Overview

Overview is the landing screen of a workspace and the one to open first each morning. It compresses the selected window into five numbers, two charts, a competitive table and the raw answers behind them.

Brand or product scope

A scope toggle switches the whole screen between your brand and your products.

Under product scope the ranked rows are products rather than brands, and the KPI strip describes your best-performing product. This is not a cosmetic filter: a company can be well known while the product it is trying to sell is invisible, and only the product view shows it. See Products.

The KPI row

Grounded Answers is a data-quality gauge, not a performance target. When it falls, the engines are answering more from memory and less from the live web, which changes what any optimization work can achieve that week.

The noise floor

Each KPI carries a caption like noise floor ±2.1 pp · 3 runs/prompt, and a delta is marked significant only when it clears that floor.

This is the most important habit to build on this screen. AI answers are non-deterministic; the same prompt run twice can name different brands. Each prompt is therefore run several times per engine per day, and the spread of those runs sets a 95% confidence interval on every point.

The floor quoted next to a change is wider than the interval on a single point — comparing two independent estimates carries roughly 1.4× the single-point half-width. The interface always captions the change floor, never the narrower per-point interval, because the change is what you are about to act on.

A two-point move inside a ±2.1-point floor is noise. Treating it as a trend is how teams end up rewriting pages that were never the problem.

Visibility Trend

Your visibility plotted daily against every tracked competitor, per engine and reasoning mode, with a shaded 95% confidence band on your own series.

What to look for:

• Bands that do not overlap — a genuine separation from a competitor.
• Bands that do — a lead you cannot yet claim.
• Crossovers and cliffs — a sharp single-day drop is rarely your site; check whether an engine changed which sources it favors in Sources.

Citation Index

Whether the pages engines read actually name each tracked entity — plotted daily as the share of read-page citations landing on pages whose fetched text names them.

This is not visibility restated, and the difference is the point. Visibility counts being named in an answer; the Citation Index counts being named in the material behind it. High visibility over a low citation index means the position rests on the engine's memory rather than on anything currently published — the two moving apart is the finding. Full methodology in Citations and sources.

How the number is built:

• From fetched body text, never titles. After each day's collection, cited pages are fetched and read most-cited first, and top pages are re-checked every 14 days.
• Visibility prompts only, like every other number on this screen. A perception prompt names the brand, so the pages read for it name the brand by construction; counting them would move the share with the prompt mix rather than with anything published.
• A fixed 30-day window, unlike the rest of the page. Pages are read one at a time off a queue, so a short slice would sample the reading schedule rather than anything that happened.
• Coverage is stated, not implied. Reading is paid for a page at a time, so the read set always trails the cited set; the caption says N of M cited pages read rather than letting a share imply a census.

Under product scope the chart follows the toggle and plots product entities instead of brand and competitors.

Competitor table

One row per tracked entity, ranked by visibility, with each change marked significant or not on the same noise-floor basis as the KPI strip: rank, entity, visibility and its change, mentions, average position and its change, sentiment and its change.

Sort by change rather than absolute value when hunting for what moved — but check the significance marker before you build a plan on it.

Top Sources

The domains the engines leaned on most across your answer set, with use counts. Counted over visibility prompts only — perception probes name the brand on purpose and would drag in its own support and corporate pages — and it follows the scope toggle and the prompt-tag filter. The shortest route to a content plan: if directories and community threads sit at the top, review velocity beats another blog post. Full detail in Sources.

Recent Chats

Cards showing real captured answers: the engine, the prompt, the answer text as the engine rendered it, whether it was grounded, the brands named with their rank inside that answer, and the sources cited — marked according to whether each was quoted inline or merely listed.

Open one and read it. Five minutes of verbatim answers usually explains a week of metric movement, and it is the fastest way to catch an engine describing your product wrongly.

Filters

The date range is shared across Monitoring tabs and carried in the URL.

The prompt-tag filter scopes live data: KPIs, the visibility trend, the competitor table and Top Sources are recomputed over answers to prompts carrying any selected tag — and the denominator moves with the filter, so a filtered visibility is your share of the selected conversation, not of the day's whole run. The Citation Index and Recent Chats take no tags and keep their own populations; the banner above the dashboard says so whenever a tag is active.

Competitor-tag and platform filters are present but do not yet scope live data; they are captioned as sample-backed until the engine dimension is wired through. See Product status.