🗺️ Google Maps Reviews Data Enricher

An Apify Actor that enriches Google Maps reviews data using AI to answer boolean questions. This Actor fetches reviews using the Google Maps Reviews Scraper and uses LLMs to extract structured information by filtering reviews by keywords and answering yes/no questions.

✨ Perfect for enriching business directories, competitive analysis, and structured dataset creation from unstructured review text.

🔧 How it works

1. 🔍 Fetch Reviews: Uses tool calling to fetch Google Maps reviews via the web_wanderer/google-reviews-scraper Actor
2. 🎯 Filter by Keywords: Filters reviews containing specified keywords (case-insensitive)
3. 💰 Cost Optimization: Limits the number of reviews analyzed per question (default: 5 reviews) to reduce LLM costs
4. 🤖 AI Enrichment: Sends limited filtered reviews to OpenAI GPT-4o-mini LLM to answer boolean questions
5. 📊 Structured Output: Returns enriched data with yes/no answers for each question per place

🎯 Example Use Case: Daycare Center Enrichment

Enrich daycare center listings with amenities information:

🔑 Keyword	❓ Question	✅ Answer
outdoor	Do they have an outdoor play area?	yes/no
camera	Do they have security cameras for parents to view?	yes/no
meals	Do they provide meals or snacks?	yes/no

🔐 Environment Variables

Before running the Actor, you must set the following environment variable:

Variable	Required	Description
OPENAI_API_KEY	✅ Yes	Your OpenAI API key (e.g., sk-...)

Setting Environment Variables

1. Navigate to Settings → Environment variables
2. Click Add environment variable
3. Set Key: OPENAI_API_KEY and Value: Your API key
4. Click Save

🤖 The Actor uses GPT-4o-mini model exclusively.

⬇️ Input

The Actor accepts the following input:

{
  "placeIds": ["ChIJoTXWl8dbwokRpKA2BJFVsGA"],
  "placeUrls": [],
  "enrichments": [
    {
      "keyword": "outdoor",
      "question": "Do they have an outdoor play area?"
    },
    {
      "keyword": "camera",
      "question": "Do they have security cameras for parents to view?"
    },
    {
      "keyword": "meals",
      "question": "Do they provide meals or snacks?"
    }
  ],
  "reviewsLimit": 100,
  "enrichmentReviewsLimit": 5
}

Input Parameters

Parameter	Type	Required	Description
placeIds	array	Optional	Array of Google Maps Place IDs
placeUrls	array	Optional	Array of Google Maps Place URLs
enrichments	array	✅ Required	Array of enrichment objects (see below)
reviewsLimit	integer	Optional	Max reviews per place to fetch (default: 100, max: 1000)
enrichmentReviewsLimit	integer	Optional	Max reviews to analyze per enrichment (default: 5, max: 50) 💰
debug	boolean	Optional	Enable debug logging (default: false)

Enrichment Object Structure:

Field	Type	Required	Description
keyword	string	✅ Required	Keyword to filter reviews (case-insensitive)
question	string	✅ Required	Boolean question to ask about the keyword
negativePattern	string	Optional	Pattern to exclude false positives (e.g., "no outdoor", "lacking outdoor")

💡 Note: You must provide either placeIds or placeUrls (or both).

🚫 Using Negative Patterns to Reduce False Positives

Negative patterns help exclude reviews that mention your keyword but in a negative context. This reduces false positives where the LLM might incorrectly answer "yes" based on complaints about missing features.

Example without negative pattern:

• Keyword: "outdoor"
• Question: "Do they have an outdoor play area?"
• Problem: Reviews mentioning "no outdoor area" or "wish they had outdoor space" will be included

Example with negative pattern:

{
  "keyword": "outdoor",
  "question": "Do they have an outdoor play area?",
  "negativePattern": "no outdoor|lacking outdoor|wish.*outdoor|need.*outdoor|missing.*outdoor|without.*outdoor"
}

This filters out reviews like:

• ❌ "Great place but no outdoor play area"
• ❌ "They're lacking outdoor space"
• ❌ "Wish they had outdoor activities"

⚠️ LLM False Positive Risk: Even with keyword filtering, LLMs can misinterpret negative reviews as positive features. Using negative patterns reduces analyzed reviews from ~20 to ~5 relevant ones, improving accuracy from ~70% to ~95% for boolean questions.

