Review Digest & AI Reply Drafts: Google Reviews Monitor

Turn the reviews you already scrape into something you can act on. Feed this Actor the dataset of a Google Maps Reviews Scraper run (or any review export) and get, per review: sentiment, themes, an "unanswered" flag and an AI reply draft in the reviewer's language — plus one digest per business with what customers like, what they complain about and up to three next steps.

Schedule it after your review scraper and you only pay for new reviews.

Who is it for?

• Agencies managing Google reviews for many clients: one digest per client, reply drafts ready to paste.
• Multi-location businesses (restaurants, shops, clinics, hotels): spot the location with a waiting-time problem this week.
• Reputation and CX teams: urgent reviews (safety, health, legal, discrimination, fraud) flagged first.

How it works

1. Run a review scraper, e.g. Google Maps Reviews Scraper or Google Maps Scraper with reviews enabled.
2. Pick its dataset in Datasets (or paste the dataset ID). You can also pass review rows as JSON.
3. Choose Only new reviews since the previous run and give the monitor a name.
4. Start. Reviews are analysed in batches; the digest is saved as a table and as Markdown.

Run it automatically after your scraper

In your review scraper's Integrations tab add Run Actor, choose this Actor and use this input:

{
  "datasetIds": ["{{resource.defaultDatasetId}}"],
  "mode": "new",
  "stateName": "client-a",
  "summaryLanguage": "en"
}

Output example

Review row:

{
  "type": "review",
  "business": "Example Bakery",
  "stars": 1,
  "publishedAt": "2026-10-22T15:40:00Z",
  "language": "nl",
  "text": "Bestelde een taart voor zaterdag, bij het afhalen was ze er niet. Geen excuses, geen oplossing.",
  "sentiment": "negative",
  "themes": ["order not ready", "no apology"],
  "urgent": false,
  "hasOwnerResponse": false,
  "unanswered": true,
  "ageHours": 66.3,
  "replyDraft": "Wat vervelend dat uw taart zaterdag niet klaarstond. Dit hoort niet te gebeuren...",
  "reviewUrl": "https://www.google.com/maps/reviews/..."
}

Digest per business (saved as DIGESTS in JSON and as DIGEST in Markdown, see the Output tab):

{
  "business": "Example Bakery",
  "reviewsAnalyzed": 3,
  "averageStars": 2.67,
  "negative": 2,
  "unansweredNegative": 2,
  "topThemes": [{"theme": "waiting time", "mentions": 2, "negative": 2, "positive": 0}],
  "summary": "Customers love the bread but Saturday mornings are chaotic...",
  "complaints": ["long waits on Saturday", "pre-orders not ready"],
  "actions": ["add a second cashier on Saturday 8–11", "confirm pre-orders by SMS the day before"]
}

Pricing

Event	Price
Review analysed (sentiment, themes, flags, reply draft)	$0.005
Digest per business	$0.05

Example: a restaurant with 40 new reviews a month = 40 × $0.005 + 4 weekly digests × $0.05 = $0.40 per month.

Not charged: star ratings without text (listed with a sentiment from the stars), reviews already analysed by the same monitor, reviews filtered out (date, stars), reviews above your limit, and anything the AI could not answer. Set a maximum cost per run and the Actor stops cleanly.

Privacy

• Reviewer names, profile links, photos and IDs are removed before anything is processed and never appear in the output.
• Review text is sent to Cloudflare Workers AI to be analysed; nothing else is sent there and nothing is kept after the run except your own dataset, the digest and the monitor's list of analysed review IDs.
• Reply drafts never name the reviewer, never promise refunds and never admit liability. Always read a draft before posting it.

Supported inputs

• Google Maps Reviews Scraper: one item per review (recommended).
• Google Maps Scraper: place items with a reviews array (spread out automatically).
• Other review scrapers or your own export: rows with keys such as text/reviewBody, stars/rating, date/publishedAtDate, reviewId, businessName/companyName, responseFromOwnerText/reply.

Good to know

• Free Apify plan: at most 30 reviews per run.
• An "urgent" flag is a hint for safety, health, legal, discrimination or fraud claims, not a legal assessment.
• Analysis comes from a language model (Llama 3.3 70B on Cloudflare Workers AI). Check a sample before acting on it.
• The digest compares only the reviews of this run; for monthly trends, run it monthly with All reviews and Only reviews since.

FAQ

Does this Actor scrape Google? No. It processes reviews you already have; pair it with a review scraper.

Which languages? Reviews in any language; reply drafts follow the reviewer's language; the digest is written in the language you choose (EN, NL, FR, DE, ES, IT, PT).

Can I post the replies automatically? No, drafts are meant to be checked and posted by you.

Can I use Trustpilot or App Store reviews? Yes, if you have them as a dataset or JSON; common field names are recognised.

Support

Open an issue on the Issues tab with the run ID.