AI visibility report

Produce a report a marketing lead can forward: the headline numbers for one brand, how they changed, which engines drive the change, which sources the engines lean on, and the three things to do about it — every claim backed by a number or a quoted answer.

When to use

• "How are we doing in ChatGPT / Perplexity / AI search?"
• A recurring (weekly, monthly, quarterly) AI visibility or AEO report.
• Before a planning session, to ground it in measured data.

Not for: a prompt-level fight with one competitor (competitor-gap), a source list for outreach (citation-sources), or a page audit (geo-audit).

Inputs

• Brand — a name; resolve it with list_brands. Ask if ambiguous.
• Window — days, default 30. For "what changed", compare to the previous window of the same length when the data allows (90-day analytics contain the daily series needed for a 30-vs-30 comparison).
• Audience — executive (short, decisions) or practitioner (full tables). Default executive.

Steps

1. Resolve the brand. list_brands → brand_id, brand website, tracked providers (list_providers if the dashboard does not list them).
2. Headline numbers. get_brand_dashboard(brand_id, days) for visibility, share of voice, sentiment, average position, citation counts. These are the five numbers at the top of the report.
3. Trend and per-engine breakdown. get_brand_analytics(brand_id, days) for the daily series and the per-provider split. Compute: the window average, the change versus the first half of the window (or the previous window), and the best and worst engine for visibility. Name the engine where the move happened.
4. Share of voice context. get_share_of_voice(brand_id, days) for the competitor table. Rank by current share of voice; note who gained and who lost.
5. Sentiment. get_sentiment(brand_id, days) — overall score and the positive/neutral/negative counts. If negatives exist, pull two or three with list_answers + get_answer and quote the sentence that carries the sentiment.
6. Citations. get_top_citations(brand_id, days) and list_cited_domains(brand_id, days, limit=15). Split into the brand's own domain, competitor domains, and third parties. The third-party list is where the engines are getting their facts.
7. Explain the biggest move. For the metric that changed most, list recent answers on the engine that moved (list_answers filtered by provider, and by date_from / date_to), open three to five with get_answer, and say what the engine did differently: a competitor it started naming, a source it started citing, a question it now answers without brands at all. For more than five answers, hand the ids to the answer-evidence-reader subagent (ships with this plugin) and use its digest, so the full answer texts never crowd the report.
8. Charts (when the client renders them): chart_brand_dashboard, chart_brand_visibility, chart_share_of_voice, chart_sentiment. Put them after the summary, not before.
9. Write the report using the template in references/report-template.md.

Output

Lead with a five-line summary block:

Brand · window · engines monitored
Visibility  42%  (▲ 6 pts vs previous 30 days)   best: Perplexity 61% · worst: Gemini 18%
Share of voice  27%  (#2 of 6; leader: Competitor A 34%)
Sentiment  +0.42  (31 positive · 12 neutral · 3 negative)
Citations  118 total · 9 to our own domain · top third-party: g2.com, reddit.com, techradar.com

Then, in order: What moved and why (evidence: quoted answers, named sources), Engine by engine (table), Competitors (table), Where the engines get their facts (domain table with own/competitor/third-party labels), Three actions (each with the metric it should move and how you would know within a month), Method (window, prompt count, providers, definitions in one line).

Numbers are quoted from the tool payloads, rounded to whole percentages. Say "no data" rather than inventing a comparison when a previous window does not exist.

Guardrails

• Read-only. This skill never changes prompts, competitors or content.
• Do not infer causality the data cannot show. "Visibility on Gemini fell after Gemini began citing competitor.com for 'best X'" is fine when the answers show it; "because of the algorithm update" is not.
• One brand per report. For a portfolio, run the skill per brand and let the user compare.
• Keep the executive version under 600 words plus tables.

Era tools used

• list_brands, list_providers (read) — resolution.
• get_brand_dashboard, get_brand_analytics, get_brand_visibility, get_share_of_voice, get_sentiment (read) — metrics.
• get_top_citations, list_cited_domains (read) — sources.
• list_answers, get_answer (read) — evidence.
• chart_brand_dashboard, chart_brand_visibility, chart_share_of_voice, chart_sentiment (read) — charts.
• get_glossary (read) — definitions, if the user asks what a metric means.

REST equivalents: GET /api/v1/brands/{brand_id}/dashboard?days=30, /analytics, /citations/domains, /citations/urls, /answers, /answers/{answer_id} with X-API-Key.

Without Era

Without monitoring data there is no report. Offer to run geo-audit on the brand's key pages instead, and explain in one sentence that visibility and share of voice come from running the brand's prompts against the engines on a schedule, which is what Era does.