This changelog highlights releases that materially changed Trustpilot Reviews Scraper for users. Dates use UTC; small internal refactors and routine maintenance are intentionally omitted.
Reliability
- The Actor now needs at least 512 MB of memory. At 256 MB the headless browser ran out of memory and the run failed to launch or hung until its timeout. Keep the 1024 MB default for multi-company runs.
Reliability
- When a proxy connection fails, the Actor now switches to its other proxy pools for later attempts and keeps the faster retry allowance. Before, a company could fail after 3 quick attempts on the same failing proxy pool.
- A company now gets at most 6 attempts in total, whatever mix of blocks, proxy failures and slow pages it meets. Before, proxy failures after earlier failures could add up to 11 attempts.
Fixes
- Selecting many languages works. With more than 5 language codes, the Actor now reads Trustpilot's all-languages listing once and keeps the reviews in your languages. Before, each code was a separate request, Trustpilot rate-limited long lists and a run could end with 0 reviews.
- Language codes are no longer case-sensitive:
DE works like de.
- A rate-limited page (HTTP 429) is retried once after a short wait instead of being skipped.
Reliability
- A run stopped close to its timeout now waits for reviews it is still saving. Previously it could end as failed with "Got 0 reviews" while those reviews were charged and written a moment later.
- The final status message of a run stopped before its timeout can no longer be replaced by a progress line from a request still in flight.
- A run that has no time left when it starts, for example a very short timeout after a slow start, no longer charges the start fee. It fails with a message to increase the timeout.
Fixes
- Selecting more than one language works again. Since 2026-11-03, runs with two or more language codes failed on every company because Trustpilot rejects that request. Each language is now read as its own listing, and
maxReviewsPerCompany is shared between the languages, with any unused share going to the others.
Reliability
- A run that saved reviews now ends as succeeded even if some companies could not be read, and its status message gives how many reviews you got, from how many companies, and why the rest are missing. Previously it ended as failed although the saved reviews were delivered and charged. A run that delivered nothing still fails.
- A run close to its timeout now stops in time, keeps the saved reviews and ends as succeeded with a clear status, instead of timing out. If nothing was saved it fails and says to increase the timeout.
- Reviews are written only after their charge is confirmed, and a run stops at its spending limit without saving reviews beyond it. Prices are unchanged.
Filters and input
- Star, language, verified-only and replies-only filters are now applied by Trustpilot itself, so filtered runs page through matching reviews. Previously the filters were applied only to the newest unfiltered pages: for example, a 5-star-only run on a large company returned 8 reviews instead of up to 200. Selecting several ratings now returns up to 200 reviews per rating. The first page of every filtered listing now uses the filters too; before, it always showed the newest unfiltered reviews and ignored the date filter.
languages accepts all for reviews in every language. An empty languages keeps Trustpilot's default view, which is usually English only; the README now says so.
- Company website URLs with a page path (for example
https://www.example.com/en/pricing) are reduced to the domain. Entries that are not a domain are skipped with a warning, and a run with no usable entry now fails before the $0.005 start fee is charged.
- The final status message now summarises the run, for example "Done: 0 reviews from 0 of 1 company. No reviews matched the filters", including profiles not found and skipped entries.
authorImage is null instead of an empty string when the reviewer has no avatar.
- The README now states the real per-company ceiling (about 1,000 reviews per run) instead of "no limit". Prices are unchanged.
Reliability
- After Trustpilot blocks one proxy identity, each later identity is first checked with one small request, and blocked ones are skipped before the browser uses them. Previously, during widespread block episodes, a company could still fail after six blocked attempts because each blocked network range had to be discovered by a full browser attempt.
- A third datacenter proxy pool is now available for these episodes. Runs without a block are unchanged, and prices are unchanged.
Reliability
- When Trustpilot blocks several proxy network ranges in the same run, later identities now come from a second datacenter proxy pool. Previously a company could fail after six blocked attempts during these short, widespread block episodes.
- Fast bot-block retries no longer use up the two-retry limit for slow page loads. Previously a company that was blocked twice and then hit one slow page failed without another try.
- Choosing a fresh proxy identity after a block now checks at most six candidates in total across both proxy pools, so retries during block episodes start sooner.
Reliability
- When Trustpilot blocks a browser identity, remember the proxy network range it came from and pick later identities from other ranges. Previously a retry could land in the same blocked range again, so some companies failed after six blocked attempts.
Reliability
- Recognize Trustpilot's explicit bot-block response immediately and retry the company with up to five fresh browser identities, instead of waiting out a 30-second page deadline on each attempt. Single-company runs that were blocked on every attempt now usually recover, and jobs that remain blocked fail faster. Slow navigation and page-load failures keep the existing two-retry limit.
Reliability
- Keep the last observed batch milestone in periodic run status updates instead of replacing it with a generic crawler message. Its timestamp continues to show when work was observed, not when status was refreshed.
Fixes
- Reject malformed company lists before scraping with a clear, indexed validation message instead of an internal parsing error. Company entries must be non-empty domain or URL strings, not URL objects.
Reliability
- Show identity-free batch progress, pagination and storage milestones so long-running and timed-out jobs can be diagnosed without mistaking an old loading message for the current operation.
Reliability
- Skip profiles explicitly removed by Trustpilot with a clear warning and no review charges, rather than failing other companies in the job. Unknown and blocked responses still fail after bounded retries.
- Clarified the recommended browser memory allocation for callers overriding the default.
Reliability
- Skip profiles that Trustpilot explicitly confirms do not exist, with a clear warning, instead of retrying and failing otherwise useful multi-company jobs. Blocked or unrecognized responses still fail after bounded retries.
Reliability
- Wait for complete application data on streaming company pages without waiting for unrelated page scripts.
- Read available review data without waiting for unrelated page scripts, and renew browser identity after failed attempts.
- Mark jobs with exhausted company requests as failed rather than reporting incomplete output as successful.
Integrations
- Added private, sanitized failure diagnostics to help investigate failed review jobs; documented diagnostic retention and privacy protections.
Reliability
- Reduced unnecessary browser rendering resources while preserving review extraction.
Features
- Added an exact
publishedAfter cursor for incremental review monitoring, with strict timestamp boundaries and early pagination cutoff when sorting by recency.
Reliability
- Continued large unfiltered review jobs beyond Trustpilot's new anonymous 10-page limit by safely partitioning requests and removing duplicates.
Features
- Added public Store task examples so users can start common Trustpilot review scraping workflows faster.
Changes
- Improved input form labels, examples, and help text to make setup clearer.
Reliability
- Reduced run cost risk by tightening navigation timeouts, retry behavior, memory defaults, and slow WAF waits for review scraping runs.
Reliability
- Improved timeout handling for default review-scraping runs so local and scheduled health checks complete more reliably.
Reliability
- Restored the Actor after Apify quality checks flagged repeated failed Trustpilot review runs.
Features
- Initial release for extracting Trustpilot reviews with ratings, replies, reviewer information, and company trust data.