I’ve spent twenty years building studio setups and teaching photographers how to light subjects properly. In the last six months, I’ve noticed something troubling: my inbox fills with questions from photographers who’ve built studios based on AI-generated diagrams, and almost every one of them is fighting with something that doesn’t work.

The diagrams look fantastic. They’re rendered in clean 3D, they have professional labels, they show modifiers positioned at precise angles. A beginner sees one and thinks, “Great, I have a blueprint.” Then they spend $800 on equipment and discover the setup doesn’t actually work in their space.

This isn’t about being anti-technology. AI tools have genuine value—they can help organize ideas, generate variations on concepts, or spark thinking about unfamiliar techniques. But there’s a critical difference between an idea generator and a technical specification. Lighting diagrams are specifications. They need to be correct.

The Power Ratio Problem

Here’s a concrete example. I looked at an AI-generated three-light portrait setup that was shared on a photography forum. The diagram showed:

  • Main light at 250W
  • Fill light at 100W
  • Hair light at 150W

The accompanying text claimed this would give a “2:1 ratio with controlled highlights.” That math doesn’t work. If your main is 250W and your fill is 100W, you’re looking at roughly 2.5:1 when light from both sides hits the face. But here’s the real problem: the hair light at 150W is close to the fill ratio. Depending on where it’s positioned, it’ll either blow out the background or create competing shadows on the face.

A human lighting professional would immediately recognize this inconsistency. An AI trained on thousands of images and forum posts can produce something that looks right without understanding the actual physics of inverse square law or how ratios compound when multiple light sources hit a subject.

I tested this setup myself—spent ninety minutes dialing it in before I realized the fundamental problem. A photographer with less experience might have spent a week troubleshooting something that was broken from the start.

The Geometry Problem

AI excels at rendering things that look spatially plausible. It struggles with the constraints of actual rooms.

One popular AI guide I found showed a softbox positioned 18 inches from a subject’s face, angled at 45 degrees, with a reflector on the opposite side. The diagram showed an 8-foot ceiling. In reality, if you’re shooting head-and-shoulders work and your main light is only 18 inches away at a 45-degree angle, your light falloff across the face becomes extreme—bright cheekbone, dark forehead. The diagrams didn’t account for this. They just showed the setup looking balanced in 3D space.

Worse, I’ve seen diagrams that position hair lights directly overhead in standard studio rooms. Anyone who’s actually hung equipment knows that most studios have drop ceilings at 8 or 9 feet, with HVAC ducts, sprinkler systems, and electrical conduit taking up the space above. You can’t position a light where the diagram says if there’s structural infrastructure in the way. AI doesn’t know about drop ceilings.

The Generic Setup Trap

The most subtle problem is generalization. AI-generated guides tend toward universal solutions—“the perfect setup for product photography” or “the ideal portrait lighting.” But lighting isn’t universal. A setup for a pale-skinned subject with a white background needs different ratios than one for a darker-skinned subject with the same background due to how cameras meter and how skin reflects light differently across the spectrum.

I’ve seen AI setups that work reasonably well for 5-foot subjects but would be completely wrong for someone 6'4". The modifier positions don’t scale logically because the AI isn’t thinking about why the light goes there—just that it should go somewhere reasonable.

What This Actually Costs You

Time. Money. Frustration. A photographer follows one of these guides, buys the recommended gear, builds the setup, and discovers it doesn’t produce the results shown in the diagram. Then they adjust—move the light, change the power, reposit the reflector. They’re essentially learning through trial and error when they thought they had a professional specification.

I’m not suggesting AI-generated diagrams are worthless. But they’re not blueprints. They’re sketches. Treat them that way. Use them as starting points, not destinations. If you’re building a setup based on an AI diagram, bring the understanding that you’ll need to adjust for your actual room, your actual subject, and your actual equipment.

The professionals who’ve been doing this work have their process down because they understand the underlying principles, not because they found the perfect diagram. That’s the actual work worth doing.