Mobile App Reviews Scraper & Version Radar

Unofficial Actor. Not affiliated with, endorsed by, or sponsored by Apple Inc.

Find iPhone and iPad apps by Apple ID or broad search term, collect recent public App Store customer reviews, and flag evidence-backed rating drops between observed app versions. Stable state deduplicates review IDs across successful runs.

Search-friendly use cases include Apple App Store review scraping, iOS app review monitoring, competitor review tracking, release-quality monitoring, version sentiment comparison, and rating regression detection.

What the Actor returns

Each default-dataset item is one newly observed public review with:

• Apple app ID, app name, storefront, and App Store link;
• rating, title, text, public author display name, review timestamp, and public App Store reviews-page link;
• the current store version reported by Apple's Lookup/Search API;
• the store-version cohort in which the Actor first observed the review, plus accumulated sample counts and average ratings for the current and immediately preceding monitored versions;
• regression, no_regression, or insufficient_evidence, plus the measured rating delta when the configured evidence gate is met.

The run summary and bounded fetch failures are stored in the default key-value store. They are not dataset results.

Quick start

{
  "searchTerms": ["budget planner", "expense tracker"],
  "appIds": ["284882215"],
  "country": "us",
  "maxApps": 5,
  "searchResultsPerTerm": 5,
  "maxReviewPages": 2,
  "maxReviews": 100,
  "minimumReviewsPerVersion": 5,
  "regressionThreshold": 0.5,
  "stateKey": "weekly-finance-watch"
}

Explicit appIds are resolved first. Search results then fill the remaining maxApps slots in Apple response order, with duplicate app IDs removed. The first successful run emits currently visible reviews up to maxReviews; later runs using the same stateKey emit only review IDs not already stored. A zero-item repeat run is valid.

How the regression signal works

Apple's public web review response does not report the reviewer's installed app version. The Actor therefore attributes each newly observed review to the store version reported by Apple's Search/Lookup API during that run. This is collection-cohort attribution, not a claim about the version installed by the reviewer.

For each app, the Actor accumulates rating count and rating sum once for every emitted review ID in that observed store-version cohort. It compares the current store version with the greatest lower, naturally ordered version observed in state. A signal is only classified after both versions reach minimumReviewsPerVersion:

• regression: current average minus comparison average is at most the negative regressionThreshold;
• no_regression: both evidence gates are met, but the drop is smaller;
• insufficient_evidence: one version is missing or either sample is too small.

This is a deterministic monitoring-cohort comparison, not sentiment AI, causal analysis, anomaly prediction, exact per-review version data, or proof that a release caused a rating change. Meaningful cross-version evidence requires using the same stateKey across at least one store-version change.

Bounded operation and PPE

Inputs cap one run at 20 apps, 5 search terms, 20 search candidates per term, 10 review pages per app, and 1,000 emitted reviews. HTTP responses are capped at 5 MB. The Actor uses only public HTTPS Apple endpoints and needs no secrets, cookies, login, proxy, browser, generative model, or fal credential. It runs once and exits; Standby is not used.

The launch hypothesis uses Apify's synthetic apify-default-dataset-item event at $0.0005 per visible review. The code makes exactly one Actor.push_data() call for each visible review result and makes no explicit charge call, so every visible default-dataset review maps to one synthetic event. Summary and failure records do not produce dataset-item events. Platform usage and final pricing must match the live Apify Console configuration.

Public sources

• Apple Search API: https://itunes.apple.com/search
• Apple Lookup API: https://itunes.apple.com/lookup
• Apple App Store public web review JSON: https://apps.apple.com/api/apps/v1/catalog/{country}/apps/{appId}/reviews?platform=web&sort=recent&offset={offset}

Important limitations

Apple controls the public endpoint availability, search ranking, storefront coverage, review ordering, moderation, localization, fields, and history depth. The public web API is undocumented and can change or stop working without notice; it is a bounded recent sample, not a complete historical export. Some apps/storefronts have no accessible reviews. The review response does not expose reviewer-installed app version, and the reviews-page link may not deep-link to one review. Natural version ordering handles ordinary numeric versions but is not a complete SemVer implementation.

Deduplication is stateful and best-effort across successful runs. It retains at most 5,000 review IDs and 50 version buckets per app. Changing stateKey, deleting storage, exceeding retention, or a process interruption between dataset and state writes can allow a review to appear again. Reviews beyond maxReviews are not marked seen, so they can appear in a later run.

Public review text and author display names are user-generated content. Follow applicable Apple terms, privacy rules, and downstream platform policies. Do not treat the signal as a release verdict, support priority, market-research guarantee, or substitute for reading the underlying reviews.

Local tests

From this Actor directory:

python -m unittest discover -s tests -v

Tests use fixtures only and do not call Apple or Apify.