ChatGPT Scraper & AI Search Brand Visibility Tracker

AI Search Answer Scraper asks ChatGPT, Google Gemini and Google AI Overviews your prompts, exactly as a real logged-out user would, and returns each AI answer as Markdown, the cited sources (URL, title, domain, position) and a brand visibility report: is your brand mentioned, how many times, in what order next to your competitors, and is your website cited.

It is built for GEO (generative engine optimization), AEO (answer engine optimization) and LLM SEO: track your share of voice in AI search, see which pages the AI engines cite for your category, and monitor competitors over time. No API keys, no ChatGPT account, no login.

What does the AI Search Answer Scraper do?

• ChatGPT scraper: types your prompt into chatgpt.com (logged out, with ChatGPT's own web search) and reads the full answer with every citation.
• Gemini scraper: asks gemini.google.com the same way and captures the answer plus the web sources Gemini grounded it on.
• Google AI Overview scraper: fetches the normal Google results page and extracts the AI Overview (the AI summary at the top), with its citations. Queries without an AI Overview are reported as such and are free.
• Brand mention tracker: for every brand you list (yours and your competitors), each answer row says mentioned, mentionCount, rankAmongBrands (1 = named first), citedDomain (your website is a source) and bestSourcePosition.

Every row is labelled source: "ui": the data comes from the public consumer interfaces, not from an API model that answers differently. That is what your customers actually see.

What data can you extract from ChatGPT, Gemini and AI Overviews?

🤖 AI engine	❓ Your prompt	📝 Full answer in Markdown
🏷️ Brand mentioned or not	🥇 Brand rank in the answer	📊 Share of voice
📍 Mention position score	🙂 Brand stance	🆚 Competitors mentioned
🔗 Your site cited as a source	📚 Cited sources (URL, title, position)	🌐 Cited domains
🔢 Citation count	🕒 Scraped at	⚠️ Status and errors

Why scrape AI answers instead of calling the OpenAI or Gemini API?

The API models (OpenAI Responses with web search, Gemini grounding, Perplexity Sonar) are different products from the chat apps: other system prompts, other search back-ends, other citations. For AI search visibility you want the answer a real person gets on chatgpt.com or google.com, so this Actor scrapes the real interfaces. It is also cheaper than most API setups for this job, because you pay one flat price per answer.

How to use it

1. Add your prompts the way customers ask them ("What is the best CRM for a small business?").
2. Pick the AI engines: ChatGPT, Gemini, Google AI Overview.
3. List your brands, one per line. Add aliases and the website after |: HubSpot | hubspot.com, Zoho CRM | Zoho | zoho.com.
4. Choose the country and language. Click Start.
5. Open the Answers and brand visibility, Brand mentions or Cited sources view, or export JSON, CSV or Excel.

Input example

{
    "queries": ["What are the best CRM tools for small businesses?"],
    "engines": ["chatgpt", "gemini", "google_ai_overview"],
    "brands": ["HubSpot | hubspot.com", "Salesforce | salesforce.com", "Zoho CRM | Zoho | zoho.com"],
    "country": "US",
    "language": "en"
}

Output: AI answer, sources and brand mentions

One row per prompt and engine:

{
    "query": "What are the best CRM tools for small businesses?",
    "engine": "chatgpt",
    "engineName": "ChatGPT",
    "status": "ok",
    "source": "ui",
    "answerMarkdown": "For most small businesses, I'd start by comparing **HubSpot, Zoho CRM, and Pipedrive**. ... ([Spotsaas](https://www.spotsaas.com/blog/...))",
    "answerText": "For most small businesses, I'd start by comparing HubSpot, Zoho CRM, and Pipedrive. ...",
    "sources": [
        { "position": 1, "url": "https://www.spotsaas.com/blog/best-crm-software-2026-reviews-pricing-feature-comparison", "title": "Best CRM Software 2026", "domain": "spotsaas.com", "citedInline": true },
        { "position": 4, "url": "https://www.hubspot.com/products/crm", "title": "Explore HubSpot", "domain": "hubspot.com", "citedInline": true }
    ],
    "sourceCount": 18,
    "citedDomains": ["spotsaas.com", "decisioncircuit.com", "hubspot.com", "zoho.com"],
    "brandMentions": [
        { "brand": "HubSpot", "mentioned": true, "mentionCount": 8, "rankAmongBrands": 1, "firstMentionOffset": 272, "citedDomain": true, "citedSourceCount": 3, "bestSourcePosition": 4 },
        { "brand": "Salesforce", "mentioned": false, "mentionCount": 0, "rankAmongBrands": null, "firstMentionOffset": null, "citedDomain": false, "citedSourceCount": 0, "bestSourcePosition": null }
    ],
    "brandsMentioned": ["HubSpot", "Zoho CRM", "Pipedrive"],
    "model": null,
    "country": "US",
    "language": "en",
    "note": null,
    "error": null,
    "attempts": 1,
    "durationSec": 30.3,
    "scrapedAt": "2026-10-24T10:02:11.000Z"
}

