Why we do / don’t

Why do you not report a synthesized content class?

We do not report a synthesized content class because we cannot prove it. Telling you a passage was synthesized rather than quoted requires knowing how an engine composed its answer, and no engine exposes that. An unprovable label would read as measurement while resting on a guess, so it stays unreported.

It is a tempting category. A dashboard cell that separates "the engine paraphrased you" from "the engine quoted you" would look like a deeper read of the same data.

The problem is that nothing in a collected answer distinguishes the two reliably. Wording similarity is not provenance — a paraphrase of your page and a paraphrase of a competitor page describing your product can look identical from the outside.

So the rule is the same one that governs the unknown state elsewhere in the product: if we cannot establish it, we label it as not established rather than inventing a class that sounds precise.

What this answer is based on

Reported measurement classes and what each one asserts