The label as a tool-category claim

"AI-powered" used to mean something, back when it distinguished a tool from one that ran a fixed formula on a tape measurement. It no longer does that job. At this point the phrase is closer to "electric" on a kettle: true of nearly everything in the category, and therefore useless for choosing between them.

What the label actually denotes

Underneath the phrase is almost always the same shape: a hosted vision model, general-purpose and not built by the tool in question, wrapped in a rubric and a scale that the tool did build. What that underlying model is actually doing is a real question, and it is largely the same answer across the category - most services are licensing or calling a small number of foundation models rather than training their own from scratch.

So when a tool says "AI-powered," it is telling you which layer is generic. It is not telling you anything about the layer that is specific to that tool: the rubric, the calibration, the scale, what gets reported back to you. That layer is where two tools built on the same model produce different numbers for the same photo, and the label does not touch it.

It helps to picture the two layers as separate teams doing separate jobs. One team trained a general vision model on a huge, unrelated corpus and exposed it as an API that any product can call - that team has no idea a rating tool exists, let alone what its rubric looks like. A second, much smaller team wrote the rubric, picked the scale, decided how to map a raw model output onto a number out of ten, and built the page you actually see. "AI-powered" is a fact about the first team's work. Every other thing you might want to know about the tool - whether it is honest, whether it is consistent, whether it is worth your time - is a fact about the second team's work, and the label is silent on all of it.

Why it stopped differentiating

A claim differentiates when some competitors have it and some do not. "AI-powered" cleared that bar for a couple of years and then stopped, once every tool in the category adopted a version of the same underlying approach. Once adoption is universal, the phrase moves from information to noise - present in every listing, meaningful in none of them, there because leaving it off would read as a gap rather than an omission.

The tell is what the phrase gets asked to do on a landing page. It sits next to "instant," "accurate," and "advanced," a cluster of words chosen for how they feel rather than what they claim. None of the four tells you the rubric, the scale, or what a result looks like before you hand over a photo.

What actually separates two tools

If the model layer is roughly interchangeable, the differences worth checking sit above it. Rubric, calibration, transparency, cost and privacy handling are the five that consistently move the experience, and none of them is visible from a badge that says AI-powered. A rubric is a good place to start precisely because it is checkable before you commit anything: does the tool report components or one number, and are those components named.

Rate Cock is one instance worth naming for exactly this reason - it reports six named axes rather than a single figure, which is a property you can verify by looking at a public entry rather than a claim you have to take on trust. That is the kind of detail "AI-powered" cannot substitute for, because it says nothing about what gets shown back to you once the model has run.

Running that check properly takes a few minutes and a short list of things to look at, rather than one thing - rubric, disclosure, consistency, and what the results page actually shows are all separate questions, and the label answers none of them.

Where the label still tells you something

There is one honest use of the phrase left: distinguishing an automated result from something with a different provenance entirely. A tape measurement is not "AI-powered" and never will be - it is a method applied by hand, and on that comparison the label is doing real, if narrow, work.

It also separates an automated result from a human one. A human reviewer is not running a model at all, so on that specific axis "AI-powered" versus "human-rated" is a real distinction rather than a marketing tic - it tells you who or what produced the number, even if it says nothing about the number's quality.

Everything else the phrase is asked to carry - accuracy, sophistication, being the newer or better option - is doing work it cannot do. The model choice matters less than what a tool built around it decided to show you, and that is a question the badge on the homepage was never going to answer.