Google AI Mode Scraper & API ($4.00/1k) 🤖

🤖 Google AI Mode Scraper & Citation API: Conversational Answers in Markdown, Cited Web Sources, Follow-Ups & GEO Brand Audits

---

---

📖 Overview

Google AI Mode is Google's dedicated conversational AI search experience — decomposing complex multi-part B2B comparisons, technical architecture trade-offs, financial regulation questions, and buyer research prompts into deep, structured answers backed by live web grounding.

Whereas classic Google AI Overviews appear selectively above traditional SERPs, Google AI Mode generates rich conversational responses on virtually every informational, comparative, and analytical prompt — making it the #1 frontier for Generative Engine Optimization (GEO), LLM brand share-of-voice tracking, and web-grounded RAG dataset curation.

Until now, collecting Google AI Mode responses at scale meant paying $150 to $200 per 1,000 prompts for slow, memory-hungry headless browsers that frequently stall on verification walls, mangle Markdown tables and code blocks, or leak opaque Google redirect links.

Google AI Mode Scraper by UnitBytes runs on a tiny 256 MB memory footprint with zero browser overhead — delivering complete Markdown answers, unwrapped canonical source URLs, follow-up prompts, and brand citation audits starting from $4.00 per 1,000 queries (50x cheaper), while charging $0.00 in query events if a prompt targets an unsupported region.

⚡ UnitBytes · Google AI, Search, News, Trends, Shopping & Local Intelligence Ecosystem
🤖 Google AI Mode Scraper 📍 You are here (From $4.00/1K) Deep conversational AI Mode answers, web sources & GEO brand audits	✨ Google AI Overviews Scraper From $4.00/1K • SERP + AEO Classic SERP AI Overviews + Page 1 organic ranks & People Also Ask	📈 Google Trends Scraper From $0.30/1K • Search Demand Breakout topics, interest over time, regional heatmaps & rising queries
📰 Google News Scraper From $1.00/1K • PR & Grounding Direct publisher URLs, recency filters & clean full-article text	🛍️ Google Shopping Scraper From $2.00/1K • Product Offers Live retail prices, deals, multi-seller offers & store checkout URLs	📍 Google Maps & Jobs From $0.25/1K • Local & Hiring B2B Leads, Reviews & Google Jobs

---

🌟 Why GEO Platforms, Brand Trackers & AI Teams Choose UnitBytes

Capability / Metric	UnitBytes Google AI Mode Scraper	Browser-Based AI Mode Scrapers	Generic SERP APIs
💰 Price per 1,000 Prompts	From $4.00 ($0.0040 / prompt — 50x cheaper)	$150.00 – $200.00 per 1K + heavy GB-hour compute	$15.00 – $30.00 per 1K
📝 GFM Tables, Code & LaTeX Fidelity	✅ Multi-col GFM tables, syntax-tagged code (python, sql) & $LaTeX$	⚠️ Collapsed whitespace or leaked UI disclaimers	⚠️ Often flattened to plain text
🔗 Inline citations vs. sources	✅ Explicit section: "cited" vs "related" + unwrapped canonical URLs	❌ Lumps sidebar cards together or leaves redirects	⚠️ Partial URL resolution
💬 Conversational followUpQuestions	✅ Extracted cleanly as string array	⚠️ Frequently missed	⚠️ Inconsistent
🎯 Built-In GEO brandCitationAudit	✅ cited / related_only vs competitor_cited / not_cited	❌ Manual post-processing required	❌ Not included
🌍 Global Markets + $0 Region Guard	✅ US, GB, DE, FR, CA, AU, IN, BR, JP + $0 on unsupported regions	❌ Fails on UK/EEA consent or bills full browser runtime	⚠️ Charges per request regardless

---

🔗 Inline citations vs. Corroborating Web sources

Google AI Mode distinguishes between pages explicitly cited inline to back specific claims inside the answer and corroborating web cards displayed in the source panel. This Actor preserves both cleanly:

1. citations (citedSourcesCount): Every source with section: "cited" that Google AI Mode directly linked inside the body of its answer.
2. sources (sourcesCount): The complete ordered list of all web sources surfaced for the prompt — combining section: "cited" inline references and section: "related" corroborating web pages, each with unwrapped url, normalized root domain, title, publisher, snippet, and faviconUrl.
3. followUpQuestions: Suggested next-turn questions generated by Google AI Mode — ideal for automated prompt-tree expansion, topic cluster mapping, and buyer-journey research.

