Trustpilot Reviews Scraper: TrustScore, Stars, Replies

Give it company domains or Trustpilot review pages and it returns one row per company (TrustScore, review count, star split, categories, reply rate) and one JSON row per review: stars, title, text, date, language, the verified flag and the company's reply with its date. From $0.52 per 1,000 reviews on the free plan, and you pay only for the review rows you get.

Company mode: give it Trustpilot categories (bank, electronics_store, any sub-category) and countries, and it returns every listed company with TrustScore, stars, review count, categories, city and country, $2 per 1,000 companies on the free plan. Or switch reviews off and get only the company rows for your own company list.

Why this actor

• Drop-in input. Input written for automation-lab/trustpilot, memo23/trustpilot-scraper-ppe or blackfalcondata/trustpilot-reviews-scraper works as it is (see "Coming from another Trustpilot scraper?").
• Tiered price per review: $0.52 per 1,000 on the free plan, down to $0.27 on Business (Platinum and Diamond also pay $0.27). The start fee is $0.005 per GB of memory ($0.01 at the default 2 GB).
• More than 200 reviews per company via filter combinations. Trustpilot shows an anonymous visitor at most 10 pages (200 reviews) per filter combination. When you ask for more, the actor also reads each star rating on its own, the other sort order and date windows, and drops repeats. The free coverage row says when the site's limit was reached.
• Company mode for lead lists and market maps: every company of a Trustpilot category (top level, sub-category or leaf, up to 10,000 per category and country) at $2 per 1,000 on the free plan, down to $1.50, sorted the way Trustpilot ranks them, deduplicated across categories and countries. Input written for shareze001/trustpilot-companies-by-category, data_direct/trustpilot-category-companies-scraper, khadinakbar/trustpilot-company-scraper or saswave/trustpilot-company-infos works too.
• Company row included with reviews, free: TrustScore, stars, review count, the star split from 1 to 5, review counts per language, categories, claimed status and reply statistics.
• onlyNewReviews for monitoring: a run returns only reviews it has not delivered before, and repeats are not charged.
• Reviewer names, avatars, profile links, countries and review counts are never collected, the reviewer's name is removed from the review and reply text, and company contact details are dropped. Free error rows with a machine-readable code (not_found, blocked, unavailable, invalid_input), so an agent knows whether to retry.

Built by Arcadium Data: public data, dated and agent-ready, never personal data.

Use with AI agents

Returns Trustpilot company ratings and reviews for company domains. Agents can call it through the Apify MCP server (https://mcp.apify.com, open source as apify/actors-mcp-server) as arcadium/trustpilot-reviews: the input schema is read from the actor, the result is a dataset of JSON rows.

Input example (minimal):

{
  "companies": ["www.bestbuy.com"],
  "maxReviewsPerCompany": 100
}

What comes back: one free company row per company, one review row per review (charged), a free coverage row per company (stated total vs delivered) and a free error row for an input that could not be read. Filter on the rowType field. The error row carries rowType: "error", a code, a message and the input it is about. Retry on blocked and unavailable later; do not retry not_found or invalid_input with the same input.

Cost of the call above: 100 reviews at $0.52 per 1,000 plus the $0.01 start fee is about $0.062 on the free plan; paid plans pay less.

Company mode (the first 200 banks Trustpilot lists for the UK):

{
  "categories": ["bank"],
  "countries": ["GB"],
  "maxCompaniesPerCategory": 200
}

That returns 200 charged company rows plus one free coverage row (stated = the companies Trustpilot lists, delivered, stopReason): 200 x $0.002 + $0.01 = $0.41 on the free plan. For company rows of your own list without reviews: { "companies": ["www.bestbuy.com"], "includeReviews": false }.

Input

Field	What it does
companies	Company domains (bestbuy.com, www.zalando.co.uk) or Trustpilot review URLs on any country subdomain. Company names are not looked up: give the domain.
maxReviewsPerCompany	Reviews per company, up to 50,000 (0 = as many as the site shows). Default 100.
stars	Only these star ratings (1 to 5), filtered by Trustpilot.
languages	Only these languages (two-letter codes). Empty = every language.
sort	recency (newest first, default) or relevance.
date	Trustpilot's date filter: last30days, last3months, last6months, last12months.
dateFrom	Only reviews published on or after a date (2026-01-01 or 30 days).
verified, withReplies	Only verified reviews; only reviews with a company reply.
includeCompanyInfo	The free company row (default on).
splitFilters	More than 200 reviews per company via filter combinations (default on).
onlyNewReviews, stateKey	Incremental mode for monitoring.
categories	Company mode: category ids (bank, electronics_technology, hi-fi_shop) or category pages (https://www.trustpilot.com/categories/bank?country=DE). Ids are the last part of the category page URL.
countries	Two-letter codes the categories are listed for (US, GB, DE, ...). Default US.
companySort	most_relevant (default), reviews_count or latest_review.
claimedOnly	Only companies that claimed their profile.
maxCompaniesPerCategory	Companies per category and country, up to 10,000. Default 200.
includeReviews	Off: the companies list gives charged company rows and no reviews. Default on.
watch, watchKey	Watch mode: only new or changed reviews and companies (see below).

