A brand mention means the brand's name appears in the text of the answer. A source citation means the answer was built from the brand's own content, which can happen even when the name is never mentioned. Rixe measures the two separately because they answer different questions: whether the model recommends you, and whether it trusts you as a source.
The two measurements combine into four states, and each one calls for something different. Mentioned and cited is the healthy case: the model recommends you and trusts your material. Mentioned but not cited means the model has an opinion about you assembled entirely from other people's writing — comfortable while the opinion is good, and very hard to correct when it is not.
Cited but not mentioned is the one that surprises teams. Your documentation or your blog is feeding the answer, but the brands named in it are your competitors. You are doing the work of establishing the category and someone else is collecting the recommendation. The fix is usually structural: content that explains a problem without clearly attaching your name to the solution.
Neither mentioned nor cited is the starting position for most brands in most categories, and the only one of the four where the first step is obvious: get into the sources the model already reads.
Four states, not two
Mentioned and cited is the healthy case: the answer recommends you and builds the recommendation on your own material. Mentioned but not cited means the reputation exists somewhere you do not control — good, but fragile, because you cannot edit the page that produced it.
Cited but not mentioned is the most common shape and the most fixable. A page of yours was useful enough to quote and your brand was not attached to the useful part — usually a wording problem on your own page rather than a decision the engine made. Neither mentioned nor cited is the honest starting position for most brands in most categories.
Why the cited-not-mentioned case is worth the work
When an engine quotes your page without your name, the material is already trusted; only the attribution is missing. That usually comes down to where the brand name sits on the page — a claim written as “the platform does X” instead of “Rixe does X”, a table whose header carries the product name and whose rows do not, a figure in a paragraph that never repeats who measured it.
It is also the case where a fix is testable. Change the wording, wait four weeks, ask the same questions again and watch whether the citation starts arriving with your name attached. That loop is short enough to run several times a quarter, which is not true of most work in this field.
When the fix is not on your site at all
If the sources feeding a question never mention you, no amount of rewriting will help, because your page is not in the pool the answer is drawn from. The work moves outward: a profile on the directory the engines actually cite, a contribution to the thread they keep quoting, a comparison article that includes you because someone was asked to consider it.
Both halves are measurable in the same table, which is the point of separating them. One column says whether the page did its job; the other says whether the category's sources carry you at all. A programme that only ever reads the first column will keep rewriting pages that were never the problem.
