Every AI claim is backed by an exact-match quote from a real review

Google Maps Reviews → AI Competitive Report

New Actor. Turn an existing Apify reviews dataset or your own JSON into one evidence-backed competitive report per place. This Actor does not scrape Google Maps, Yelp, or any website. You bring the reviews you already have, plus your own OpenAI API key.

It analyzes review text and ratings only. Reviewer names, profile URLs, and other reviewer personal data are ignored and never returned.

Why the output is safer to use

AI can make a plausible claim without adequate evidence. This Actor rejects that failure mode in code: every AI-produced competitive claim, strength, complaint, and recommendation must carry a quotedReview that exactly equals one original review body. A quote that is edited, shortened, invented, or merged is dropped from the output.

Source-derived rating fields (averageRating and sentimentBreakdown) are calculated directly from supplied star ratings. sentimentBreakdown uses 4–5 stars = positive, 3 = neutral, and 1–2 = negative.

Input

Provide exactly one review source and an OpenAI API key:

Field	Required	Description
datasetId	One source required	Existing Apify dataset ID containing place records and/or review rows.
reviewsJson	One source required	Array of place records with reviews arrays and/or individual review rows.
openaiApiKey	Yes	Your OpenAI API key. Used only for this run; never emitted in output or logs.
language	No	en (default) or ja for the report wording.
maxReviewsPerPlace	No	Maximum source reviews per place; default 200.

For compatibility with common Google Maps dataset shapes, review text may be in reviews[].text, reviewText, comment, or content; ratings may be in stars, rating, or score. Each place is compressed to no more than 12,000 review-text characters before the AI call.

Output

One record per place:

{
  "placeName": "Example Cafe",
  "sourceReviewCount": 84,
  "reviewsAnalyzed": 71,
  "averageRating": 4.23,
  "sentimentBreakdown": { "positive": 62, "neutral": 4, "negative": 5 },
  "competitiveSummary": { "claim": "Guests praise the coffee quality.", "quotedReview": "The espresso was excellent and beautifully balanced." },
  "strengths": [{ "theme": "Coffee quality", "quotedReview": "The espresso was excellent and beautifully balanced." }],
  "complaints": [{ "theme": "Slow service", "quotedReview": "It took thirty minutes to get our order." }],
  "recommendations": [{ "action": "Improve peak-hour service speed.", "quotedReview": "It took thirty minutes to get our order." }],
  "aiTokenUsage": { "promptTokens": 0, "completionTokens": 0, "totalTokens": 0 },
  "sourceDatasetId": "existing-dataset-id",
  "analyzedAt": "2026-08-02T00:00:00.000Z"
}

If no exact source quote supports an AI claim, the claim is omitted (competitiveSummary: null or an empty array). This is intentional.

Pricing and limits

The Actor is priced per output place report. Your OpenAI usage is billed by your own API provider, separately from Apify. It does not provide or resell an AI key.

Notes

• This Actor analyzes supplied data only; it does not retrieve reviews itself.
• A run fails clearly before analysis if openaiApiKey is missing.
• The default dataset contains the evidence-backed output records. A SUMMARY key-value record reports success rate, AI token total, and rejected unsupported-claim count.