GPT Image 2.5 Sunburst vs Nano Banana Pro - what AI picks when your prompt is lazy

Why do so many AI images look the same?

While recreating a bunch of photos, I always made a lazy version first, just to see what the AI does on its own. After a while I noticed it's not random. Every time I left something out, the AI filled the gap with the same handful of defaults. Here are the five I kept running into.

1. It picks who

I asked for "a stylish man in a suit" and didn't say anything else about him. The AI picked a young guy with tousled hair and a model's jawline. That's who it reaches for when you don't describe anyone.

If the person matters, say who they are: age, skin tone, hair, build.

2. It picks the body

For a runner, I asked for a man after a run. I got a gym ad model with huge arms and a smartwatch.

So I added his age and a grey beard. Still way too muscular.

Adding age and a beard didn't change the body at all. It only got fixed when I wrote the build out: lean, average, not muscular.

3. It picks the light

Look at those two runners again. Both are golden hour with the sun behind them and a blurry skyline. That's what "dramatic lighting" and "cinematic" turned into for me almost every time.

If you want a certain light, describe it instead. Where it comes from, how hard it is and what it hits.

4. It picks the polish

This one's from an early skin test. I stacked every "quality" word I could think of: flawless, glowing, perfect, stunning, high quality, 4k.

You get a beauty ad. Glossy, smooth, perfect skin that reads a little plastic. Stacking quality words just made it more polished. When I wanted real skin, I had to ask for it: visible pores, uneven tone, a few tiny imperfections.

5. It picks the room, on every model

Here's the part that surprised me. I ran almost the same lazy red room prompt on a second model. The only change was saying the man is Black. Compare this one with the first image in this post.

Both went straight to the same room: a velvet armchair, table lamps, frames on the wall. So it's not one model's quirk. Both models reach for the same stuff when the prompt is lazy.

Why it works

Every time I left something out, the AI decided it for me, and it kept making the same choices. Writing out who it is, their build, the light, how polished it should be and the place covered most of it.

Your turn: take your last prompt and check it for those five.

And if you'd rather skip the prompting, Emero writes the prompts for you and remembers your preferences so you're not repeating yourself every chat. Everything's at cost.