Three related terms, defined

Exposure is how much light reaches the sensor, ISO is the amplification applied afterward, and noise is the static that amplification brings with it. For a rating submission the rule follows: get enough light and keep ISO low, because noise degrades exactly the fine detail a score reads.

Why this is not just "brightness"

Two photos can arrive at the same final brightness by very different routes. One had plenty of light and needed little amplification. The other was underexposed and got pushed up by a high ISO to match. They can look nearly identical at a glance and be nowhere near identical as files - the second one carries speckle in its shadows and softness in its fine detail that the first one does not, because that detail was amplified along with everything else and amplification does not distinguish signal from noise.

Why a rating tool cares

Texture, edge definition and surface detail are exactly what noise degrades first. A standardised submission is meant to hold the input constant so a rescore reflects the subject, not the file; a noisy file from a high-ISO shot is a less stable input even before anything else about the photo changes, and it can quietly explain a chunk of the error bar between two otherwise identical retakes - on a tool that reports per-axis results, like Rate Cock, a noisy file tends to drag the texture axis specifically rather than every axis equally. Image classifiers share the weakness: Dodge and Karam (2016) tested networks against blur, noise, contrast and JPEG compression and found them "susceptible to these quality distortions, particularly to blur and noise". How much of that noise the model can see through depends on the model, but no model recovers detail the sensor never resolved.

The one rule

Fix exposure with enough light in the first place, and keep ISO low and constant rather than letting the camera raise it to compensate for a scene that is too dark. Adequate, even light does two jobs at once: it keeps dynamic range inside what the sensor can hold, and it keeps ISO low enough that noise stays out of the file. A dim room "fixed" by cranking ISO produces a technically bright photo that is worse raw material than a slightly dimmer one shot with more light and less amplification - the amplified version costs equipment-side accuracy too, since a noisy reference photo undermines a tape measurement's own record just as much as it undermines a rating. The same low-light, high-ISO problem is why a human judge working from a dim upload complains about grain before they complain about anything else in the frame. Same room, same lamp, same ISO setting every time: three lines in a log, and the noise stops being a variable.