AI Search Visibility Tracker: ChatGPT, Perplexity, Claude, Grok

AI Search Visibility Tracker: brand mentions, citations and AI share of voice in ChatGPT, Perplexity, Claude & Grok

Find out whether ChatGPT, Perplexity, Claude and Grok recommend your brand, and track it over time. A generative engine optimization (GEO, also called AEO or LLM SEO) tracker for marketers, SEO agencies and founders. Give it the questions your customers ask AI assistants. For every answer, it reports whether your brand is mentioned, where it ranks against competitors, whether your website is cited as a source, and your share of voice.

It uses each provider's official API with your own API key. No scraping of chat apps, no shared accounts, no proxies. That makes it reliable. Requests run under your own provider account, so your provider's terms and usage policies apply to your prompts and to how you use the answers.

What you get

For each prompt × engine × run, one row:

Field	Example	Meaning
prompt	What is the best project management tool for a small team?	The question asked
engine	openai	openai, perplexity, anthropic or xai (Grok)
brandMentioned	true	Your brand or an alias appears in the answer
brandRank	2	Position of your brand among all tracked brands, in order of first mention
brandsMentionedInOrder	["Asana", "Trello", "ClickUp"]	Every tracked brand, in the order the AI named them
brandCited	true	The answer cites a page on your domain
brandCitationPosition	3	Position of your first cited page among the answer's sources
shareOfVoice	0.5	Your mentions ÷ mentions of all tracked brands
competitors	[{"name": "Asana", "mentioned": true, "rank": 1, "cited": true}]	The same metrics for each competitor
citations	["https://trello.com/guide", …]	URLs the AI cited
answer	Full answer text	Optional

A SUMMARY record in the run's key-value store gives, per engine: mention rate, citation rate, average rank, average share of voice and each competitor's mention rate. With Compare with previous runs on, it also shows changeSincePrevious, such as "mentionRate": 0.2 (up 20 points since the last run).

Sample output

The shape of one row, from the free demo run (shortened). The demo answer is illustrative sample text written for the demo, not a live AI answer; with your key, each row is the engine's real answer:

{
  "prompt": "What is the best project management tool for a small team?",
  "engine": "openai",
  "brand": "ClickUp",
  "brandMentioned": true,
  "brandRank": 4,
  "brandsMentionedInOrder": [
    "Trello",
    "Asana",
    "monday.com",
    "ClickUp"
  ],
  "brandCited": false,
  "shareOfVoice": 0.2,
  "competitors": [
    {
      "name": "Trello",
      "mentioned": true,
      "mentionCount": 2,
      "rank": 1,
      "cited": true
    },
    {
      "name": "Asana",
      "mentioned": true,
      "mentionCount": 1,
      "rank": 2,
      "cited": true
    }
  ],
  "citations": [
    "https://asana.com/compare",
    "https://trello.com/en/guide"
  ]
}

Use cases

• Brand monitoring in AI answers: see how often AI assistants recommend you for "best X for Y" questions, and which competitors they name instead.
• GEO / AEO reporting for clients: run one tracker per client on a weekly schedule and export the trend to Google Sheets or a dashboard.
• Content and PR impact: check whether a new comparison page, review or press mention changes your mention rate and citations a few weeks later.
• Citation gap analysis: the citations field lists every page the AI cited, so you can see which third-party sites (review sites, listicles, forums) shape the answer and target them.
• Competitor benchmarking: share of voice and rank against the competitors you name, per engine.

How to use it

1. Enter your brand, its domain, and a few competitors.
2. Add 5 to 50 prompts your customers would ask, such as "best CRM for a 10-person agency" or "alternatives to Asana".
3. Add an API key for at least one engine: OpenAI, Perplexity or Anthropic or xAI. Keys are stored encrypted by Apify and used only to call that provider.
4. Optionally set a country (for example US or CA) to localize web search. xAI's API has no location option, so Grok answers are not localized.
5. Run it. For tracking, add a schedule (daily or weekly) in Apify Console. Each run is compared with the previous one.

Tip: AI answers vary between runs. Set Runs per prompt to 3 to 5 for a reliable mention rate.

