AI Brand Visibility Tracker: ChatGPT, Perplexity, Gemini, Claude

🔦 What is the AI Brand Visibility Tracker?

When a buyer asks an AI assistant "what's the best X?", does it name your brand? In what position? Does it link to your site, or to a competitor's?

This Actor asks your buyer questions to several AI models through their official APIs (with live web search on), then measures for every answer:

• Mentioned or not, how many times, and the sentence it appears in
• Rank among brands: 1 = your brand is the first brand named in the answer
• List position: "#2 in the list" when the answer is a numbered or bulleted list
• Cited: whether your domain is among the sources the model used
• Share of voice against the competitors you list
• Sentiment and a one-line summary of how the answer describes you (optional)
• Other brands the AI recommends that you didn't list, so you find competitors you weren't tracking

Plus a summary per model and per prompt, the most cited domains (the pages AI models are learning from in your niche), and the change since your last run, so a weekly schedule becomes a visibility trend line.

No scraping of chatbot websites, no logins, no API keys required.

🎯 Why use it

• GEO / AEO reporting: share-of-voice numbers for clients or your own brand, broken down by model.
• Content strategy: the topCitedDomains list shows which review sites and publications the models trust for your category. Get featured there.
• Competitive intelligence: see which competitors AI assistants push, and how they describe them.
• Launch tracking: schedule it weekly and watch mention rate move after a PR push or new content.

⚙️ How it works

1. Each prompt is sent to each selected model runsPerPrompt times. AI answers vary, so 2-3 runs give steadier rates.
2. With web search on, OpenAI, Gemini, Claude and Grok use their providers' native search and Perplexity Sonar uses its own, so answers and cited sources come close to what users see in those apps. Turn it off to test what models "know" from training alone.
3. Mentions are matched on the brand name, your aliases and your domain. Matching is word-boundary aware ("On" won't match "Online") and handles straight and curly apostrophes.
4. An optional second pass with a small, cheap model reads each answer for sentiment and other brand names.

Models are reached through OpenRouter, either through Apify's own OpenRouter proxy, billed to your Apify account, or through your own OpenRouter key if you add one.

⬇️ Input

Field	What to enter
brand	Allbirds | allbirds.com. Add aliases and domains after a |.
prompts	Real buyer questions without your brand name, e.g. "best wool shoes for travel".
competitors	Veja | veja-store.com, one per line (max 25).
models	Pick from OpenAI GPT-5 mini / GPT-5.1, Perplexity Sonar / Sonar Pro, Gemini 2.5 Flash, Claude Haiku 4.5, Grok 4 Fast, DeepSeek, Llama. Add any other OpenRouter model ID in customModels.
webSearch	On by default.
location	Optional market, e.g. "United Kingdom".
runsPerPrompt	1-5.
aiAnalysis	Sentiment and unlisted-brand discovery (on by default).
maxLlmCostUsd	Hard stop for model spend per run (default $2).
trackChanges	Compare with the previous run of the same brand, prompts and models.
openRouterApiKey	Optional. Your own key; stored encrypted as a secret input.

Example:

{
  "brand": "Allbirds | allbirds.com",
  "prompts": ["What are the best sustainable sneakers for everyday wear?", "Most comfortable wool shoes for travel"],
  "competitors": ["Veja | veja-store.com", "Rothy's | rothys.com", "On Running | on.com"],
  "models": ["openai/gpt-5-mini", "perplexity/sonar", "google/gemini-2.5-flash", "anthropic/claude-haiku-4.5"],
  "runsPerPrompt": 2,
  "location": "United States"
}

⬆️ Output

One dataset row per prompt × model × run (excerpt from a real run, answer shortened):

