AI brand sentiment: Named is half of it. How AI describes you is the rest.

Both, on every answer that names you. Each one gets a label (positive, neutral or negative) and a score from minus 100 to plus 100. The number at the top of the card is the average of the scored answers rescaled to 0 to 100, shown next to how many answers it was averaged over and what share of them were positive.

Nothing is scored, and the card says so rather than showing a zero. Sentiment only exists where a brand is named, so an unmentioned brand has no sentiment to read. Head-to-Head is the exception: it still builds, from the answers that name the competitor, and that absence is what it reports.

From the answer text, not from anything you typed. Verand reads a window around each brand mention across the last thirty days, merges near-duplicates, and returns up to fifteen attributes, each with the number of answers that touched on it, a positive, neutral or negative reading, and a reason drawn from those answers.

Open it. Each negative attribute keeps the answers that said it and the pages those answers were reading, and comes with a recommended fix of two to four steps. Every step points at one of those pages: fix your own page in Page Tracker, see what someone else's page or thread says and what to tell its author, or, when none of your pages is behind it, create a brief for the page you don't have yet. It may be true, so the fix is context, not denial: nothing in it claims the opposite, it adds no number the answers and pages didn't give, and every step runs through your compliance packs before you see it. Verand doesn't send anything. The fix is rewritten only when the evidence behind it changes, and Regenerate asks for a fresh one. It needs answers that name you: a brand the assistants don't mention has no attributes yet.

The ones on your tracked list in the Brand Hub, up to ten, so you are compared with who you say you compete with rather than whoever an answer happened to name. Brands the assistants name alongside you that are not on that list are discovered separately and shown on the Visibility tab.

No. All of it recomputes after each scan over the trailing thirty days and is cached, so the tab is already filled in when you open it. The only buttons are bootstrap ones, for a site whose first scan has not produced anything to read yet, or a competitor nobody has compared you with.

Every phrase is grounded in the answers, and a pass that fails writes nothing instead of guessing: the panel stays empty and the next scan retries. It reports how you are described, not whether the description is correct. That check is Brand Facts, the panel beside it.