⬆️ Output

The Actor outputs structured enrichment data:

{
  "results": [
    {
      "place_id": "ChIJoTXWl8dbwokRpKA2BJFVsGA",
      "place_name": "Sunshine Daycare Center",
      "place_address": null,
      "total_reviews": 47,
      "enrichments": [
        {
          "keyword": "outdoor",
          "question": "Do they have an outdoor play area?",
          "answer": "yes",
          "matching_reviews_count": 12,
          "analyzed_reviews": [
            "Great daycare with a large outdoor playground...",
            "They have a nice outdoor area where kids can play..."
          ]
        },
        {
          "keyword": "camera",
          "question": "Do they have security cameras for parents to view?",
          "answer": "yes",
          "matching_reviews_count": 23,
          "analyzed_reviews": [
            "Love being able to check the cameras anytime...",
            "The security camera system gives us peace of mind..."
          ]
        }
      ]
    }
  ],
  "metadata": {
    "processed_at": "2025-01-XX...",
    "total_places": 1,
    "total_enrichments": 2
  }
}

✨ Features

Feature	Description
🔧 Tool Calling	Uses Apify tool calling to fetch reviews from trusted scraper
🎯 Keyword Filtering	Filters reviews by keywords before LLM analysis
🚫 Negative Patterns	Exclude false positives with regex patterns
🤖 LLM Support	Uses OpenAI GPT-4o-mini model
✅ Boolean Answers	Returns yes/no answers with confidence indicators
🗺️ Multiple Places	Process multiple places in a single run
❓ Multiple Questions	Ask multiple enrichment questions per place
📊 Structured Output	Clean JSON output ready for downstream processing
🔐 Secure API Keys	API keys configured via environment variables

💼 Use Cases

Use Case	Description
📋 Business Directory Enrichment	Add structured data about amenities, services, or features
📊 Competitive Analysis	Compare features across multiple businesses
🔍 Market Research	Extract specific information from customer reviews
🔗 Data Pipeline Integration	Feed structured data into ML/AI pipelines
✓ Quality Assurance	Verify business information from customer feedback

💰 Pricing Structure

This Actor uses a Pay Per Event (PPE) pricing model that automatically adjusts based on your Apify discount tier:

Event	Description	FREE Tier	BRONZE Tier	SILVER Tier	GOLD Tier
actor-start-*	Actor initialization	$10 / 1,000 ($0.01)	$8 / 1,000 ($0.008)	$6 / 1,000 ($0.006)	$5 / 1,000 ($0.005)
enrichment-processed-*	Per enrichment question	$5 / 1,000 ($0.005)	$4 / 1,000 ($0.004)	$3 / 1,000 ($0.003)	$2 / 1,000 ($0.002)
task-completed	Task completion	$0.10	$0.10	$0.10	$0.10

Pricing Examples

Example 1: Small run (FREE tier)

• 1 place, 2 enrichment questions
• Cost: $0.01 + (2 × $0.005) + $0.10 = $0.12
• Plus: User's scraper costs + User's OpenAI costs

Example 2: Medium run (BRONZE tier)

• 10 places, 3 enrichment questions each (30 total)
• Cost: $0.008 + (30 × $0.004) + $0.10 = $0.228
• Per place: $0.0228
• Plus: User's scraper costs + User's OpenAI costs

Example 3: Large run (GOLD tier)

• 100 places, 5 enrichment questions each (500 total)
• Cost: $0.005 + (500 × $0.002) + $0.10 = $1.105
• Per place: $0.01105
• Plus: User's scraper costs + User's OpenAI costs

Example 4: Bulk run (GOLD tier)

• 1,000 places, 3 enrichment questions each (3,000 total)
• Cost: $0.005 + (3,000 × $0.002) + $0.10 = $6.105
• Per place: $0.006105
• Plus: User's scraper costs + User's OpenAI costs

