AEO Tracker: AI Search Visibility Monitor for ChatGPT, Gemini, Claude & Perplexity

Measure how often a brand appears, is recommended, and receives official citations in current AI search answers. This Apify Actor is built for answer engine optimization (AEO), generative engine optimization (GEO), AI SEO, brand monitoring, competitor research, and citation tracking.

The Actor sends the same neutral buyer-intent questions across four fixed, cost-conscious OpenRouter models representing ChatGPT, Gemini, Claude, and Perplexity. It returns comparable visibility results by answer engine without exposing model selection to end users.

What the AEO Tracker measures

• Brand mention rate across AI-generated answers
• Official website citation rate
• Mention count and sentence-level mention context
• Explicit recommendation position in ranked answers
• Positive, negative, neutral, or mixed brand sentiment
• Competitor mentions and competitive visibility signals
• Raw grounded answer text and web citations
• Per-engine visibility breakdown for ChatGPT, Gemini, Claude, and Perplexity

The run summary is stored in both OUTPUT and SUMMARY. It includes successful and failed checks, total visibility score, model list, and an engineBreakdown with mention rate, citation rate, and visibility score for every configured engine.

How it works technically

1. Provide a brand, optional official domain, industry, custom AEO queries, and competitor names.
2. The Actor uses supplied queries or generates neutral category questions from industry.
3. Each question is sent to the owner-controlled model matrix through the OpenRouter Chat Completions API.
4. OpenRouter's openrouter:web_search server tool grounds answers in current web results and returns standardized citation annotations.
5. The Actor extracts brand, domain, competitor, sentiment, recommendation-position, and citation signals.
6. One dataset row is written for each successful or failed query/model check, followed by the run summary.

This measures model-level AI answer visibility, not every user's personalized or logged-in ChatGPT, Gemini, Claude, or Perplexity interface. Keep the query set, model matrix, location, and timing stable when comparing runs over time.

Fixed model matrix

The model list is intentionally not a public input. The Actor owner controls it in src/main.ts so pricing and result quality remain predictable:

Engine	OpenRouter model	Role
ChatGPT	openai/gpt-4o-mini	Low-cost OpenAI baseline
Gemini	google/gemini-3.1-flash-lite	High-volume Google model with cost-controlled Exa grounding
Claude	anthropic/claude-haiku-4.5	Current efficient Anthropic model with web-search support
Perplexity	perplexity/sonar	Search-first model with citations

OpenRouter model availability and pricing can change. See the OpenRouter model catalog, web-search documentation, and the individual GPT-4o mini, Gemini 3.1 Flash Lite, Claude Haiku 4.5, and Perplexity Sonar pages for current provider pricing.

Configuration

Set OPENROUTER_API_KEY as an Apify Actor environment secret. It is never accepted as public input and is never written to the dataset.

Use queries for a fixed monitoring set. If omitted, three neutral questions are generated from industry. The default maxChecks is 12, which runs three generated queries across all four engines. Increase or decrease the cap deliberately: every completed query/model check is one billable AEO event.

Use searchContextSize to control the web-search context budget:

• low: lowest provider usage and suitable for routine monitoring
• medium: more grounding context for general research
• high: broadest context for complex categories, with higher provider usage

Apify pricing

The configured pay-per-event price is $0.01 per successful AEO answer check. A 12-check run can therefore generate up to $0.12 in Apify event charges, plus the platform's synthetic Actor-start charge. Failed OpenRouter checks are written with an error and are not charged as successful checks. OpenRouter model-token and web-search usage is billed separately to the configured OpenRouter account.

Example input

{
  "brandName": "Notion",
  "brandDomain": "notion.so",
  "industry": "productivity software",
  "queries": [
    "What are the best productivity software tools for a small remote team?",
    "Which productivity apps are easiest for teams to adopt?",
    "What should a startup look for when choosing productivity software?"
  ],
  "competitors": ["Asana", "Coda"],
  "maxChecks": 12,
  "searchContextSize": "low"
}

Dataset output

Each dataset record includes engine, modelRequested, modelUsed, query, mentioned, cited, recommendationPosition, sentiment, competitorMentions, citations, answerText, token usage, and any error diagnostic.

The summary's engineBreakdown makes cross-engine comparison straightforward:

{
  "engine": "Gemini",
  "model": "google/gemini-3.1-flash-lite",
  "mentionRate": 0.5,
  "citationRate": 0.25,
  "visibilityScore": 43
}

The overall score is a directional snapshot: 70% mention coverage plus 30% official-domain citation coverage among successful checks. It is not a universal share-of-voice metric. Inspect error before interpreting a missing mention or a partial engine result.

Local development

npm install
npm test
npm start

For a local run, set OPENROUTER_API_KEY and provide Apify input through the normal Actor storage environment. No cloud deployment or paid run is performed by the local build.

Scope and responsible use

This Actor is for legitimate brand, content, competitor, and answer-engine visibility research. It does not guarantee rankings, reproduce every personalized AI answer, or replace primary-source verification.