---

💰 Transparent Pay-Per-Event Pricing (50x Cheaper Than Browser Scrapers)

This Actor uses Pay-Per-Event (PPE) billing on a 256 MB container footprint so you pay strictly for answered AI Mode prompts — never for unsupported regions (status: "unsupported_region" = $0.00 event fee).

Apify Plan Tier	Event Name	Price per Prompt	Price per 1,000 Prompts	Savings vs. $200/1K Browser Scrapers
Free Plan (FREE)	ai-mode-query	$0.0050	$5.00 / 1,000 prompts	40x cheaper (~995 prompts/mo free with $5 credit)
Paid Volume Plans (GOLD+)	ai-mode-query	$0.0040	$4.00 / 1,000 prompts	50x cheaper (~1,248 prompts per $5)
Actor Run Start	apify-actor-start	$0.0050	$5.00 / 1,000 runs	One-time startup fee per run (256 MB allocation)
Unsupported Territories	—	$0.00 / prompt	$0.00 event fee	Instant detection with zero ai-mode-query charge

💡 Budget Caps Built In: Set Maximum cost per run in Apify Console or via API (maxTotalChargeUsd), and the Actor stops cleanly with exit code 0 the instant your exact budget ceiling is reached.

---

💡 High-Impact Use Cases

1. 🎯 Generative Engine Optimization (GEO / AEO) Share-of-Voice Tracking: Pass buyer-intent prompts ("best SOC2 compliance automation platform", "hubspot vs salesforce vs pipedrive for a 25 person b2b saas team") with targetDomains to measure how often Google AI Mode recommends and cites your brand versus competitors across US, GB, DE, FR, CA, AU, IN, BR, and JP.
2. 🧠 Web-Grounded RAG & Fine-Tuning Dataset Curation: Collect deeply structured Markdown explanations complete with verified GFM tables, runnable code blocks, LaTeX equations, and provenance URLs at $0.0040 per prompt.
3. 🔎 Prompt-Tree & Content Gap Discovery: Harvest followUpQuestions and topCompetitorDomains to uncover the exact comparison matrices and third-party review publications (pcmag.com, g2.com, niche benchmarks) driving AI Mode recommendations in your category.

---

🛠️ How to Use via API (Python, Node.js & cURL)

🐍 Python Example (GEO Brand Citation Monitoring)

$pip install apify-client

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_API_TOKEN")

run_input = {
    "queries": [
        "compare postgresql vs mysql for high write workloads",
        "hubspot vs salesforce vs pipedrive for a 25 person b2b saas team",
    ],
    "country": "US",
    "targetDomains": ["hubspot.com", "pipedrive.com", "postgresql.org"],
    "resolveUrls": True,
    "maxQueries": 25,
}

run = client.actor("unitbytes/google-ai-mode-scraper").call(run_input=run_input)
items = client.dataset(run["defaultDatasetId"]).list_items().items

for row in items:
    audit = row.get("brandCitationAudit") or {}
    print(
        f"[{row['country']}] {row['query']} | status={row['status']} "
        f"| cited={row['citedSourcesCount']}/{row['sourcesCount']} "
        f"| audit={audit.get('citationStatus')}"
    )

---

🟨 JavaScript / Node.js Example

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_API_TOKEN' });

const run = await client.actor('unitbytes/google-ai-mode-scraper').call({
    queries: [
        'stocks and shares isa vs lifetime isa uk rules and fee comparison',
    ],
    country: 'GB',
    targetDomains: ['gov.uk', 'moneysavingexpert.com'],
    resolveUrls: true,
    maxQueries: 10,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0].markdown);
console.log('Follow-up prompts:', items[0].followUpQuestions);

---

🌐 cURL Example

curl -X POST "https://api.apify.com/v2/acts/unitbytes~google-ai-mode-scraper/runs?token=YOUR_APIFY_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "queries": ["tfsa vs rrsp vs fhsa contribution limits and tax comparison canada"],
    "country": "CA",
    "targetDomains": ["canada.ca"],
    "resolveUrls": true,
    "maxQueries": 5
  }'

