AI Visibility Monitor — Is ChatGPT recommending your brand?

More and more buyers ask ChatGPT, Perplexity and Gemini instead of Google. This Actor asks those AI engines the questions your customers ask and tells you:

• Appearance rate — in what % of answers your brand appears, measured over repeated runs of each question
• Position when mentioned — how high you tend to be listed when AI does recommend you
• Share of voice vs competitors — who AI recommends instead of you
• Cited sources — which websites AI trusts for your category, and whether your own site is one of them
• Sentiment — whether each answer describes you positively, neutrally or negatively
• Your cited pages — which URLs on your site AI engines use as sources
• Gaps — the exact prompts where competitors show up and you don't

Use it for GEO / AEO (generative engine optimization), AI brand monitoring, agency client reports, or competitor research.

Why check more than one engine?

Each AI engine searches the web differently, so they often disagree. In a test run (September 2026) we asked all three "¿Cuál es la mejor billetera virtual en Argentina?" (best digital wallet in Argentina), tracking Mercado Pago against Ualá and Naranja X:

Engine	Mercado Pago ranked	Sources it relied on
ChatGPT	1st	mercadopago.com.ar, uala.com.ar
Perplexity	1st	clarin.com, cronista.com, perfil.com…
Gemini	3rd — behind Ualá and Naranja X	infobae.com, lavoz.com.ar, cronista.com…

Checking only ChatGPT would have missed that Gemini recommends two competitors first — and the cited sources show which publications to target to change that.

How to use

1. Enter your brand and a few competitors.
2. Either describe your market in topic (e.g. "CRM for small agencies") and the Actor writes realistic buyer questions for you, or add your own prompts like "alternatives to HubSpot". Any language works.
3. Pick the engines and run. Schedule it weekly to track changes over time.

Each prompt is sent to each selected engine with live web search enabled, so results reflect what users see today.

Why repeated runs

Ask ChatGPT the same question twice and you'll often get a different list of brands. A single answer (and the rank inside it) is mostly noise. That's why the Actor asks every prompt several times per engine (runsPerPrompt, default 3) and reports the share of runs where you appear, per prompt and per engine. Use 5–10 runs for decisions that matter.

Example input

{
  "brand": "Pipedrive",
  "brandDomains": ["pipedrive.com"],
  "competitors": ["HubSpot", "Salesforce", "Zoho"],
  "topic": "CRM for small sales teams",
  "promptCount": 10,
  "prompts": ["HubSpot alternatives for agencies"],
  "engines": ["openai", "perplexity", "gemini"],
  "country": "US"
}

Generated prompts never include your brand name, so a mention is always earned. Generating prompts, sentiment analysis and the report are free; you only pay per answer checked.

Output

Dataset — one row per prompt × engine:

{
  "prompt": "What is the best CRM for a small sales team?",
  "engine": "openai",
  "brandMentioned": true,
  "brandPosition": 2,
  "brandDomainCited": true,
  "brandContext": "…HubSpot first, then Pipedrive, which is simpler for small sales teams…",
  "brandSentiment": "positive",
  "competitors": [{ "name": "HubSpot", "mentioned": true, "position": 1, "count": 2 }],
  "citedDomains": ["g2.com", "pipedrive.com"],
  "answer": "full AI answer…",
  "error": null
}

SUMMARY record (Key-value store) — the scorecard, overall and per engine, plus byPrompt (appearance rate for every prompt × engine across runs), sentiment counts and brandCitedPages (your URLs that AI engines cited):

{
  "brand": "Pipedrive",
  "overall": {
    "checks": 9,
    "brandMentionRate": 0.667,
    "brandAvgPosition": 1.5,
    "brandTop1Rate": 0.444,
    "brandDomainCitedRate": 0.333,
    "entities": [
      { "name": "HubSpot", "mentions": 8, "shareOfVoice": 0.4, "avgPosition": 1.2 },
      { "name": "Pipedrive", "isBrand": true, "mentions": 6, "shareOfVoice": 0.3, "avgPosition": 1.5 }
    ],
    "topCitedDomains": [{ "domain": "g2.com", "count": 5 }]
  },
  "byEngine": { "openai": { "…": "same fields" }, "gemini": {}, "perplexity": {} },
  "promptsWithoutBrand": [
    { "prompt": "HubSpot alternatives for agencies", "engine": "gemini", "competitorsMentioned": ["Zoho"] }
  ]
}

Download the dataset as JSON, CSV or Excel from the run page, or read the summary via API: https://api.apify.com/v2/actor-runs/{RUN_ID}/key-value-store/records/SUMMARY

Client-ready report

Every run also saves a formatted HTML report (REPORT in the key-value store, linked in the Output tab): a written summary with recommendations, appearance by engine, share of voice, a prompt-by-prompt table across runs, sentiment, and the sources AI engines cite. It's in English or Spanish (auto-detected from country), and agencies can add their name with agencyName to send it as their own. Open it and print to PDF.

Pricing

Pay only for answers checked — one check = one prompt on one engine. Failed requests are never charged.

	Price per check
Using our API keys (default)	$0.05
Using your own OpenAI / Perplexity / Gemini keys	$0.01

One check = one answer from one engine. Examples with our keys:

• 10 prompts × 3 engines × 1 run = 30 checks = $1.50
• 10 prompts × 3 engines × 3 runs (default) = 90 checks = $4.50
• Weekly tracking of 10 prompts × 3 engines × 3 runs ≈ $18 / month

You can set a maximum cost per run; the Actor stops before exceeding it.

Integrations

• Schedules — run weekly or daily from Apify Schedules to build a visibility trend.
• API — start runs and fetch results from any language with the Apify API.
• AI agents — call it from Claude, Cursor or any MCP client through the Apify MCP server.
• n8n, Make, Zapier — use the Apify integration to send results to Google Sheets, Slack or your client reports.
• Ready-made n8n workflow — AI Visibility Weekly Report template: runs this Actor every Monday, logs results to Google Sheets and posts a summary to Slack. 5-minute setup.

Tips

• Write prompts the way customers speak, and don't include your brand name in them — otherwise you'll always be "mentioned".
• 10–20 prompts × 3 engines × 3 runs gives a stable picture. Weekly schedules show the trend over time.
• Name variants matter: AI may say "Seguros Rivadavia" instead of "Rivadavia Seguros". The Actor auto-detects common variants (autoAliases), and you can add your own with brandAliases or Competitor | variant in the competitors list. The variants used are listed in SUMMARY.aliasesUsed.
• Positions are ranked among the brands you list (your brand + competitors). If AI recommends a brand you didn't list, it won't count — read answer on a few rows to spot competitors you're missing.
• Use country (e.g. US, AR, ES) to localize web search where the engine supports it.

FAQ

Which models are used? OpenAI Responses API with web search, Perplexity Agent API with web search, and Gemini with Google Search grounding. You can override models in the Advanced section.

Does it scrape chatgpt.com? No. It uses the official APIs, so it's stable and within each provider's terms. API answers can differ slightly from the consumer apps, which may add personalization.

Are my API keys safe? Keys you provide are stored as secret inputs and are only sent to the provider they belong to.

Why did an engine get skipped? If an API key is invalid or out of credits, that engine is disabled for the rest of the run and the remaining engines continue. The reason is in the error field.