Field	Meaning
status	ok, no_answer (e.g. Google showed no AI Overview) or failed
answerMarkdown / answerText	The full answer, with citations kept as Markdown links / as plain text
sources[]	Cited or listed sources in the engine's order: position, url (tracking params removed), title, domain, citedInline
brandMentions[]	Per brand: mentioned, mentionCount, rankAmongBrands, citedDomain, bestSourcePosition
brandsMentioned	Brands named in the answer, in order of first mention
model	Model shown by the engine when it says (e.g. Gemini "3.5 Flash-Lite")

How much does it cost to scrape ChatGPT answers?

Pay per event, no subscription:

• $0.02 per AI answer ($20 per 1,000), for every engine.
• A small start fee per run.
• Failed prompts and Google queries without an AI Overview are free.

Tracking 50 prompts on 3 engines every week costs about $3 per week. That is more than 10x cheaper than tools charging $0.20 per search.

Supported AI engines

Engine	How	Notes
ChatGPT	Real browser, logged out, chatgpt.com	Uses ChatGPT's own web search when it decides to search, as for any visitor
Google Gemini	Real browser, logged out, gemini.google.com	Logged-out Gemini runs the Flash-Lite model with Google Search grounding
Google AI Overview	Google results page	Not every query triggers an AI Overview; those rows are no_answer and free

Perplexity, Microsoft Copilot and Google AI Mode now require a signed-in account or block automated visitors, so they are not supported yet.

Tips for AI visibility tracking

• Ask the same prompts on a schedule (daily or weekly) and chart rankAmongBrands and citedDomain over time. AI answers vary from run to run, so look at trends across several prompts, not one answer.
• Use several phrasings of the same intent ("best CRM for startups", "HubSpot alternatives", "cheapest CRM for a 5 person team").
• Check citedDomains to find the review sites, listicles and Reddit threads the AI engines trust in your niche; that is where GEO work pays off.
• Run it from different countries to compare local answers.

Run it on a schedule or via API

Create a task with your prompts and brands, then add a schedule in Apify Console (Schedules > Create). Results can go to Google Sheets, Slack, a webhook or any of the Apify integrations. You can also call the Actor from code with the Apify API or the apify-client package, or from an AI agent through the Apify MCP server.

FAQ

Is it legal to scrape ChatGPT and Gemini answers? The Actor only reads public answers that any logged-out visitor gets. It does not log in or bypass paywalls. Check the terms of each service for your use case and do not use the data to build a competing model.

Why do answers differ between runs? Generative engines are not deterministic, and ChatGPT decides per prompt whether to search the web. Track several prompts over time.

Official API vs this Actor? The OpenAI, Gemini and Perplexity APIs answer from different models and search stacks than the consumer apps, need API keys and are billed per token plus per search. This Actor returns what users see, for a flat price per answer.

What are the alternatives? Separate ChatGPT, Perplexity or AI Overview scrapers on the Apify Store, or SaaS GEO tools (Profound, Peec AI, Otterly). This Actor combines several engines and brand tracking in one run at a per-answer price.

Can I track Perplexity or Copilot? Not yet: both now require a signed-in session for automated visitors. Open an issue if you need them.

Related actors

• Video Transcript Scraper: transcripts from TikTok, Instagram, YouTube and X.
• Reddit Scraper: posts and comments, useful to see the Reddit threads AI engines cite.
• Amazon Product Scraper: product data for the brands AI engines recommend.
• Similarweb Scraper: see how much traffic ChatGPT, Claude and Perplexity actually send to a site.
• Google AI Mode Scraper: Apify's own scraper for Google AI Mode answers.