The Quattr Method
R4 · Represented, does AI describe your brand per your positioning?
Partially instrumented today, and why saying so is the point.
A prospect emails you a question that has your own product wrong in one specific way, and quotes an assistant as the source. You check. The assistant cited your page. It also described a version of you from two releases back, in the category's generic vocabulary, and pointed at somebody else at the end.
R4 · Represented is the fourth of the Quattr Method's five questions: does AI describe you per your positioning and recommend you with confidence? It sits a step above appearing in the answer at all. Being in an answer is not the same as being described accurately, and neither is the same as being recommended.
It is also the one question the method ships with a flag attached: partially instrumented today. That flag is not an apology. It is the point, and most of this article is about why a visible gap in the instruments beats a polished number pretending there isn't one.
Being cited is not being recommended
An assistant can name your page among its sources and still describe you a version behind, or hedge its recommendation until it means nothing.
Teams feel this one before they measure it. It is the screenshot somebody posts internally: the answer used us, and look at what it said. A citation, an AI answer that names one of your pages as a source, is necessary here and nowhere near sufficient.
The gap has two flavors and they fail differently. A model describing your product in the category's generic vocabulary is a clarity problem. A model that lists you and then steers its confident recommendation to a rival is a trust problem. The method tracks them as halves. The work that fixes each one is different work.
What can AI sentiment measure today?
The instrumented half is trust.
Positive-sentiment share, read the only way that stays honest: as a trend against your own trailing window, and as a gap against the competitors you track on the same prompts. Never against an absolute threshold, because nobody can tell you what a good score is.
The prompts matter as much here as anywhere. The sentiment number runs on the same first-party prompt baskets as the visibility number, meaning the set of questions replayed in AI engines to measure you, generated from what your buyers actually search for. Same denominator, same rules.
One check reports sentiment beside citations and share of voice per engine, so the trust picture arrives with the visibility picture. Share of voice is your slice of the AI answer space against tracked competitors. A citation counts for more there than a mention, which is an answer naming your brand without linking a page, and the exact weight appears in-product rather than in public.
The competitor gap keeps the number honest in both directions. A falling positive share during an industry-wide rough patch reads differently once you see every rival fell further. A rising one means little if the category rose faster than you did.
See also: reading AI sentiment without absolute thresholds →
The workflow that does this: AI platform scorecard →
What it can't measure yet, said plainly
The other half is clarity: tracking, systematically, whether AI's description of you matches your positioning. That instrumentation is on the roadmap. Until it ships, the method prints the flag on every surface where this question appears. Partially instrumented today.
Half-instrumented beats un-instrumented. It beats pretending by a wider margin than that. This industry is full of complete-sounding measurements of things nobody can fully measure yet, and asking what a packaged representation score actually reads usually collapses it into sentiment with better marketing. The smaller true claim is worth more.
The flag costs something in a demo and earns it back every week afterwards, because a set of questions that admits its gaps is a set whose other answers you can believe.
Ask it yourself
How is our positive sentiment trending across AI platforms, and where's the gap vs competitors?
The work that reaches Represented
No lever works this question as its primary target yet. Authority and off-site work moves it secondarily, because what third parties say about you feeds how models describe you. So does AI-answer visibility work, because the prompts you track are where a misdescription surfaces first.
Watching beats forcing here, because both failure modes get worked at other questions anyway. A clarity gap usually traces back to the vocabulary on your own pages. That is refresh and net-new work. A trust gap usually traces to what third parties have on record, which is authority work. This question tells you which of the two to fund.
So this is the question to watch rather than the one to force. Read the trust half with the rules intact, wait for the clarity half's instruments, and let the flag do its quiet work of proving the other four answers tell the truth. The Quattr Method page carries the full grid, and the flag is printed on it there too, partially instrumented today.