{
  "prompt": "What are the best sustainable sneakers for everyday wear?",
  "model": "google/gemini-2.5-flash",
  "modelLabel": "Google Gemini 2.5 Flash",
  "webSearch": true,
  "status": "ok",
  "brandMentioned": true,
  "brandMentionCount": 3,
  "brandRank": 1,
  "brandListPosition": 1,
  "brandCited": false,
  "brandSentiment": "positive",
  "brandDescription": "Praised for comfort and eco-friendly materials like merino wool; Certified B Corp.",
  "shareOfVoice": 0.3333,
  "entitiesInOrder": ["Allbirds", "Rothy's", "Veja"],
  "competitors": [{"name": "Veja", "mentioned": true, "mentionCount": 2, "rank": 3, "listPosition": 4, "cited": false}],
  "otherBrandsMentioned": ["Cariuma", "Thousand Fell", "Native Shoes"],
  "citations": [{"url": "https://www.thegoodtrade.com/...", "domain": "thegoodtrade.com", "title": "..."}],
  "answer": "Here are some of the best sustainable sneakers...",
  "llmCostUsd": 0.0104,
  "checkedAt": "2026-10-08T14:49:40Z"
}

The SUMMARY record in the run's key-value store contains:

• overall, byModel, byPrompt: mention rate, top-3 rate, average rank, citation rate, share of voice, sentiment counts and the same metrics for every competitor
• topCitedDomains, otherBrandsMentioned
• changeSincePreviousRun: deltas versus your last comparable run

Failed checks appear as rows with status: "error" and are never charged.

💰 Cost

• Actor: $0.02 per successful prompt check, pay-per-event (see the pricing table below). Failed checks are free.
• AI model usage is passed through at the model provider's price via Apify's OpenRouter proxy (or billed by OpenRouter if you use your own key). A typical check with web search costs about $0.005-0.015 in model usage, and the optional analysis pass adds about $0.0003. The run log and every row show the exact model cost.
• Note: Apify's OpenRouter proxy charges accounts on the free Apify plan 10x the model price. If you're on the free plan and run more than a handful of checks, add your own OpenRouter key.
• maxLlmCostUsd stops the run once model spend reaches your limit.

💡 Tips

• Use 10-30 prompts that match real buying intent ("best X for Y", "X vs Y", "alternatives to Z"). Avoid putting your own brand in the prompt unless you're testing brand perception specifically.
• Run 2-3 times per prompt before drawing conclusions. Single answers are noisy.
• Schedule weekly with the same input to build a trend. changeSincePreviousRun fills in from the second run.
• These are answers from the models' APIs. Consumer apps (ChatGPT, Gemini app, Perplexity.ai) can add personalization and other features, so treat results as a strong, repeatable proxy, not a pixel-perfect copy of any one user's screen.

🔌 Use it via API, Python or integrations

Run it from the Apify Console, on a schedule, or from code with the Apify API, the Python or JavaScript apify-client, or MCP. Connect results to Google Sheets, Slack, Zapier, Make, n8n or a webhook from the Integrations tab.

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("lanternlane-data/ai-brand-visibility-tracker").call(run_input={"brand": "Allbirds | allbirds.com", "prompts": ["What are the best sustainable sneakers?"], "competitors": ["Veja | veja-store.com"]})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

❓ FAQ

Is this allowed? Yes. The Actor only calls official, paid model APIs under their standard terms. It doesn't scrape chatbot websites or automate logins.

Which permission level does it need? Limited permissions. It reads and writes only its own run storage and one named store for run-to-run comparison.

Can AI agents use it? Yes. The input is plain JSON and the output is structured, so it works well from the Apify MCP server or the API.

🏮 More from Lantern Lane Data

Jobs & hiring data is our focus: job postings straight from employers' applicant tracking systems and official public feeds, in one consistent schema. More ATS platforms and hiring-signal tools are on the way.

• 💼 ATS Jobs Scraper: Every open job from 7 ATS platforms (Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Workable, Recruitee) in one schema, with new and closed job alerts.
• 🌍 Remote Jobs Scraper: Remote jobs from 7 boards via their official public feeds, deduplicated, with salary and location filters.

Other tools

• 🛍️ Shopify Products Scraper: Every product, price, variant and stock level from any Shopify store, with price-change tracking.
• 🗺️ Sitemap URL Extractor: Every URL from any website's XML sitemaps, with lastmod, hreflang and new-page alerts.

---

Built and maintained by Lantern Lane Data. Questions or feature requests: open an issue on the Actor's Issues tab.