Reviews Monitor: Shopify, Amazon & Trustpilot

This reviews monitor reads Shopify reviews from product pages of shops that use Judge.me, Yotpo, Okendo, Loox or Stamped (most Shopify stores), Amazon ASINs and businesses on Trustpilot, and gets each product's (or company's) average rating, review count, share per star, its newest reviews (rating, title, text, date, verified flag, review id) and complaint and praise themes such as "sizing runs small" or "arrived damaged", each with example review ids. Schedule it and it says what changed since your last run: new reviews, the rating moving, the review count jumping or dropping, a product taken down. It runs once, on a schedule or through the API, with no sign-in.

Also known as a product reviews scraper, Judge.me, Yotpo, Okendo, Loox and Stamped reviews scraper, Amazon rating tracker, Trustpilot reviews scraper and review alert tool.

How to monitor product reviews

1. Open Reviews Monitor on Apify and go to its input form.
2. Paste product page links in Product URLs, one per line. Add Amazon products in Amazon ASINs (with Amazon marketplace for plain ASINs) and companies in Trustpilot businesses.
3. Set Reviews from the last (days) and Reviews per product (0 gives the rating summary only), and keep Complaint and praise themes on.
4. Start the run. Open the Rating summaries view to read each product's Average, Reviews, Per star, Top complaints and Top praise, and the Reviews view for each review's Date, Rating, Title and Review.
5. To be told about new reviews and rating moves, turn on Changes only, give the watchlist a Monitor name, save it as a task and schedule it daily. The Changes view then lists new reviews with their Lowest rating, and rating and count moves with From, To and Delta.

What the reviews monitor gives you

• "Tell me when a competitor gets a bad review." Put their product URLs in a watchlist, turn on Changes only and schedule it daily: only new reviews come back, with a change row giving their lowest rating.
• "What do shoppers complain about?" Every product's summary row lists complaint and praise themes from the reviews read, with how many reviews mention each and which ones.
• "Is my rating slipping?" A rating move of 0.1 stars or more, and a review count jumping or dropping by 10% or more, become change rows. Both thresholds are yours to set.
• "Compare my product with theirs." Mix your own and competitors' URLs and ASINs in one run: every row has the same fields, whatever the source.
• "Feed reviews to an AI agent or a spreadsheet." Compact JSON with stable review keys, a field picker and a dataset view per job.

How the reviews monitor reads reviews

1. Shop product pages: the actor reads the page, finds the review app it uses and the public ids that app's widget needs (shop, app key, subscriber, store id), then reads the same public review data the widget loads in a shopper's browser, newest first:
   • Judge.me: reviews from its widget endpoint; totals and stars per level from the widget rendered in the page.
   • Yotpo: reviews, totals and stars per level from its widget API.
   • Okendo: reviews (with the variant reviewed) and its aggregate with stars per level.
   • Loox: reviews and totals from its widget.
   • Stamped: reviews and totals from its widget API.
2. Amazon ASINs: the product page signed out: star rating, ratings count, the share per star and the top reviews it shows (about 8 to 13). Outside the US, the page is read from addresses in that marketplace's country.
   • amazon.com (US) shows no review text to signed-out shoppers. Since October 2026 its product page says "Sign in to see customer reviews" where the top reviews were, from every kind of address we tried (Apify's own, datacenter, a home connection and a page-unlocking service). This actor never signs in, so a US ASIN gives the rating, ratings count and share per star, and no reviews or themes. This is a lasting Amazon limit, not a failure: the run stays COMPLETE (or NO_CHANGES), the rating summary row says why in reviewsNotShown (reviewsNotShownKind: "sign-in"), and RUN_SUMMARY lists the products in reviewsNotShown with one line in warnings; the status names the count. Set Reviews per product to 0 for a US watchlist of ratings only. Outside the US, Amazon still shows its top reviews (checked on the UK, Germany and Spain, 2026-10-27), and they are read as before.
3. Trustpilot businesses (a domain such as allbirds.com, or its trustpilot.com/review/… page): the company's reviews, newest first, with its overall rating, review count and share per star. Trustpilot is read through CleanPull's marketplace data service, a third-party web data provider; CleanPull holds the account and its key, which never appears in your input, logs or results.
   • The newest first, up to Reviews per product, within Reviews from the last (days) (the service filters by 30 days, 3, 6 or 12 months, so the smallest of those covering it is asked for, and older reviews are dropped). With Changes only, each business reads at most its Trustpilot reviews per business (changes only) most recent reviews (5 by default): new ones are returned, the others are reported as re-checked. Once a business has been read twice, a run first reads only as many as its review total grew last time, plus one (one review for a quiet business), and reads the full number when that short read did not reach a review seen before or the total grew by more than the new reviews it found, so no new review is missed. Raise the setting for a business that gets more new reviews than that between runs.
   • Each review delivered, and each recent review re-checked on a watchlist run, is a provider record on your bill (RUN_SUMMARY gives recheckedReviews, and the status says "N recent reviews re-checked"). Reviews the service brings back from before that window are dropped and not charged per record: the business's provider search covers them. A read takes about 30 to 60 seconds and businesses are read up to 4 at once (Products at a time), so 100 businesses take about 15 to 25 minutes.
   • With Reviews per product at 0, one review is read for the company's totals and no review rows are returned.
   • A business with no review in the window gets no totals from the service on that run: the last known rating and count stand.

