AI Search Visibility Tracker: ChatGPT, Gemini, AI Overviews

See how AI assistants answer the questions your buyers ask, and whether they name your brand.

For each prompt (for example "What is the best CRM for a small business?"), the actor asks ChatGPT, Gemini and Google's AI Overview and returns one row per answer with:

• Which of your brands were named, how often, and in what order (position 1 = named first)
• Which sites the answer cited, and whether your own domains were among them
• The full answer text (optional)

The run summary adds share of voice for every brand, overall and per engine: the share of answers that mention it, its average position, and how often it was named first.

Use it to track generative engine optimisation (GEO) over time, benchmark against competitors, or check which publications AI answers rely on in your category. Schedule it weekly on Apify to build a trend. No accounts, API keys or logins needed.

How much does it cost?

Event	Price	When
answer-checked	$0.03	One captured AI answer (one prompt on one engine), analysed and delivered

10 prompts on all three engines cost at most $0.90 per run. Apify also bills its start event (apify-actor-start): $0.00005 per GB of run memory, so $0.0001 per run at the default 2 GB.

You are not charged when an engine blocks, times out or replies with an error instead of an answer, when Google shows no AI Overview for a prompt, when an overview could not be captured, or for duplicate prompts. Every such outcome is listed in the free status dataset.

Input

Field	Example	Notes
prompts	["What is the best CRM for a small business?"]	Up to 25 per run, 500 characters each
brands	["HubSpot | HubSpot CRM", "Salesforce", "Zoho"]	Your brand and competitors, up to 20. Add spelling variants after |
engines	["chatgpt", "gemini", "google-ai-overview"]	Any subset
trackedDomains	["hubspot.com", "zoho.com"]	Optional: domains to look for in citations (subdomains count)
includeAnswerText	true	Turn off for smaller datasets

Brand matching ignores case and matches whole words only, so "Zoho" does not match "Zohodesk".

Output

{
  "engine": "chatgpt",
  "prompt": "What is the best CRM for a small business?",
  "brandsMentioned": ["HubSpot", "Zoho", "Pipedrive"],
  "firstBrandMentioned": "HubSpot",
  "brandResults": [
    { "brand": "HubSpot", "mentioned": true, "mentions": 2, "position": 1 },
    { "brand": "Salesforce", "mentioned": false, "mentions": 0, "position": null },
    { "brand": "Zoho", "mentioned": true, "mentions": 1, "position": 2 },
    { "brand": "Pipedrive", "mentioned": true, "mentions": 1, "position": 3 }
  ],
  "citedDomains": ["pcmag.com", "zapier.com"],
  "trackedDomainsCited": [],
  "citations": [{ "url": "https://www.pcmag.com/picks/the-best-small-business-crm-software", "domain": "pcmag.com" }],
  "answerChars": 3284,
  "answerText": "For most small businesses, I recommend HubSpot CRM…",
  "checkedAt": "2026-10-16T09:00:00.000Z"
}

Google AI Overview rows also carry sourceNames: the source labels and link titles Google shows with the overview (for example "PCMag" or "Best CRM for Small Business in 2026 | Slack").

The share-of-voice summary is saved as OUTPUT in the run's key-value store.

Honest limits

• Logged-out answers from a US data-centre address. ChatGPT and Gemini are asked in a fresh, logged-out browser session, with no history or personalisation. Signed-in users, other countries and paid plans can see different answers.
• Answers vary. The same prompt can get a different answer an hour later. Track trends over several runs rather than reading too much into one answer.
• Gemini answers have no links when logged out, so Gemini rows have no citations.
• Google AI Overview: US English results only. The answer text is the part of the overview Google shows before "Show more"; brand mentions are counted in that text, while citations cover every source attached to the overview. When Google hides a source's article link behind an encrypted redirect, the citation is the publisher's homepage: the domain is exact, the page is not. For some prompts Google builds the overview in the browser after the page loads; those are reported as ai_overview_not_captured and are free.
• Not covered: Perplexity and Google AI Mode. Both require sign-in or block automated access.
• Speed: each ChatGPT or Gemini answer takes about 30 seconds. The engines run in parallel, so 25 prompts take roughly 15–20 minutes.

Telemetry

When telemetry is configured, each run records anonymous usage events to the developer's own storage: a salted hash of the caller account ID (never the raw ID), whether the caller is on a paid plan, the run ID, the run origin (e.g. web console vs API), timestamps, and the number of charged events. No prompts, brands, answers or other inputs and results are collected; this only measures adoption (distinct and returning users) and revenue.