AI Visibility Rank Tracker — ChatGPT, Perplexity, Claude & Gemini

Track how your brand ranks in AI answers over time. This actor queries ChatGPT, Perplexity, Claude, and Gemini with the prompts your customers actually ask, then measures whether your brand is mentioned, where it ranks versus competitors, its share of voice, whether your domain is cited, the sentiment of each mention — and how all of this changes run over run.

Built for Generative Engine Optimization (GEO): the new discipline of getting your brand recommended by AI answer engines. Schedule it daily or weekly to build a visibility history and catch ranking drops before they cost you traffic.

---

What it does

For every prompt × engine combination, the actor returns a single clean row containing:

• mentioned — was your brand named in the answer?
• position — your rank by order of first appearance among all tracked brands (you + competitors)
• share_of_voice — your % of total tracked-brand mentions in the answer
• cited / brand_cited_url — was your domain referenced as a source?
• citations — the source URLs the answer relied on
• sentiment — positive / neutral / negative framing of your mention
• competitors_mentioned — which rivals showed up alongside you
• position_change / visibility_status — the delta versus the previous run (new, gained, lost, improved, declined, unchanged, absent)

A run-level rollup (visibility score, average position, share of voice, gained/lost counts, per-engine breakdown) is saved to the key-value store under SUMMARY.

Export results, run via API, schedule and monitor runs, or integrate with other tools.

---

Why track AI visibility?

Buyers increasingly ask ChatGPT, Perplexity, Claude, and Gemini "what's the best tool for X?" instead of scrolling Google. If your brand isn't named — or is ranked below competitors — you lose the recommendation. Traditional SEO rank trackers don't see this. This actor does, across four engines, with trend history.

---

Input

Field	Type	Description
brandName (required)	string	The brand/product to track.
brandDomain	string	Your domain (e.g., notion.so) for citation detection.
industry	string	Used to auto-generate prompts when you don't supply your own.
prompts	string[]	The exact questions to monitor. Falls back to auto-generated prompts from industry.
platforms	string[]	Any of perplexity, chatgpt, claude, gemini.
competitors	string[]	Competitor brands for position & share-of-voice.
extractCitations	boolean	Extract source URLs and detect domain citations.
analyzeSentiment	boolean	Classify the sentiment of each brand mention.
trackingId	string	Stable ID used to compare history across runs (defaults to a slug of the brand).
useWebSearch	boolean	Enable OpenRouter's web plugin for non-native-search models.
webSearchResults	integer	Web results per query for the plugin (1–10).
models	object	Optional per-platform OpenRouter model-slug overrides.
maxPrompts	integer	Cap on prompts per run (cost control).

Example input

{
  "brandName": "Notion",
  "brandDomain": "notion.so",
  "industry": "productivity software",
  "prompts": [
    "What is the best note-taking app for teams?",
    "Recommend an all-in-one workspace tool."
  ],
  "platforms": ["perplexity", "chatgpt", "claude", "gemini"],
  "competitors": ["Evernote", "Obsidian", "Coda"],
  "extractCitations": true,
  "analyzeSentiment": true
}

---

API key (required) — one key for all models

All engines run through OpenRouter, so you only need a single API key. Set it as an environment variable on the actor (Console → Settings → Environment variables):

OPENROUTER_API_KEY = sk-or-...

If the key is missing, every check is reported with an error and not charged.

Platform	Default model	Web search
Perplexity	perplexity/sonar	native
ChatGPT	openai/gpt-4o-mini	OpenRouter web plugin
Claude	anthropic/claude-sonnet-4	OpenRouter web plugin
Gemini	google/gemini-2.5-flash	OpenRouter web plugin

Note: For non-Perplexity models, web grounding uses OpenRouter's web plugin (it adds live search results to any model) rather than each vendor's own search product — the right trade-off for one-key simplicity, and it still reflects current web data. You pay OpenRouter directly for model + web-plugin usage; the actor's PPE events cover the tracking/analysis layer. Override any model slug with the models input.

---

Output

Each row in the dataset:

{
  "brand": "Notion",
  "brand_domain": "notion.so",
  "engine": "perplexity",
  "model_used": "sonar",
  "prompt": "What is the best note-taking app for teams?",
  "mentioned": true,
  "position": 1,
  "share_of_voice": 60,
  "cited": true,
  "citations": ["https://notion.so/product", "https://example.com/review"],
  "brand_cited_url": "https://notion.so/product",
  "mention_count": 2,
  "mention_context": "Notion is one of the most popular all-in-one workspaces...",
  "sentiment": "positive",
  "competitors_mentioned": ["Evernote", "Obsidian"],
  "previous_mentioned": true,
  "previous_position": 3,
  "position_change": 2,
  "visibility_status": "improved",
  "error": null,
  "checked_at": "2026-06-07T12:00:00.000Z"
}

---

Pricing (Pay Per Event)

Charged events:

Event	When
actor-start	Once per run.
prompt-engine-check	Each successful prompt × engine check.
citation-extraction	Each successful check when citation extraction is enabled.
sentiment-analysis	Each mention scored when sentiment analysis is enabled.

Failed checks (e.g., a missing API key) are never charged. You pay your own LLM provider for the underlying API calls.

---

Scheduling for trend tracking

1. Run once to seed a baseline (first run shows visibility_status: new / absent).
2. Create a Schedule (daily or weekly) with the same trackingId.
3. Each run compares against the previous snapshot and fills position_change and visibility_status.

History is stored in a named key-value store (ai-rank-history-<trackingId>), so multiple brands can be tracked independently.

---

Use cases

• GEO / AEO monitoring — prove and improve how AI engines recommend you.
• Competitive intelligence — see who AI names ahead of you, and where.
• Content ROI — confirm whether new content gets you cited.
• Agency reporting — deliver weekly AI-visibility reports to clients.

---

Notes

• Results depend on each engine's live web access and model behavior; answers vary between runs, which is exactly what trend tracking is for.
• Keep prompts stable across runs for the cleanest deltas.