Reading stops at the first review older than your window (Reviews from the last (days)) or at your limit per product, so a product with 5,000 reviews costs no more than one with 20.

In tests from Apify's servers, every source above answered every request from Apify's own addresses (no proxy needed; Loox refused one datacenter proxy address in three), and a watchlist of 7 products from all six sources read twice gave no false changes.

Reviews monitor accuracy

• A block is never an empty result. Each site gets its own proxy ladder (direct, then datacenter, then residential only where a site blocks the others), so one strict site never makes the rest of the run pricier.
• New means new. A review counts as new only when it was not seen before and is not older than what the last run read, so raising your limit is not a flood of "new" reviews. Amazon's top reviews count as new only when written in the last 14 days.
• Noise is not a change. Rating moves below 0.1 stars and small count moves are not reported, and a read that lacks a total carries the last one over instead of reading as a change.
• Removed is confirmed. A product page that answers 404 or redirects away is reported as removed only when two runs in a row see it; the first is a warning and is not billed. A redirect to the same brand's site for another country (brand.com to brand.co.uk) is followed.
• Themes are reproducible. They are computed in the actor with keyword and phrase matching (no AI call), with a negation check ("not comfortable" is no praise), so the same reviews always give the same themes.
• Each review once. A review a source repeats within one read (Trustpilot's data can hold a review twice) is delivered, counted in themes and charged once, by its id.
• Only delivered rows and checks are billed. Each delivered summary and review is charged, and with Changes only a product with nothing new, or a removal confirmed, is charged as a check (see Pricing below). A page without a supported review app, a block or a missing product is listed in the run summary with the reason, never charged.
• Only public addresses are read. A URL or redirect pointing into a private network is never fetched, and no site gets more than 2 requests at once.

Reviewer names and privacy

Reviewer names are personal data, so they are off by default (Reviewer names, includeReviewerNames: false). Review rows still carry the review id, rating, text, date and verified flag, which is what themes, alerts and comparisons need. Turn names on only if you have a lawful reason to process them; you then get the display name the source itself shows publicly (often a first name and an initial), never anything more.