Key Benefits

• Competitive pricing: Aligned with Google Maps Scraper pricing model
• Low per-place cost: Scales efficiently from $0.012 (small) to $0.006 (bulk) per place
• Tiered discounts: Up to 60% savings with GOLD tier vs FREE tier
• Transparent: User pays separately for scraper and LLM costs
• Bulk-friendly: Per-place cost decreases significantly with volume

Additional Costs

Users are responsible for:

• Scraper costs: Paid directly to web_wanderer/google-reviews-scraper Actor
• LLM costs: Paid directly to OpenAI (user provides their own API key)

💡 Additional Costs: Users are responsible for scraper costs (paid to web_wanderer/google-reviews-scraper) and LLM costs (paid to OpenAI via your API key).

📝 Examples

Example 1: Daycare Center Amenities

{
  "placeUrls": ["https://www.google.com/maps/place/..."],
  "enrichments": [
    {
      "keyword": "outdoor",
      "question": "Do they have an outdoor play area?",
      "negativePattern": "no outdoor|lacking outdoor|wish.*outdoor|need.*outdoor"
    },
    {
      "keyword": "camera",
      "question": "Do they have security cameras for parents to view?",
      "negativePattern": "no camera|no security|can't view|cannot view"
    },
    {
      "keyword": "meals",
      "question": "Do they provide meals or snacks?",
      "negativePattern": "no meals|no food|bring.*food|pack.*lunch"
    }
  ],
  "enrichmentReviewsLimit": 5
}

Example 2: Restaurant Features

{
  "placeIds": ["ChIJ..."],
  "enrichments": [
    {"keyword": "parking", "question": "Do they have parking available?"},
    {"keyword": "outdoor", "question": "Do they have outdoor seating?"},
    {"keyword": "vegetarian", "question": "Do they offer vegetarian options?"}
  ],
  "reviewsLimit": 200,
  "enrichmentReviewsLimit": 5
}

💡 Cost Optimization Tips

Setting	Use Case
5 reviews (default)	Usually sufficient for accurate yes/no answers while minimizing costs
10-15 reviews	Higher confidence for critical questions
3 reviews	Processing many places/questions and want to minimize costs
Use negative patterns	Reduce false positives and improve accuracy by 20-25%

🎯 Pro Tip: The Actor first filters reviews by keyword and negative patterns, so only relevant reviews are analyzed, maximizing cost efficiency.

⚠️ Limitations

Limitation	Description
🔑 Keyword dependency	Requires reviews to contain relevant keywords for filtering
🤖 LLM accuracy	Answers depend on review content quality and model accuracy
⏱️ Processing time	Increases with more places and questions
🔄 Scraper limits	Limited by scraper Actor availability and rate limits
🎯 False positives	LLMs may misinterpret ~5-10% of negative reviews (use negative patterns to reduce)

🔧 Troubleshooting

Issue	Solution
🔍 No reviews found	• Verify place IDs/URLs are correct • Check if places have public reviews on Google Maps • Increase reviewsLimit if needed
🤖 LLM errors	• Verify OPENAI_API_KEY environment variable is set correctly • Check that API key has credits and is valid • Check OpenAI API rate limits • Verify key has access to GPT-4o-mini model
🎯 Keyword not matching	• Keywords are case-insensitive substring matches • Try variations or more common terms • Check if reviews actually mention the keyword
❌ False positives	• Add negative patterns to exclude irrelevant reviews • Use regex to catch variations (e.g., "no.*outdoor|lacking.*outdoor")

📚 Resources

Resource	Link
📖 Apify SDK Documentation	docs.apify.com/sdk/python
🔗 LangGraph Documentation	langchain-ai.github.io/langgraph
🗺️ Google Maps Reviews Scraper	web_wanderer/google-reviews-scraper
🏢 Apify Platform Documentation	docs.apify.com/platform
💬 Join Apify Discord Community	discord.com/invite/jyEM2PRvMU

📄 License

This Actor is published on Apify Store and available for use under Apify's Terms of Service.