AI Visibility Tracker: ChatGPT, Gemini & Google AI Overviews

When your customers ask AI "what's the best ___?", does it recommend you, or your competitors?

This Actor asks ChatGPT, Google Gemini and Google AI Overviews the questions your buyers ask, and tells you:

• 📈 Your AI visibility score (0-100), overall and per engine: how often AI names you, and how high in its answer
• 🥊 Share of voice against your competitors: who AI recommends first in your category, engine by engine
• 🕳 Gaps: the exact prompts where AI names your competitors and leaves you out
• 🎯 Sites to get featured on: the review sites, directories, listicles and forums AI cites when it recommends competitors instead of you. That's your PR and outreach list
• 🔗 Which of your own pages AI cites, and the sources AI relies on most in your niche
• 📅 Changes over time: schedule it weekly and see your score trend, plus the prompts you won or lost since last run
• 📄 A client-ready HTML report, white-label for agencies ("Prepared by your agency", no Apify branding)

Every answer comes from the real, public, logged-out ChatGPT, Gemini and Google search, as a new visitor sees them. They are not from an API model that answers differently from the product your customers use. You don't need an OpenAI or Google account or any API key.

Who it's for

• Marketing teams and founders who want to know whether AI search sends buyers to them or to a competitor
• SEO and GEO / AEO agencies reporting AI visibility to clients: one scheduled task per client, a branded HTML report each week
• PR and content teams who need the list of publications and pages AI trusts in their category
• Local businesses and franchises: "best dentist in Brisbane", "plumber near Austin", with the country set to where the customers are

How to use it

1. Your brand: name, other names, website, e.g. Pipedrive, Pipedrive CRM, pipedrive.com.
2. Competitors: one per line, same format.
3. What to ask: type a category ("CRM for small business") and the Actor writes the buyer questions for you, or paste your own prompts, or both.
4. Pick the engines (all three by default) and the country your customers are in.
5. Optional: set a project name and schedule the Actor weekly to track changes.
6. Run it. Open the Visibility report (HTML) from the Output tab, or use the JSON report and the dataset in your own dashboards via the API, Google Sheets, Zapier or Make.

The report

Section	What it tells you
AI visibility score	100 if you're the first tracked brand named, 85 second, 70 third, 55 fourth, 40 lower, 0 if not named, averaged over every answer. Overall and per engine
Who AI recommends	Every tracked brand's score, share of voice, % of answers naming it, % ranked first, average position, % citing its site
Prompt by prompt	A prompt × engine grid with your position in each answer
Gaps	Prompts where an engine named competitors but not you, and who it named
Sites to get featured on	Domains cited in answers that recommended a competitor and not you, with how often and for whom
Your pages AI cites	Which of your URLs AI uses as a source, and on which engines
Sources AI relies on most	The top cited domains across all answers
Since the last run / Score over time	With a project name: score changes and prompts gained or lost

Prompts that name a tracked brand ("Is HubSpot any good?") are answered and shown, but left out of the scores, since an answer to them names the brand whatever its real visibility is.

Output

Dataset: one row per prompt per engine (views: Brand visibility, Answers & sources, Every tracked brand):

{
  "prompt": "What is the best CRM for small business?",
  "engine": "gemini",
  "engineName": "Gemini",
  "status": "answered",
  "brandedPrompt": false,
  "yourBrand": "Pipedrive",
  "yourBrandMentioned": true,
  "yourBrandRank": 2,
  "yourBrandCited": false,
  "brandsMentioned": ["HubSpot", "Pipedrive", "Salesforce"],
  "brands": [
    {
      "brand": "Pipedrive",
      "mentioned": true,
      "mentions": 2,
      "rank": 2,
      "firstMention": "Pipedrive: Best for sales-focused teams who want a clean, visual drag-and-drop pipeline to track deals from initial contact to close.",
      "cited": false,
      "citedUrls": []
    }
  ],
  "answerText": "Choosing the \"best\" CRM for a small business depends heavily on your budget, technical setup, and primary goals...",
  "answer": "(the answer as markdown)",
  "sources": [
    {"title": "Best small business CRM of 2026 - TechRadar", "url": "https://www.techradar.com/best/best-crm-for-small-business",
     "domain": "techradar.com", "siteName": null, "citedInline": true}
  ],
  "sourceCount": 8,
  "citedDomains": ["techradar.com", "pcmag.com", "g2.com", "zendesk.com"],
  "askedAt": "2026-10-01T14:11:36Z",
  "seconds": 13.5,
  "error": null
}

status is answered, not shown (Google showed no AI Overview for that search, which is itself useful to know) or failed (the engine wouldn't answer after retries). Only answered rows are charged.

Key-value store: REPORT.html (the report above) and REPORT (the same as JSON: visibilityScore, scoreByEngine, brands, gaps, citationOpportunities, yourCitedPages, topCitedDomains, coverage, history, trend).

Pricing

You pay per AI answer delivered. Searches where Google shows no AI Overview, failed prompts and the report are free. Example: 10 prompts on 3 engines is at most 30 answers. Set Max cost per run in the run options to cap any run.

Good to know

• AI answers vary. The same question can name different brands each time. For steadier numbers, track 10-30 prompts and/or ask each 2-3 times (Ask each prompt this many times).
• Google AI Overviews appear only for some searches; the report shows how many did (coverage). They're searched for the country you pick.
• ChatGPT is asked with web search, from the chosen country. For countries other than the US it goes through Apify residential proxy, which your Apify plan must include.
• Gemini answers logged-out visitors from the server's own location (the US). It doesn't localise by country setting.
• Speed: about 10 prompts × 3 engines in 3-5 minutes. Google AI Overviews are asked one at a time and slowly, to stay within Google's limits.
• If Google temporarily refuses the server ("unusual traffic"), the Actor pauses that engine and tries again. If it's still refused, the remaining prompts for that engine are marked failed and aren't charged.

FAQ

Why not use the OpenAI or Gemini API? API models don't browse the way the consumer apps do, and they recommend different brands. What matters is what your customers see on chatgpt.com, gemini.google.com and google.com, and that's what this Actor reads.

Can I track several clients? Yes. Create one task per client with its own project name, agency name and schedule.

Can I get the data into Looker Studio or Sheets? The dataset and the JSON report are available through the Apify API and integrations (Google Sheets, Zapier, Make, webhooks).