---

📤 Example Output Record

{
  "query": "compare postgresql vs mysql for high write workloads",
  "country": "US",
  "language": "en",
  "status": "ok",
  "statusMessage": null,
  "hasAnswer": true,
  "markdown": "When choosing between **PostgreSQL** and **MySQL** for **high write workloads**, the decision comes down to how each database handles **concurrency**, **storage architecture**, and **background maintenance**.\n\n#### Core Architectural Differences Under Writes\n\n| Feature | PostgreSQL | MySQL (InnoDB) |\n| --- | --- | --- |\n| MVCC Model | Append-only row versions (requires `AUTOVACUUM` tuning) | In-place clustered primary key updates via undo logs |\n| Concurrent Writes | Strong under complex mixed read/write & JSONB indexing | Predictable throughput for simple primary-key OLTP updates |",
  "text": "When choosing between PostgreSQL and MySQL for high write workloads, the decision comes down to how each database handles concurrency, storage architecture, and background maintenance...",
  "textBlocks": [
    {
      "type": "paragraph",
      "text": "When choosing between PostgreSQL and MySQL for high write workloads, the decision comes down to how each database handles concurrency, storage architecture, and background maintenance."
    },
    {
      "type": "heading",
      "level": 4,
      "text": "Core Architectural Differences Under Writes"
    },
    {
      "type": "table",
      "headers": ["Feature", "PostgreSQL", "MySQL (InnoDB)"],
      "rows": [
        ["MVCC Model", "Append-only row versions (requires AUTOVACUUM tuning)", "In-place clustered primary key updates via undo logs"]
      ]
    }
  ],
  "sources": [
    {
      "position": 1,
      "title": "MySQL or PostgreSQL? A Workload-by-Workload Decision Guide",
      "url": "https://panelica.com/blog/mysql-or-postgresql-workload-decision-guide-2026",
      "domain": "panelica.com",
      "publisher": "Panelica",
      "snippet": "PostgreSQL stores multiple versions of each row (called tuples). Dead tuples accumulate and must be cleaned up by the VACUUM process.",
      "faviconUrl": "https://encrypted-tbn2.gstatic.com/faviconV2?url=https://panelica.com&size=32",
      "section": "cited"
    }
  ],
  "citations": [
    {
      "position": 1,
      "title": "MySQL or PostgreSQL? A Workload-by-Workload Decision Guide",
      "url": "https://panelica.com/blog/mysql-or-postgresql-workload-decision-guide-2026",
      "domain": "panelica.com",
      "publisher": "Panelica",
      "snippet": "PostgreSQL stores multiple versions of each row (called tuples). Dead tuples accumulate and must be cleaned up by the VACUUM process.",
      "faviconUrl": "https://encrypted-tbn2.gstatic.com/faviconV2?url=https://panelica.com&size=32",
      "section": "cited"
    }
  ],
  "sourcesCount": 1,
  "citedSourcesCount": 1,
  "followUpQuestions": [
    "What is the approximate write volume (e.g., rows per second)?",
    "Is it mostly inserts, updates, or a mix of both?"
  ],
  "brandCitationAudit": {
    "targetDomains": ["postgresql.org", "dev.to"],
    "citationStatus": "competitor_cited",
    "isCited": false,
    "citedTargetUrls": [],
    "citedTargetCount": 0,
    "topCompetitorDomains": [
      { "domain": "panelica.com", "citationsCount": 1 }
    ]
  },
  "googleAiModeUrl": "https://www.google.com/search?q=compare+postgresql+vs+mysql+for+high+write+workloads&gl=us&hl=en",
  "scrapedAt": "2026-11-14T12:00:00Z"
}

Complete Output Fields Reference (17 Root Fields)

Field	Type	Description
query	string	Normalized prompt or search query processed.
country	string	Uppercase 2-letter ISO 3166-1 country code ("US", "GB", "DE", "JP", etc.).
language	string	Effective interface language code used ("en", "de", "fr", "pt-BR", "ja", etc.).
status	string	"ok" when answered, or "unsupported_region" ($0 event charge) when Google AI Mode is unavailable in country.
statusMessage	string | null	Upstream status note when status is "unsupported_region", otherwise null.
hasAnswer	boolean	true when status == "ok" and a non-empty AI Mode answer was extracted.
markdown	string	Full Google AI Mode response in clean GitHub-Flavored Markdown (headings, tables, lists, code, math).
text	string	Clean plain-text representation of the AI Mode answer.
textBlocks	object[]	Structured block list (paragraph, heading, list, table, code) for RAG ingestion.
sources	object[]	All web sources (position, title, url, domain, publisher, snippet, faviconUrl, section).
citations	object[]	Subset of sources directly cited inline inside the answer (section: "cited").
sourcesCount	integer	Total count of web sources (len(sources)).
citedSourcesCount	integer	Count of inline-cited web sources (len(citations)).
followUpQuestions	string[]	Suggested conversational follow-up prompts returned by Google AI Mode.
brandCitationAudit	object | null	GEO brand audit (targetDomains, citationStatus, isCited, citedTargetUrls, citedTargetCount, topCompetitorDomains) when targetDomains is set.
googleAiModeUrl	string	Direct Google Search / AI Mode URL for the prompt and market.
scrapedAt	string	UTC ISO 8601 extraction timestamp.