Reviews monitor input

Field	Key	What it does
Product URLs	productUrls	Product pages of shops using Judge.me, Yotpo, Okendo, Loox or Stamped; Amazon product URLs too
Amazon ASINs	asins	Amazon ASINs (or URLs), one per line: star rating, ratings count, share per star and the top reviews the page shows
Trustpilot businesses	trustpilotDomains	Businesses on Trustpilot: domains or trustpilot.com/review/… pages (those URLs also work in Product URLs)
Amazon marketplace	amazonMarketplace	The marketplace for plain ASINs (US by default; URLs keep their own)
Reviews from the last (days)	sinceDays	Only reviews from the last N days (30 by default)
Reviews per product	maxReviewsPerProduct	The most reviews per product (20 by default; 0 = the rating summary only)
Trustpilot reviews per business (changes only)	trustpilotWatchlistReviews	With Changes only, the most recent Trustpilot reviews read per business at most (5 by default)
Complaint and praise themes	includeThemes	Complaint and praise themes on each summary row (on by default)
Reviewer names	includeReviewerNames	The public display name on each review (off by default, see Reviewer names and privacy)
Changes only	changesOnly	Return only what changed since the last run with the same monitor name; the first run is the baseline
Monitor name	monitorName	Runs with the same name compare against each other (default default)
Smallest rating move (stars)	minRatingChange	A rating move smaller than this is not reported (default 0.1)
Review count jump (%)	countJumpPercent	Report the count jumping or dropping by at least this percentage and at least 3 reviews (default 10)
Max cost per run (USD)	maxRunCostUsd	Optional. The most a run may charge: products are delivered in order, summary first, then reviews, while they fit, and a Trustpilot business is read with as many reviews as are left to pay for. Empty: only the run's spending limit applies
Only these fields	fields	Keep only these fields, for example ["productUrl", "rating", "text", "date"]; recordType is always kept
Products at a time	maxConcurrency	Products read in parallel (default 4); no site gets more than 2 requests at once
Proxy	proxyConfiguration	Optional. The actor picks proxies itself per site: direct, then datacenter, then residential only for sites that block the others
Start URLs	startUrls	Product URLs, same as Product URLs
History store	stateStoreName	Optional key-value store for what earlier runs saw (default cleanpull-reviews-monitor); pick one when running with a scoped token in Restricted access mode

Example inputs

A competitor watchlist, checked daily for new reviews:

{
  "productUrls": [
    "https://www.tentree.com/products/kessler-sweater-warm-oak-heather-fade",
    "https://mvmt.com/products/odyssey-ii-ceramic-fusion-grey-40mm-28000461"
  ],
  "asins": ["https://www.amazon.co.uk/dp/B0F1C56SR1"],
  "changesOnly": true,
  "monitorName": "competitors",
  "maxRunCostUsd": 0.5
}

The last 90 days of reviews with themes, up to 100 per product:

{
  "productUrls": ["https://mvmt.com/products/odyssey-ii-ceramic-fusion-grey-40mm-28000461"],
  "sinceDays": 90,
  "maxReviewsPerProduct": 100
}

Ratings only (average, count, share per star) for a list of ASINs in the UK:

{ "asins": ["B000MXIB0G"], "amazonMarketplace": "GB", "maxReviewsPerProduct": 0 }

Reviews monitor output

Three kinds of rows; recordType says which.

Example rows

A real rating summary:

{
  "recordType": "rating-summary",
  "source": "yotpo",
  "sourceName": "Yotpo",
  "productUrl": "https://mvmt.com/products/odyssey-ii-ceramic-fusion-grey-40mm-28000461",
  "productKey": "cbf4e98e01272e1b",
  "domain": "mvmt.com",
  "productTitle": "Odyssey II Ceramic Fusion Grey 40mm",
  "productId": "10221707821304",
  "average": 4.83,
  "count": 6,
  "histogram": { "1": 0, "2": 0, "3": 0, "4": 1, "5": 5 },
  "histogramUnit": "count",
  "reviewsRead": 5,
  "newReviews": 0,
  "newestReviewDate": "2026-09-08T00:00:00.000Z",
  "oldestReviewDate": "2025-11-30T00:00:00.000Z",
  "themes": [
    {
      "theme": "uncomfortable",
      "polarity": "complaint",
      "kind": "known",
      "reviews": 1,
      "share": 0.2,
      "exampleReviewIds": ["882971486"]
    },
    {
      "theme": "looks great",
      "polarity": "praise",
      "kind": "known",
      "reviews": 4,
      "share": 0.8,
      "exampleReviewIds": ["894671816", "882971386", "882971411"]
    }
  ],
  "topComplaints": ["uncomfortable"],
  "topPraise": ["looks great", "durable"],
  "changes": [],
  "firstSeen": false,
  "readThroughResidentialProxy": false,
  "scrapedAt": "2026-10-14T23:13:47.822Z"
}

