Review Intelligence Monitor

Stop reading raw review exports. This Actor scrapes Trustpilot reviews and gives you back a prioritized, classified complaint report plus a fake-review risk signal — not just a CSV of text.

What it does

1. Scrapes recent reviews from any public Trustpilot company page.
2. Classifies every negative review (configurable threshold) using keyword-based rules into:
   • Complaint category (shipping, product quality, customer service, billing, etc.)
   • Urgency (low / medium / high — flags legal threats, regulator mentions, viral-complaint risk)
   • A short summary (the review's own title/text, since classification is rule-based, not AI-generated)
3. Detects suspicious review patterns: volume spikes, near-duplicate review text, and thin/new reviewer accounts — surfaced as a risk score (0-100), not a binary accusation.

No AI API costs

Classification and fraud detection both run on local keyword rules and statistics — no LLM calls, no external API key required, no per-review AI cost. You only pay for this Actor's own Apify compute usage, same as any scraper.

Trade-off to know about: keyword rules catch obvious cases well but won't understand sarcasm, implicit complaints, or unusual phrasing as well as an AI model would. Treat the categorization as a strong first pass, not a perfect one.

Why this instead of a plain review scraper

Plain scrapers give you rows of text. You still have to read 500 reviews to find the 5 that need a response today. This Actor does that triage for you and tells you which reviews are urgent and whether your review set shows signs of manipulation.

Input

Field	Required	Description
trustpilotUrl	Yes	Trustpilot company URL or bare domain
maxReviews	No	Cap on reviews fetched (default 100)
negativeRatingThreshold	No	Ratings at or below this are classified (default 3)
runFraudDetection	No	Toggle the fraud-signal analysis (default true)

Output

Three record types in the dataset:

• summary — company name, trust score, counts, top-level risk label
• complaint — one row per classified negative review, sorted by urgency
• fraud_report — volume spikes, duplicate-text pairs, thin-account flags, overall risk score

Important disclaimers

• Fraud detection is a statistical signal, not proof. Legitimate review-request campaigns can also trigger volume spikes.
• Classification is keyword-based, not AI-based — accuracy depends on how closely review wording matches the built-in keyword lists (currently English-only).
• This is an independent tool, not affiliated with or endorsed by Trustpilot.