What you get

Review rows:

Field	What it holds
reviewId	Trustpilot___ plus Trustpilot's own review id. Stable between runs, so you can dedupe exports.
reviewUrl	Link to the single review.
ratingOverall	Stars, 1 to 5.
title, text	Headline and review text, with the reviewer's name, emails and phone numbers removed.
date	Publication date, ISO yyyy-mm-dd; the full timestamp is extras.publishedAt.
language	Language code of the review.
reviewerVerified	Whether Trustpilot marks the review as verified; extras.verificationLevel and extras.verificationSource say how.
vendorResponse	The company's reply, with the reviewer's name removed; extras.vendorResponseDate is its date.
helpfulVotes	"Useful" votes.
extras.experiencedDate	Date of the experience the reviewer gives.
extras.reviewSource	How the review came in (Organic, an invitation tool).
product	Company summary repeated on every row: name, TrustScore, review count, primary category, website.

Company rows (rowType: "company"): entityId (the Trustpilot domain), name, url, website, trustScore, stars, reviewCount, reviewCountLast12Months, starSplit ({ "1": 6315, ..., "5": 415 }), languages ({ "en": 6842, "fr": 162 }), primaryCategory, categories, categoryPath, country, isClaimed, claimedDate, isClosed, isCollectingReviews, usesPaidFeatures, replyPercentage, averageDaysToReply, negativeReviewsWithReplies, negativeReviewsTotal, lastReplyToNegativeReview, hasConsumerAlert, alertType, isVerified. The same data is in the key-value store record COMPANY.

Company rows from a category (source: "category") carry what the category listing shows: businessUnitId, entityId, name, url, website, trustScore, stars, reviewCount, categories, primaryCategory, city, country, hasConsumerAlert, alertType, isGamblingRelated, isRecommendedInCategories, isClaimed (true with claimedOnly, else null), and where it was found: listingCategory, listingCountry, listingRank. Profile-only fields (star split, reply statistics, claimed date) are null there; run the domains with includeReviews: false to get them. Profile rows have source: "profile". Email addresses, phone numbers and street addresses are never output.

Output example (company row from a category, shortened):

{
  "rowType": "company",
  "source": "category",
  "businessUnitId": "5d2f4f05dc509f00010c9b44",
  "entityId": "itrustcapital.com",
  "name": "iTrustCapital",
  "url": "https://www.trustpilot.com/review/itrustcapital.com",
  "trustScore": 5,
  "stars": 5,
  "reviewCount": 9946,
  "primaryCategory": "Software Company",
  "country": "US",
  "hasConsumerAlert": true,
  "alertType": "InfoHighRiskInvestment",
  "listingCategory": "electronics_technology",
  "listingCountry": "US",
  "listingRank": 1
}

Output example (review row, shortened):

{
  "rowType": "review",
  "platform": "trustpilot",
  "productName": "Best Buy",
  "entityId": "www.bestbuy.com",
  "reviewId": "Trustpilot___6ad7fab995ae7024fff4278e",
  "reviewUrl": "https://www.trustpilot.com/reviews/6ad7fab995ae7024fff4278e",
  "ratingOverall": 1,
  "title": "BUYERS BEWARE: BEST BUY TOOK MY $249 AND I NEVER RECEIVED MY LAPTOP!",
  "date": "2026-10-21",
  "language": "en",
  "reviewerVerified": false,
  "vendorResponse": null,
  "extras": { "experiencedDate": "2026-09-25", "verificationLevel": "not-verified", "reviewSource": "Organic" }
}

Coming from another Trustpilot scraper?

Their field	Ours
companyUrls (automation-lab), startUrls (memo23, blackfalcondata), companyDomains, companyDomain (blackfalcondata)	companies (all of them are merged)
maxReviewsPerCompany (0 = all), maxResults, maxItems	maxReviewsPerCompany
filterStars, filterLanguages, filterDateRange, filterVerifiedOnly, filterRepliesOnly (memo23)	stars, languages, date, verified, withReplies
sortBy (recent, relevancy, auto)	sort
publishedAfter, startDate, lastDays, lookbackDays	dateFrom
includeCompanyDetails	includeCompanyInfo
scrapeAllReviews	splitFilters
incrementalMode	onlyNewReviews

maxItems is a per-run total in memo23's actor; here it becomes the per-company limit.

Company scrapers:

Their field	Ours
categoryId (shareze001), category_id (data_direct), /categories/... entries in queries (khadinakbar)	categories
domains (saswave), other queries entries (khadinakbar)	companies
country	countries
size (shareze001), maxResultsPerQuery (khadinakbar)	maxCompaniesPerCategory
sort (reviews_count, latest_review), sort_by (recommended, recently_reviewed)	companySort
claimed	claimedOnly
mode (khadinakbar)	includeReviews: false

TrustScore minimum filters (trustscore, min_rating) are not applied: filter on trustScore in the output.