A review:

{
  "recordType": "review",
  "source": "okendo",
  "sourceName": "Okendo",
  "productUrl": "https://www.tentree.com/products/kessler-sweater-warm-oak-heather-fade",
  "productTitle": "Womens Kessler Sweater",
  "reviewId": "0c3c5dc4-3d01-4c81-8590-4ee671ba3fda",
  "reviewKey": "9f2f9369161887dd4a52eaa1ee1bee61234851ca23a6c435bde0572cf4c94663",
  "rating": 5,
  "title": "Girlfriend love it",
  "text": "Same old quality as tentree always offered",
  "date": "2026-10-14T12:23:30.510Z",
  "verified": true,
  "variant": "DEEP BURGUNDY HEATHER FADE / XS",
  "helpfulVotes": 0,
  "language": "en",
  "isNew": null,
  "scrapedAt": "2026-10-14T23:15:03.782Z"
}

Changes and other fields

A change looks like { "recordType": "change", "changeType": "new-reviews", "newReviews": 2, "reviewIds": ["r3", "r2"], "lowestRating": 1, "averageOfNew": 1.5 }; the other types are rating-moved (from, to, delta in stars), count-jumped and count-dropped (delta, percent) and removed.

• histogramUnit is count for the review apps and percent for Amazon (Amazon publishes shares, not counts). Stamped and Loox publish no counts per star (histogram: null).
• isNew is null on a product's first run (the baseline), then true or false.
• reviewKey is a SHA-256 of source, product and review id: stable across runs, for joining and deduplication.
• Also: asin, marketplace (Amazon), finalUrl (after a redirect), country (Amazon, Stamped), mediaCount, previous (last run's average and count, on rows with changes).

Dataset views: Rating summaries, Reviews, Changes and Themes (one row per theme). RUN_SUMMARY in the key-value store gives the outcome (COMPLETE, PARTIAL, NO_CHANGES, EMPTY, FAILED), the sources seen, charges, and every product that could not be read, with the reason.

Pricing

Event	Free	Bronze	Silver	Gold and above
Product rating summary (one product read, with totals, stars and themes)	$3.00	$3.00	$2.70	$2.40
Review	$0.50	$0.50	$0.45	$0.40
Product checked: unchanged in Changes only, or a removal confirmed	$0.50	$0.50	$0.45	$0.40
Request through residential addresses (add-on, only where a site blocks others)	$6.00	$6.00	$6.00	$6.00
Provider record (a Trustpilot review delivered, or re-checked on a watchlist run)	$6.00	$3.00	$2.70	$2.40
Provider search (one Trustpilot business read)	$6.00	$3.00	$3.00	$3.00