Pricing

You pay two things, and both are low:

• This Actor: a small fee per analyzed answer (see the Pricing tab), plus Apify's standard start fee of $0.00005 per run. Failed answers are never charged.
• The AI provider: billed directly to your own API account at their standard rates. The cost per answer depends on the engine, the model you pick and how much web content the search pulls in, plus each provider's web-search fee. Smaller models (for example gpt-4.1-mini, sonar, or a Claude Haiku or Sonnet model) cost much less than the largest ones. Check your provider's pricing page and start with a small run.

Set a maximum cost per run in Apify, and the Actor stops cleanly when it is reached.

Automate it: schedules and integrations

• Weekly tracking: in Apify Console, open Schedules, create a weekly schedule and add this Actor (or a saved task) to it. Every run is compared with the previous one, and the change appears in the SUMMARY record.
• Send results where your team works: under the Actor's Integrations tab you can connect Google Sheets, Slack, email, Make, Zapier, n8n or a webhook, so each finished run lands in a spreadsheet or channel without code.
• API: start runs and download results with the Apify API or the official Python and JavaScript clients. The API tab shows ready-to-copy examples.
• AI agents (MCP): connect Claude, Cursor or any MCP client to Apify's MCP server at https://mcp.apify.com?tools=stellalytics/ai-search-visibility-tracker and ask it to check your brand.

Related Actors

Tech Stack & Domain Lookup shows whether a site's robots.txt blocks GPTBot, ClaudeBot or PerplexityBot and whether it publishes an llms.txt file, for your site and your competitors'. If an AI engine can't read your pages, it is less likely to cite them.

Bulk PageSpeed & Core Web Vitals Checker bulk-tests your pages with Google PageSpeed Insights and shows whether real visitors pass Core Web Vitals, with a 6-month trend.

Why this instead of a chat scraper?

	This Actor	Chat-app scrapers
Data source	Official APIs	Automating the consumer chat website
Reliability	No CAPTCHAs, no login, no layout changes	Breaks when the site changes
Cost	Small fee + your API usage	Per-answer fee, no key needed

One honest limitation: the Actor asks each provider's API, not the consumer apps. The apps can use a different model than the one you pick here (the OpenAI default, gpt-4.1-mini, is chosen for low cost and is not the ChatGPT app's model), and they add their own search settings, personalization and memory. Treat results as a consistent, repeatable benchmark of what these AI engines answer, not an exact copy of one person's screen. Pick a larger model in the input to get closer to the apps.

FAQ

Can I try it without an API key? Yes. Run it with no key and it switches to demo mode: it analyzes illustrative sample answers (written for the demo, not live AI output) for your brand, marks every row with "demo": true and a note, and charges no answer fees (only Apify's $0.00005 start fee). Add a key for live results.

Which models are used? By default gpt-4.1-mini with web search, Perplexity sonar, claude-sonnet-5-5 with web search, and grok-4.3 with xAI's web search. You can change each model in the input.

Does it store my API keys? No. Keys are secret input fields that Apify stores encrypted. The Actor sends each key only to its own provider, never writes it to the log or the dataset, and removes anything that looks like an API key from provider error messages before logging them.

Is personal data collected? The Actor doesn't look for personal data. It sends your prompts to the providers you choose and saves their answers, which can name people if a prompt asks about them. Don't put personal information in prompts. Your provider's data policies apply to what you send.

Can I show the answers to my clients? Yes, but keep the sources. OpenAI and Anthropic require that answers based on web search are shown with their citations, so when you display answer to anyone, show the citations links with it.

What terms apply to Grok answers? Your own agreement with xAI and its Acceptable Use Policy, because the Actor calls the API with your key. Answers stay in your Apify account; the Actor does not share, pool or resell them.

Is this affiliated with OpenAI, Perplexity, Anthropic or xAI? No. ChatGPT, Perplexity, Claude and Grok are trademarks of their owners and are named only to say which APIs the Actor calls.

Can I track several prompt sets for the same brand? Yes. Give each tracker a different History name.

What happens if one key is wrong? That engine is skipped with a clear error in the log, and the other engines still run.

What if a run is about to hit its timeout? The Actor stops starting new answers a few minutes before the run timeout, saves the SUMMARY and history, and says so in the status message and in SUMMARY (stoppedBeforeTimeout, notProcessed). An answer still waiting for the provider at the last minute is dropped and not charged. For large runs, raise the timeout in the run options.

Support

Open an issue on the Issues tab. Fixes for broken runs are a priority.