Pricing

Plan	Per 1,000 reviews	Per 1,000 companies (company mode)
Free	$0.52	$2.00
Bronze	$0.45	$1.80
Silver	$0.35	$1.60
Gold	$0.27	$1.50
Platinum	$0.27	$1.50
Diamond	$0.27	$1.50

Plus a start fee of $0.005 per GB of memory ($0.01 at the default 2 GB). A company is charged once per run, however many categories or countries list it. The company row that comes with reviews, coverage, error and summary rows are free.

Watch mode (below): new and changed reviews and companies are charged as above; each unchanged review or company costs $0.06 per 1,000 (watch-check), on every plan, from the second run of a watch key on. The first run of a watch key is a normal run, charged normally.

Watch mode: pay only for changes

Set "watch": true and a watchKey (for example "weekly-competitors"), then run the same input on an Apify schedule. The first run has nothing to compare with: it is a normal run, returns every review up to maxReviewsPerCompany and is charged normally, and it saves a snapshot. Every later run with the same key reads the newest reviews again (newest first) and returns only:

• new reviews (change.type: "new"), and
• reviews whose stars, title, text, company reply or verified flag changed (change.type: "changed").

{
  "type": "changed",
  "fields": [{ "field": "rating", "old": 2, "new": 4 }, { "field": "companyReply" }],
  "previousCheckAt": "2026-10-31T13:04:42.969Z"
}

Stars and the verified flag show old -> new; title, text and company reply are named only (the row carries the new text). Unchanged reviews are not delivered; they cost only the small watch-check above, because the run still has to read them. Every watch run also adds one free watch-summary row: checked, unchanged, changed, new, gone (in the last snapshot, not found now), delivered, checksCharged, firstRun, previousCheckAt and complete (false when the run stopped early, e.g. at your cost limit: then nothing counts as gone and changes that were not delivered come again next run). A review counts as gone only when its company was read to the end and the review is newer than the oldest review read, so reviews that just moved past maxReviewsPerCompany do not count. The snapshot is kept in a named store (trustpilot-reviews-watch) in your account; use one watchKey per company list. Watch mode replaces onlyNewReviews (it also reports edits and removals).

Watch mode works for company rows too (a weekly TrustScore panel of a category, or of your competitor list with includeReviews: false): a company row comes back only when it is new or its trustScore, stars, reviewCount or consumer alert changed (profile rows also isClaimed and replyPercentage); the listing rank is not compared. Company rows get their own free watch-summary row with scope: "companies". A company counts as gone only when its category listing was read to the end in this run.

Weekly digest

Add "digest": true to a watch run to get one more free row, rowType: "watch-digest", that sums up the changes of the last digestDays days (default 7, up to 90) for this watchKey: new, changed, gone, byType (counts per change type), tags.newOneStar and tagsByLabel.newOneStar (new 1-star reviews in total and per company), topMoves (the 5 biggest moves per metric and direction, largest relative change first, each with label, old, new, delta, pct), runs in the window, windowStart, windowEnd, previousDigestAt and summary: one plain-text line an AI agent or a workflow can paste into an email or Slack message as it is. The window is rolling: put the watch on a daily schedule and every run's digest covers the last 7 days, or run weekly and read one digest a week. The first run of a key only saves the baseline and its rows are not counted as changes. The window is kept as a second record in the same key-value store as the snapshot. The digest row is free; there is no extra event.

Change types: new_review, rating_changed, reply_changed, edited. topMoves ranks each company's TrustScore (trustScore) and total review count (reviewCount) from the first to the last run in the window; labels are company domains.

In company mode (categories, or includeReviews: false) the company rows get their own digest row with scope: "companies": change types new_company, trustscore_drop, trustscore_rise, consumer_alert, review_count, and topMoves for trustScore, reviewCount and stars of every delivered company. Example summary (illustrative values):

trustpilot watch "competitors", last 7 days (7 runs): 41 new, 2 changed, 1 gone. new reviews 41, star ratings changed 2. new 1-star reviews: 6 (www.example-a.com 4, www.example-b.com 2). Biggest TrustScore drop: www.example-a.com 4.2 -> 4.1, -2.4%. Biggest review count rise: www.example-a.com 1200 -> 1241, +3.4%.

Related actors

• Yelp Reviews Scraper: All Languages, Owner Replies
• Review Monitor: Negative Review Alerts and Rating Drift

Limitations

• Trustpilot shows an anonymous visitor 10 pages per filter combination; reading further would need a login, which this actor never does. With splitFilters the actor reaches more reviews through other combinations, not all of them for very large profiles: the coverage row then says stopReason: "cap".
• With several languages or splitFilters, rows are not in one date order; sort on date if you need it.
• Category listings stop at 10,000 companies per category and country (Trustpilot's own limit, coverage stopReason: "cap"); list sub-categories to reach more. Trustpilot's TrustScore filter is not applied.
• Public data only. No login, no personal data, no contact details.