Per 1,000. Each run also has a $0.00005 start fee. Reviews are charged only when delivered: a US Amazon product (no review text signed out) costs its rating summary alone, $0.003. The add-on applies per request a site made us send through residential addresses; such requests are capped at 600 KB transferred (a larger page is not read or charged). In tests no review app or US Amazon page needed it; Amazon outside the US is read from that country's residential addresses, so each of those products carries one or two. Change rows (new reviews, rating moved, count jumped or dropped) are free: they come with the product's rating summary. A removed product's row is billed as a product check. Trustpilot's provider search and re-checked reviews are billed with the business's first row (or its check), and a delivered review's provider record with its row; a read that fails or is blocked is not billed. A Trustpilot business on a daily watchlist with the default 5 reviews costs at most $0.0185 on Bronze ($0.0365 on the Free plan) on a day with nothing new (a check, a search and 5 re-checked reviews), and usually $0.0065 on Bronze ($0.0125 on Free) once the actor knows the business (one re-checked review): about $0.20 a month on Bronze, $0.38 on Free. A daily watchlist of 100 products with Changes only is about 3,000 checks a month: $1.50 on the Free and Bronze plans, plus the summaries and new reviews on days something changed.

Limits

• Sources this actor does not read: Amazon's full review list needs a sign-in; Google Shopping publishes reviews only inside pages that need a browser running scripts. Undated Amazon reviews (a date line Amazon wrote in a way the actor cannot read) never count as new. Pages from shops using Bazaarvoice, Junip, REVIEWS.io, PowerReviews and other apps are named in the run summary and not charged.
• Review apps loaded from a theme's own scripts (some large stores) expose no ids in the page; the run summary says so.
• Loox shows no review dates in some widget layouts: those reviews are not cut by Reviews from the last (days) (the limit per product still applies), and stars per level are not published.
• Judge.me shops can group several products under one review set (for example one per scent); the totals and reviews are the group's, as the shop shows them.
• Amazon shows the top reviews, not the newest, and sometimes leaves them out of a page: the actor then reads the page's sections. A product whose reviews still did not come back is delivered with its rating summary and reviewsNotShown saying why. When Amazon left them out of every read or blocked the sections request (reviewsNotShownKind: "left-out", a passing miss), the run ends PARTIAL, never COMPLETE. A sign-in prompt ("sign-in", amazon.com, see above) is a note and the run stays COMPLETE. A product with star ratings but no written review shows an empty reviews card: nothing is missing and the run stays COMPLETE.
• Pace: at most 2 requests at once per site. Read at a pace and scale the sites' terms allow.
• Nothing found: when nothing could be read because it is not there (a product page that is gone, a page with no supported review app, input naming no product), the run still succeeds, with one free recordType: "unreadable" row per item giving the reason; nothing is charged. These free rows come only when the run delivers nothing; otherwise the misses are listed in RUN_SUMMARY. A run near its time limit stops starting products, delivers what it read and lists the rest in RUN_SUMMARY (as free rows when it delivered nothing). If the run delivers nothing and any product was blocked or failed with an error (not one of these misses), the run ends FAILED with the reasons in RUN_SUMMARY, no free rows are given, even for the misses, and nothing is charged.

FAQ

How much does review monitoring cost per product per day?

A product on a review app checked once a day with Changes only costs $0.0005 a day on the Free and Bronze plans when nothing is new (about $0.015 a month), and on a day with new reviews $0.003 for its rating summary plus $0.0005 per new review. A day when only the rating or the review count moved, with no new review, costs the $0.003 rating summary. Without Changes only, each run delivers the rating summary and up to Reviews per product reviews: $0.013 a run for a product with 20 reviews in the window. A US Amazon product (no review text signed out) costs its rating summary alone, $0.003; Amazon outside the US adds $0.006 for each of the one or two requests read through that country's residential addresses. A Trustpilot business on a daily watchlist usually costs $0.0065 a day on Bronze ($0.0125 on the Free plan) once the actor knows it, and at most $0.0185 on Bronze ($0.0365 on Free) on a day with nothing new. Change rows are free, and a page without a supported review app, a block or a missing product is never charged. Each run also has the start fee, and higher Apify plans pay less (see Pricing below).

How often should I check for new reviews?

