AI Visibility Tracker: ChatGPT, Perplexity, Gemini & Claude (GEO)

Find out whether AI assistants recommend your brand or your competitors. This Actor asks ChatGPT, Perplexity, Gemini and Claude the questions your customers ask (with live web search on), then measures for every answer:

• ✅ Are you mentioned? How many times, and in what context
• 🥇 Where do you rank? Order of first mention and position in "top 10" style lists
• 🔗 Is your website cited as a source?
• ⚔️ Which competitors appear instead of you, and in what order
• 📊 Share of voice against the competitors you track
• 📚 Which sites the AI cites most: the publications and review sites to target for PR and content

You get a 0–100 AI Visibility Score, per-engine breakdowns, a competitor leaderboard, a list of content gaps (prompts where competitors show up and you don't) and a shareable HTML report you can send to a client or your boss.

This is GEO (Generative Engine Optimization), also called AEO (Answer Engine Optimization) or LLM SEO: the new SEO for AI search.

Who is it for?

• Marketing & SEO teams tracking how AI search changes their organic pipeline
• GEO / SEO agencies producing monthly AI visibility reports for clients
• Founders & product marketers checking whether ChatGPT recommends them or a competitor
• PR teams discovering which publications AI engines trust and cite
• Local businesses (set a city) checking whether AI suggests them for "best dentist in Madrid"-style questions

How much does it cost?

Pay only for answers analysed. No subscription.

Event	Price
ChatGPT answer analysed	$0.10
Perplexity answer analysed	$0.02
Gemini answer analysed	$0.05
Claude answer analysed	$0.18
Any answer using your own API key	$0.005
Visibility report (once per run)	$0.02

Example: 25 prompts × 4 engines = 100 answers ≈ $8.77. Weekly tracking of 25 prompts on ChatGPT + Perplexity costs about $13/month, versus $29–$500/month for typical GEO SaaS subscriptions.

Failed or declined answers are never charged. Set Maximum cost per run to cap spending; the Actor stops cleanly at the cap.

Bring your own keys: add your OpenAI, Perplexity, Google or Anthropic API key in the "Your own API keys" section to pay providers directly and pay only $0.005 per answer for the analysis. On Apify's free plan, Perplexity works out of the box (up to 3 answers per run) and the other engines need your own key.

How to use

1. Enter your brand and your website domain (to detect citations).
2. Add competitors, one per line, optionally with their domain: Asana | asana.com.
3. Add prompts: unbranded questions your customers ask, e.g. "What's the best CRM for a 10-person startup?". Not sure what to ask? Enter a topic and set Number of prompts to generate.
4. Pick the AI engines, the country (and optionally a city), then click Start.
5. Open the Visibility report (HTML) from the Output tab, or export the per-answer data as CSV, Excel or JSON.

Input example

{
    "brand": "Notion",
    "brandDomains": ["notion.so", "notion.com"],
    "competitors": ["Asana | asana.com", "ClickUp | clickup.com", "Trello | trello.com", "Monday | monday.com"],
    "prompts": [
        "What is the best project management tool for a small team?",
        "Best all-in-one workspace app for startups",
        "Alternatives to Jira for non-technical teams"
    ],
    "engines": ["chatgpt", "perplexity", "gemini", "claude"],
    "country": "US",
    "runsPerPrompt": 1
}

Output

Visibility report

The report.html record (and its JSON twin REPORT) contains:

• Overall: visibility score, mention rate, citation rate, average position, share of voice
• By engine: the same metrics for ChatGPT, Perplexity, Gemini and Claude
• Leaderboard: you vs every competitor, per engine
• Content gaps: prompts where competitors are mentioned and you're not
• Most-cited sources: the domains AI engines rely on, flagged when it's your site or a competitor's

One record per AI answer

{
    "prompt": "What is the best project management tool for a small team?",
    "engine": "chatgpt",
    "model": "gpt-6.1-sol",
    "brand": "Notion",
    "brandMentioned": true,
    "brandMentionCount": 2,
    "brandRank": 2,
    "brandListPosition": 2,
    "brandCited": true,
    "brandCitedUrls": ["https://www.notion.so/product"],
    "brandContext": "1. Asana – great for larger teams. 2. Notion – flexible docs + tasks (notion.so). 3. Trello – simple boards…",
    "shareOfVoice": 0.5,
    "mentionedInOrder": ["Asana", "Notion", "Trello"],
    "competitors": [
        { "name": "Asana", "mentioned": true, "mentionCount": 1, "rank": 1, "listPosition": 1, "cited": false },
        { "name": "ClickUp", "mentioned": false, "mentionCount": 0, "rank": null, "listPosition": null, "cited": false }
    ],
    "citations": [{ "url": "https://www.notion.so/product", "title": "Notion", "domain": "notion.so" }],
    "citedDomains": ["notion.so", "zapier.com"],
    "searchQueries": ["best project management tool small team 2026"],
    "answer": "Here are some of the best options: …",
    "checkedAt": "2026-10-06T19:55:12.000Z"
}

Track your AI visibility over time

AI answers change every week as models update and the web changes. Save your input as a task, add a weekly schedule, and send results to Google Sheets, Looker Studio, Slack or a webhook through Apify integrations. Use Runs per prompt: 2–3 for steadier percentages.

How it works

Each prompt is sent to the official API of each provider with live web search enabled: OpenAI's Responses API with web search, Perplexity Sonar, Gemini with Grounding with Google Search, and Claude with Anthropic's web search tool. The Actor then matches your brand, aliases and domains (whole-word, Unicode-aware), detects ranked lists, and resolves citations to domains.

Results closely track what people see in the consumer apps, but they won't be identical: consumer apps add personalization, memory, A/B tests and their own system prompts. For decisions, look at trends across many prompts rather than single answers.

Tips for good prompts

• Write prompts like a customer who doesn't know you yet: "best X for Y", "X vs Y", "alternatives to Z", "how do I solve [problem]".
• Mix head prompts ("best CRM") and long-tail prompts ("CRM for real estate agents in Spain").
• Keep the same prompt set over time so trends are comparable.
• If your brand is a common word (Notion, Linear, Monday), enable case-sensitive matching.

FAQ

Which models are used? By default gpt-6.1-sol (ChatGPT), sonar (Perplexity), gemini-3.8-flash (Gemini) and claude-opus-5-5 (Claude). You can change each model in the "Models" section, for example to claude-sonnet-5-5 or sonar-pro.

Can I use it for any language or country? Yes. Write prompts in any language, set the country code (and optionally a city) to localize web search, and optionally force the answer language.

Is my API key safe? Keys are stored encrypted as secret inputs and are only sent to the provider they belong to.

Can I call it from my own app or an AI agent? Yes: through the Apify API, the JavaScript/Python clients, Make, Zapier, n8n, or the Apify MCP server.

Support

Questions or feature requests (more engines, sentiment analysis, Looker Studio template)? Open an issue in the Issues tab — this Actor is actively maintained.