Run a visibility scan

Measure how ChatGPT, Claude, Gemini, and AI Overviews perceive and recommend a brand. Trigger a scan, poll until it finishes, then read the score summary, the full report, historical trends, and the sources AI platforms cite.

When to use

• Reach for this when you want a fresh, authoritative visibility measurement for an agent you own (a Starter or Pro plan is required — the scan is a write plus LLM/SerpAPI spend).
• Scanning by url will create a new agent for the domain if one doesn't already exist; scanning by syntheticId re-scans an existing agent.
• After a scan, turn results into work with get_recommendations, and deep-dive a cited source with analyze_url.
• For a fast, public, shareable report without managing an agent, use the brand preview instead. For any brand's public score with no API key, use the brand score lookup.

Typical sequence

1. scan_visibility — trigger the scan. Returns immediately.
2. get_scan_status — poll every 30–60s until scanAvailable: true.
3. get_report — pull the full report (raw responses, query-level analysis, topics) for deep analysis.
4. get_cited_sources — list the URLs AI platforms cite for this brand.
5. get_scan_history — track score trends across scans to measure content impact.

Tools

scan_visibility

Trigger an AI visibility scan for a brand across the platforms on your plan: ChatGPT, Claude, Gemini, and AI Overviews on Starter, plus Perplexity on Pro. Provide either a website url (creates a new agent if needed) or a syntheticId (for an existing agent). Returns immediately — poll with get_scan_status.

Authentication: Paid (Starter or Pro). An account in read-only mode can't run a scan (it's a write plus LLM/SerpAPI spend) — the tool errors with a read-only message.

Type: Write — non-destructive.

Platforms by plan: Starter scans ChatGPT, Claude, Gemini, and AI Overviews; Pro adds Perplexity; enterprise plans can broaden coverage via platforms. The response's platforms array lists what the scan runs on. Platforms and the query cap follow the plan of the account that owns the brand. A requested platform your plan doesn't include is skipped (the scan still runs on your plan's platforms), and the response includes a droppedEnterprisePlatforms array plus an enterpriseNotice naming the plan each one needs.

Behavior notes:

• Scanning an existing syntheticId verifies ownership, checks credit balance, and — if a scan is already running — returns current progress (scanAlreadyRunning: true with a progress object) instead of starting a new one.
• Scanning a url extracts the domain, re-scans the matching agent if one exists, otherwise creates a new brand agent (website scraping, knowledge base creation, and the scan).
• scan_visibility never creates a public Brand Index page. If the brand already has one, it shows the newest scan.
• Errors when the existing agent's owner is on the free plan, even if your own plan is paid: the scan bills the owner, so the owner needs to upgrade to Pro first.
• Errors on insufficient credits (creditBalance, upgradeUrl). A free account needs the scan's full 50,000 credits before it starts; paid plans need a positive balance. Also returns a short maintenance message when new scans are paused for a platform upgrade (existing scans, get_scan_status, and get_report are unaffected).

get_scan_status

Poll the status of a visibility scan. Returns the full score summary when complete, or progress while running. Call every 30–60s after triggering.

Authentication: Required

Type: Read-only.

Key response fields:

The response no longer includes claimUrl; claim links were retired in October 2026.

While running, returns scanAvailable: false with scanInProgress: true and a progress object (completedQueries / totalQueries). Before any data exists it returns scanAvailable: false, scanInProgress: false with a "keep polling" message.

get_report

Get the full visibility report — the most detailed view available. Includes raw LLM responses, individual query analysis, cited sources, competitor data, topics, account info, and recommendations. Use after a scan completes for deep analysis.

Authentication: Required

Type: Read-only.

scanFailed behaves as in get_scan_status — true means the run reached no AI engine and the report is that run's real 0-score result, always reflecting the most recent attempt; re-run if you see it.

Compared to get_scan_status: get_report additionally returns query-level data (queryAnalysis), raw LLM responses (rawResponses), cited sources (citedSources), account info (balance, plan), and topics. Use get_scan_status for polling and quick checks; use get_report when you need per-query results, which sources AI platforms cited, or specific gaps to build content against.

get_scan_history

Get historical visibility scan scores to track trends over time and measure the impact of content you've created.

Authentication: Required

Type: Read-only.

Key fields: latestScore (most recent overall 0–100), scoreChange (difference between the two most recent scans, positive = improving), scoreTrend ("improving", "declining", or "stable"), and history (scan summaries, newest-first). Each history entry includes platformScores, personaScores, topCompetitors, citedSources, and queryAnalysis for detailed trend analysis. As with get_scan_status, a platformScores entry with queried: false means that scan didn't test that platform — its score is null, not a real 0.

get_cited_sources

Get the full list of URLs that AI platforms cite when discussing this brand, aggregated across all scans. Each source includes which platforms cite it, how often, an influence score, and which queries reference it. To produce a deep brief on a source, pass its URL to analyze_url.

Authentication: Required

Type: Read-only.

Returns a summary (total sources, opportunity count, how many mention you, how many are your own website) and a sources array with per-source platform, frequency, influence, and related-query data.

Key response fields:

Behavior notes:

• A page and its variants count as one source for pendiumContent: query strings (such as ?view=human), fragments, www. and trailing slashes don't change the label.
• If the brand's site settings can't be loaded, sources come back with pendiumContent: null instead of the call failing.