Daily suits most watchlists, and weekly works for slow-selling products. With Changes only and a schedule, each run returns only new reviews, rating moves and review count jumps or drops since the last run under the same Monitor name; Smallest rating move (stars) and Review count jump (%) keep noise out. A removed product is reported only when two runs in a row see it gone. For a busy Trustpilot business, raise Trustpilot reviews per business (changes only) above the number of new reviews it gets between runs.

What is the difference between the rating summary, reviews and themes?

The rating summary row (recordType: "rating-summary") gives the product's or company's average, review count and histogram, the share per star (histogramUnit says whether it holds counts or, on Amazon, percentages). Review rows (recordType: "review") are the individual reviews read, newest first (on Amazon, the top reviews the page shows). themes, topComplaints and topPraise on the summary row group the reviews read into complaints and praise, with how many reviews mention each and example review ids.

Why are reviews or a rating missing?

• amazon.com (US): Amazon shows review text only to signed-in shoppers, so a US ASIN gives the rating, count and share per star, with reviewsNotShown saying why (reviewsNotShownKind: "sign-in").
• Left out by Amazon: when Amazon left the reviews out of every read, the summary row has reviewsNotShownKind: "left-out" and the run ends PARTIAL.
• Outside the window: only reviews from the last Reviews from the last (days) are read, up to Reviews per product; 0 gives the summary only.
• No stars per level: Stamped and Loox publish no counts per star (histogram: null).
• Unsupported review app: pages from shops using Bazaarvoice, Junip, REVIEWS.io, PowerReviews and other apps, or apps loaded from a theme's own scripts, are named in RUN_SUMMARY and not charged.
• Nothing read: when the whole run delivers nothing and every product is a miss (gone, no supported review app), each gets a free recordType: "unreadable" row with the reason; if any product was blocked or failed with an error, the run ends FAILED with no free rows.

Which review apps and sites does it support?

Judge.me, Yotpo, Okendo, Loox and Stamped on shop product pages (most Shopify stores), Amazon product pages by ASIN or URL on the marketplace you pick in Amazon marketplace (or the URL's own), and businesses on Trustpilot by domain or trustpilot.com/review/… page. Amazon's full review list and Google Shopping reviews are not read.

Can I use the reviews monitor through the API?

Yes. Start runs through the Apify API or an API client, and read the dataset through its Rating summaries, Reviews, Changes and Themes (one row per theme) views, or the RUN_SUMMARY record with the outcome, sources, charges and every product that could not be read. Only these fields (fields) trims each row to the keys you need, and reviewKey is stable across runs for joining.

Does it work with Google Sheets, Make, Zapier or Claude?

Yes. Apify's integrations send each run's results to Google Sheets, Make, Zapier, webhooks and other tools, and a schedule keeps them updated. From Claude, the MarketMage plugin reads what customers say about a product, the complaint and praise themes, or watches for new and bad reviews, in a chat.

Is it allowed to scrape product reviews?

The actor reads public review data as a signed-out visitor sees it: no login, no account and no customer data. Review rows keep the review's rating, title, text, date and id; reviewer names are left out unless you turn on Reviewer names, and then only the display name the source shows publicly. Trustpilot reviews come through a third-party web data provider. How you use the data, and following the laws and terms that apply to you, is your responsibility. This is not legal advice.

Related actors

Other CleanPull actors for monitoring:

• Competitor Promotions & New Products Monitor: competitor sales, discount codes and new products.
• Price Tracker & Stock Monitor for Any Online Store: price and stock changes on product pages of any online store.
• Amazon Product Scraper: Amazon price, Buy Box, every seller's offer and Best Sellers Rank by ASIN.
• Watch Digest: one scheduled digest of price, stock, rank and review changes.

Use it from Claude

The MarketMage plugin runs these actors from a Claude chat.

Found it useful? A rating on the Store helps others find it.

