Position is weighting
Most image models weight earlier terms more heavily than later ones. The first three or four terms effectively decide the image, and everything after is adjustment.
That makes term order a real control rather than a stylistic choice. Putting your subject after four style adjectives produces a different image than putting it first.
What earns its place
Medium — photograph, oil painting, 3D render. Lens and film stock for photographic looks. Lighting direction and time of day. Composition framing. These reliably change the output.
What mostly does not: "masterpiece", "best quality", "8k", "highly detailed", "trending on artstation". They correlated with good images in early diffusion training captions and are close to noise on current models, while still occupying the positions that matter most.