Reading the Visibility Panel

The Visibility tab is where a run turns into something you can act on. It always shows one run: the latest by default, or any past run picked from the run selector.

The four cards

Card	Range	What it is
Final Score	0–100	0.8 × average AI Score + 0.2 × SEO Score
AI Score	0–100	Mean AI Score of every scorable probe in the current scope
Mention Rate	%	Share of discovery conversations that mention you
Sentiment	0–100	Tone, averaged over answers that actually mention you

Every card has a small i with the short methodology; the long version is in scoring methodology.

Mention Rate carries a second line — + x/y in brand-named prompts — because comparison and trust prompts guarantee a mention and would inflate the headline if counted. Sentiment counts only answers that mention you, so that a brand nobody talks about doesn’t read as a brand everyone dislikes.

The scope selector

Above the tabs: All, one entry per market, and Global engines.

Scope	Shows
All	Every probe in the run
A market	Geo-adjustable engines probed there, plus fixed-region engines whose home market it is
Global engines	Engines that retrieve without a location signal

Switching scope recomputes every card and chart on the same formulas over that subset. Two labels you’ll see and should not read as bugs:

• site-level on the SEO half — your site is the same site everywhere, so this term can’t vary by market.
• A footnote naming engines that don’t vary by region, or that retrieve from one home market only.

See engines & regions.

Overview tab

Score history plots the Final Score and AI Score across runs, so you can see whether a change moved anything. Per-region history only exists for runs made after per-region tracking shipped; earlier runs fall back to the overall trend and the chart says so.

AI analysis & recommendations is generated per run: a short objective reading of where you stand against competitors, plus prioritised actions tailored to this brand. It’s generated across all regions rather than per region.

Discovery / Sentiment sub-tabs break the run down two ways:

• Discovery splits category (head terms) from scenario (long tail) prompts. Winning the long tail and losing the head is the most common shape, and the most actionable — head terms are incumbent-dominated, while scenario prompts sit closest to a buying decision. It also shows head-to-head outcomes for comparison prompts (a win rate) and the reputation verdict for trust prompts (no win rate — it isn’t a ranking question).
• Sentiment lists every answer graded by tone, brand-named prompts by verdict and discovery prompts by sentiment score. Click a row to open the full answer. Start with the negative band: a single hostile answer that’s repeated across engines is usually one bad source.

Competitors

Mention distribution — share of voice over discovery answers, you versus each competitor. Click a slice to see the conversations and the passages that named that brand.

Visibility leaderboard — you and every competitor, with a column per engine, as a percentage of conversations that named them. The bucket toggle switches between discovery and brand-named prompts.

Share of voice by model and sentiment by model show where the picture differs by engine — usually a retrieval difference rather than a preference.

Competitor names are resolved to parent brands before counting, so sub-brands and product lines don’t fragment into separate rows. See competitors.

Sources

Every external page the engines cited in this run, aggregated by domain, with the number of citations and how many prompts pulled it in.

Source breakdown classifies domains into Authority Media, Owned Assets and User Consensus. Cross-model cited sources lists domains cited by more than one engine — the highest-leverage targets in the product, because one change there propagates to several engines.

Exports to Excel, so the list can go straight to a content or PR team. See citations & sources.

Probes

Every probe in the run: engine, market, prompt, status, score, and a link into the detail page. Filterable by intent, and exportable to CSV.

Status values:

Status	Meaning
Mentioned	The answer referred to your brand
Missed	It didn’t — every score for this probe is 0
Won / Tie / Lost	Comparison verdict
Legit / Mixed / Flagged	Trust verdict
Not analyzed	Grading failed after retries — excluded from every metric

The probe detail page is the ground truth for any number you don’t believe: the exact prompt, the full answer, the score breakdown with each component, the competitors named, and every source cited. If a score looks wrong, read the answer — it is usually right and surprising.

Region Comparison

Available once you have results from at least two markets, and computed over geo-adjustable engines only so every row has the same prompts and engines behind it.

• Region scorecard — probes, score, delta against the cross-region average, mention rate, sentiment and top competitor per market. The average weights regions equally, not probes, so a market that lost a probe to a timeout doesn’t quietly lose influence.
• Mention rate by region — where you’re found unprompted.
• Competitor visibility by region — a rival that dominates one market and is absent elsewhere is a local player.
• Citation sources: shared vs region-specific — shared sources are global leverage; region-only sources are local content opportunities.

A reading order that works

1. Final Score, then AI Score If they diverge sharply, the gap is your site's technical health — go to the Site Audit.
2. Mention Rate Low means you aren't being found at all. That's a visibility problem, and sentiment is irrelevant until it's fixed.
3. Discovery split Category versus scenario tells you whether to fight for head terms or own the long tail.
4. Competitors Who is filling the answers you're missing from.
5. Sources Which pages those answers were built from — the actual lever.