---

📊 Data Honesty & Realistic Upstream Behavior

• Regional Rollout & $0 Unsupported-Territory Protection: Google AI Mode is live across major global markets (US, GB, CA, AU, IN, BR, JP, DE, FR, etc.). If you query a restricted or unlaunched territory, the Actor detects it immediately on the first check, emits a transparent record with status: "unsupported_region" and hasAnswer: false, and never bills an ai-mode-query event.
• Pure Coding & Generative Synthesis Prompts: On self-contained coding or mathematical prompts (e.g., "write a Python binary search helper with unit tests"), Google AI Mode often answers from model synthesis with sourcesCount: 0 (sources: []) while returning complete runnable code blocks in markdown and textBlocks.
• Suggested Follow-Up Questions (followUpQuestions): Google AI Mode includes suggested follow-up prompts on roughly 75–90% of queries depending on topic complexity; exhaustive comparison responses occasionally omit follow-up chips upstream (followUpQuestions: []).

---

⚙️ Input Parameters Reference

Parameter	Type	Default	Description
queries	string[]	["compare postgresql vs mysql...", ...]	Conversational prompts, comparison queries, or Google Search URLs. Automatically deduplicated.
startUrls	object[]	[]	Optional Google Search URLs ({"url": "https://www.google.com/search?q=..."}).
maxQueries	integer	25	Maximum unique prompts to process in this run (1 – 1000).
country	string	"US"	2-letter ISO 3166-1 country code (US, GB, CA, AU, IN, BR, JP, DE, FR, etc.). Unsupported territories exit cleanly at $0 event charge.
language	string	""	Optional ISO/BCP-47 language override. Leave empty ("") to automatically use the natural language of country.
targetDomains	string[]	[]	Brand/client domains (e.g. ["hubspot.com"]) to populate brandCitationAudit.
resolveUrls	boolean	true	Unwrap Google redirect links into clean canonical external URLs.
proxyConfiguration	object	null	Optional custom proxy override (proxyUrls). Managed residential proxies are configured automatically on Cloud.

---

📋 Copy-and-Paste Directive for AI Coding Assistants (Cursor, Windsurf & Claude Code)

"Use the unitbytes/google-ai-mode-scraper Actor on Apify to extract full conversational Google AI Mode answers in clean GitHub-Flavored Markdown, unwrapped citations and sources, followUpQuestions, and brandCitationAudit (targetDomains). Pair it with unitbytes/google-ai-overviews-scraper to compare classic SERP AI Overviews against conversational AI Mode. Example call: client.actor('unitbytes/google-ai-mode-scraper').call(run_input={'queries': ['compare postgresql vs mysql for high write workloads'], 'country': 'US', 'targetDomains': ['postgresql.org'], 'maxQueries': 25})"

---

🔌 MCP Server Setup: Claude Code, Cursor & AI Agents

{
  "mcpServers": {
    "apify": {
      "type": "http",
      "url": "https://mcp.apify.com/?tools=actors,docs,unitbytes/google-ai-mode-scraper,unitbytes/google-ai-overviews-scraper"
    }
  }
}

---

⚖️ Compliance & Privacy

This Actor extracts strictly publicly available AI Mode responses and cited web references served on Google Search. It never accesses private user accounts, personal histories, or personally identifiable information (PII). Read Apify's guide on the legality of web scraping to ensure compliance with applicable regulations.

Disclaimer

This Actor is an independent software tool and is not affiliated with, endorsed by, or sponsored by Google LLC.

---

💬 Support & Feedback

Tracking LLM brand visibility across global markets or curating web-grounded AI benchmarks?

• 💬 Apify Issues: Open a ticket on the Actor Issues Tab
• 🚀 Apify Store: Google AI Mode Scraper