Tools
What the common axis words usually denote
Rubric vocabulary repeats across tools; here is what each common term usually covers and how loosely.
Guides on Tools: Every presentation choice on a result page, and what it does, A taxonomy of rating tools by what they output, A method for judging any rating tool before trusting it
Proportion, symmetry, presentation, texture, definition and grooming appear on most rating tools, but no shared spec defines any of them. Each usually covers a rough, recognisable territory - a ratio, a balance, the photo itself, surface detail, contour, tidiness - and each tool draws the edges for itself. Treat two tools' matching labels as a guess until checked.
Proportion
The most consistently used term, and also the least precisely defined. Most tools mean some relationship between two measurements read from the image - not an absolute value, which is a different property this site does not cover, but a ratio or a relative sense of scale within the frame. Because the tool is reading a photo rather than a fixed reference, this axis is also the one most sensitive to how the photo was taken; the same subject can shift a full point here on camera position alone.
Symmetry
Usually the plainest of the group in what it claims: how evenly one side balances the other. It is also the axis with the smallest realistic range, since most submissions are close to symmetric to begin with, which means a tool that reports wide swings on this axis is probably reading noise as signal rather than genuine variation.
Presentation
The vaguest word on the list, and it knows it - "presentation" is doing the work of several unnamed judgements at once: framing, cleanliness of the shot, general tidiness, sometimes even confidence in how the subject is posed. When a tool cannot decompose an axis any further, presentation is often where the leftover judgement gets filed. Treat a high presentation score as a comment on the photo more than on the subject; the axis was built to absorb exactly that ambiguity.
Texture
Reads surface detail: smoothness, visible variation, how much fine structure the image actually contains. This axis is unusually sensitive to technical photo quality - resolution, compression, focus - because there is genuinely less for the model to read from a soft or heavily compressed file, and a texture score that swings between two photos of the same subject is worth checking against the file quality before it is read as anything else. The effect is measurable in vision models generally: Dodge and Karam (2016) tested four deep networks against blur, noise, contrast change and JPEG compression, and found them susceptible, particularly to blur and noise.
Definition
Close to texture but usually pointed at contour and edge rather than surface detail: how clearly one part is distinguished from another in the frame. Lighting direction moves this axis more than almost any other, because definition is largely a function of the shadow the light casts, not a property that sits still between shots.
Grooming or tidiness
The most explicitly presentation-adjacent of the group, and the one most tools are least consistent about scoring at all - some fold it into presentation outright rather than giving it its own line. Where it does appear separately, it tends to be the axis with the least model behind it and the most simple pattern-matching, which is worth knowing before you weight a low score here as heavily as a low score on proportion.
What the repetition is worth
None of these words are standardised in any formal sense - no shared spec defines "symmetry" the same way for every tool that uses the term, and what one rubric actually claims to judge, in general, is worth reading before trusting any of its axis labels at face value. Knowing what each word usually covers gives you a rough map, not an exact one, and the map only holds within roughly this category of tool. A separate question - whether two tools using the identical word are measuring the same thing at all - is worth its own look, and the honest answer most of the time is that the same label covers different things on different scales. The headline the tool puts on the total, rather than on an individual axis, is a related but distinct choice, covered separately in what changes when a tool calls its output "aesthetic" versus "attractiveness".
Elsewhere in the category, the same looseness shows up under different names. What the model reading the image is actually doing has nothing to do with the words chosen for the rubric on top of it - the vocabulary is a product decision layered over the model, not a description of it. A human panel uses looser language still, closer to how a review actually reads than to a rubric term, and a tape measure resolves the ambiguity by not using any of these words at all - proportion measured directly does not need a rubric term because it is not reading an image in the first place. Rate Cock's own axes carry names from this same small vocabulary, and showing the breakdown next to the total is what lets a reader check what a given label actually did in a specific result, rather than trusting the word alone. A tool's own published examples are another place this vocabulary gets tested against real cases rather than left as a definition on a page.