Changelog

• 0.2.x, Shopify bot check (2026-11-07): a Shopify store whose bot check refuses its product pages is read without them: the product's id and title from the store's catalog (/products.json), its review app and the app's public ids from the store's home page (else /collections/all), then the reviews from the app's own data as before. These store reads use direct and datacenter addresses only, so they never add the residential add-on; a later product of the same store skips its page at once. On a Shopify store, a product page that asks to wait more than 5 s is tried on a fresh session at once (other stores still wait up to 60 s). When the home page does not show the review app (some stores load it only on product pages), the product fails as blocked, as before, with nothing charged. Prices and events unchanged.

• 0.2.x, small limits (2026-11-01): a small spending limit or Max cost per run now delivers what it can pay for instead of nothing. A Trustpilot business is read with as many reviews as the limit can still pay for (the read and its rating summary must fit; otherwise it is not read), and a page or Amazon product is read and delivered in order, summary first, then reviews, while they fit, instead of being refused because of a worst case of residential add-ons. Products read at the same time each keep room for their own read and summary, so several Trustpilot businesses at a small limit each get their summary. No row is delivered unpaid, and no Trustpilot record is bought past what the limit can pay. The status says what was delivered and which limit stopped the rest; RUN_SUMMARY.shortenedByLimit counts Trustpilot businesses read with fewer reviews than asked. The form's example and the empty-input example are now live product pages (the Beardbrand example page was gone). Prices and events unchanged.

• 0.2.x, reliability (2026-11-01): a product address on a store that does not exist (no such domain) now ends with a free unreadable 'not found' row instead of a failed run; the address is checked before any proxy is used, so it costs nothing. Prices and events unchanged.

• 0.2.x, fewer reads (2026-10-30): Trustpilot watchlists read only as many recent reviews as the business's review total grew last run, plus one (one for a quiet business), and read the usual number when that might miss a new review (a short read that did not reach a review seen before, or a total that grew by more than the new reviews found). You are charged only for the reviews re-checked, so quiet businesses cost less. www.shop.com and shop.com (any case, with or without a trailing slash) are one business, read once. RUN_SUMMARY gains pageReads, recordsRead, trustpilotReadAgain and rows; the final status ends with "N pages/records read for M rows". Prices and events unchanged.

• 0.2.x, fix (2026-10-27): amazon.com now shows review text only to signed-in shoppers. The actor recognises Amazon's sign-in prompt, reads the page once (no retries or sections request for it), and delivers the rating summary with reviewsNotShown saying why. Products without the reviews asked for are listed in RUN_SUMMARY.reviewsNotShown with their kind and reason (kept across a migration), with one warning line per kind, also on Changes-only runs. Reviews left out of every read or a blocked sections request end the run PARTIAL (it used to be COMPLETE with a warning); the sign-in limit is a note. The form's Amazon example is now a UK product, whose page shows its reviews. Prices and events unchanged.

• 0.2.x, reliability (2026-10-24): runs that find nothing (a product page that is gone, a page with no supported review app, input naming no product) end SUCCEEDED with a free unreadable row per item saying why, instead of FAILED; a run stops starting products before its time limit and ends with what it read, never TIMED-OUT. A run where the site refused on every route for every product still fails. Prices and events unchanged.

• 0.2, fix (23 October 2026): a review the source delivers twice in one read is kept once (Trustpilot could return the same 10 reviews twice, doubling rows, theme counts and charges).

• 0.2 (October 2026): Trustpilot businesses, read through CleanPull's marketplace data service (no setup): reviews newest first, incremental between runs, with the company's rating, count and share per star. New events: provider record (a review delivered, or re-checked on a watchlist run; reviews dropped by the date filter are not charged) and provider search ($0.003 per business read on Bronze, $0.006 on the Free plan). Watchlists read 5 recent reviews per business by default (trustpilotWatchlistReviews).

• 0.1 (October 2026): reviews and totals from Judge.me, Yotpo, Okendo, Loox and Stamped widget data and from Amazon product pages; complaint and praise themes; watchlists with new reviews, rating and count changes and confirmed removals; spending-limit and per-run cost caps; reviewer names off by default.