. Who notice when the light is impossible, when the depth of field contradicts the lens, when something feels slightly wrong even though it looks right.

That evaluation requires years behind a camera. There’s no shortcut to visual literacy. AI is what you do with it after.