App Store Review Monitor — New & Edited Reviews by Version/Country

Turn recent Apple App Store reviews into a daily product feedback feed. Monitor selected apps and countries, filter for ratings or keywords, and identify reviews that are new or edited compared with your own previous snapshot. Export clean review rows to CSV, JSON, a database, or your existing automation.

Price: $0.50 per 1,000 delivered reviews ($0.0005 each). No startup event. Failed feeds, invalid reviews, duplicates, filtered reviews and undelivered reviews are uncharged. Coverage diagnostics, the next snapshot and rating rollups of the delivered sample are included at no additional Actor event charge.

Who it helps

App product teams need a fresh review triage list after releases so they can find regressions and decide which fixes to prioritize. Agencies managing several apps need a repeatable view across markets. Instead of comparing CSV exports manually, supply the previous snapshot and receive only new or changed review content, with stable identifiers and the source feed URL.

The Actor adds country selection, bounded pagination, rating/version/keyword filters, change detection, integration identifiers, billing limits and sample rollups to Apple's public review feeds. It uses deterministic parsing; it does not generate sentiment or summaries with an AI model.

Try a small run

{
  "apps": ["618783545"],
  "countries": ["us"],
  "maxPages": 1,
  "maxResults": 50,
  "maxRunSeconds": 120
}

This inspects one page of recent US reviews for the example app. Up to 50 delivered rows cost at most $0.025 in Actor events. The default run charge cap is $0.05. Apple may return fewer reviews. You can also supply an official URL such as https://apps.apple.com/us/app/slack/id618783545; the URL supplies the app ID and countries controls the requested storefronts.

Open the Reviews and changes dataset view for usable rows. Open OUTPUT in key-value storage for coverage, the snapshot and per-version/country sample rollups. An empty dataset with OUTPUT.status = "succeeded" means the requested window was inspected and no review matched; check its filters and coverage before interpreting the result.

A recurring review workflow

1. Run once for the apps and countries you track. Review OUTPUT.complete and each entry in OUTPUT.coverage.
2. Save OUTPUT.snapshot.previousReviews in your own workflow storage.
3. On the next run, send that array as previousReviews and set changesOnly to true. Keep the same scope for a comparable window.
4. Send delivered new and updated rows into your existing triage queue. Use reviewKey to update an existing review record; use observationId to deduplicate a particular content version.
5. Replace the workflow snapshot with the returned snapshot after reviewing coverage. A partial run preserves undelivered baseline hashes and advances only delivered records.

No schedule, shared named store, webhook or external notification is created by this Actor. Configure frequency and destination in your own Apify task, API client, Make, n8n or other workflow.

Example for a second run:

{
  "apps": ["618783545"],
  "countries": ["us"],
  "maxPages": 2,
  "maxResults": 100,
  "maxRating": 2,
  "keywords": ["crash", "login"],
  "changesOnly": true,
  "previousReviews": [
    {
      "appId": "618783545",
      "country": "us",
      "reviewId": "14749821280",
      "contentHash": "d5ce7c008641885ed94ef936eaf215f57b99ddab71280b69df8fc954c35835c1"
    }
  ]
}

Use the complete snapshot from your previous run; the single-record array illustrates its shape. Changing filters does not retroactively add filtered reviews to the baseline. A review not previously delivered can therefore be marked new when it first matches.

What a review row contains

This excerpt comes from the example app's public US feed, inspected on November 5, 2026. Review text can change later. Reviewer names and author IDs are omitted.

{
  "reviewKey": "618783545:us:14749821280",
  "appId": "618783545",
  "country": "us",
  "reviewId": "14749821280",
  "changeType": "observed",
  "rating": 5,
  "title": "“Catch Up” feature is game changing",
  "text": "Plz Salesforce Don’t Ruin It",
  "appVersion": "26.10.40",
  "updatedAt": "2026-11-04T12:52:29Z",
  "matchedKeywords": [],
  "sourceUrl": "https://itunes.apple.com/rss/customerreviews/page=1/id=618783545/sortby=mostrecent/json?cc=us"
}

Each complete row also includes schemaVersion, observationId, contentHash, previousContentHash and observedAt. matchedKeywords contains the supplied terms found in the title or text, using case-insensitive substring matching. An empty array means no keyword filter was supplied.

Field	Meaning
reviewKey	Stable app/country/review identity, suitable for an upsert key.
observationId	SHA-256 identifier for that review identity and content version.
changeType	observed without a supplied baseline; otherwise new, updated or unchanged. changesOnly excludes unchanged rows.
contentHash	Fingerprint of normalized rating, title, text and app version. A timestamp-only change does not count as an edit.
previousContentHash	Prior fingerprint for the same review, or null. The minimal snapshot does not retain previous text or rating values.
updatedAt	Apple's source timestamp, normalized to UTC.
observedAt	UTC time this run began collecting its frozen result plan.
appVersion	Version reported with the review, or null when absent.

new means absent from your supplied snapshot. It does not prove the review was first written during this run. Passing previousReviews: [] intentionally starts an empty baseline. Omitting it keeps the status observed; changesOnly: true requires an explicit baseline.

Inputs and bounds

Input	Default	Accepted scope
apps	Required	1–10 app IDs or official HTTPS Apple App Store URLs. Duplicate IDs are removed.
countries	["us"]	1–5 supported country codes; maximum 20 app-country pairs. Duplicates are removed.
maxPages	1	1–10 pages per app-country, up to 50 reviews per page.
maxResults	100	1–1,000 delivered reviews across the run.
minRating, maxRating	1, 5	Inclusive integer rating range.
keywords	[]	Up to 20 terms, 100 characters each; any case-insensitive title/text substring matches.
versions	[]	Up to 20 exact app version strings; empty accepts all versions.
since	Omitted	Inclusive ISO-8601 timestamp with timezone, for example 2026-11-01T00:00:00Z.
changesOnly	false	Deliver only new/updated rows compared with explicit previousReviews.
previousReviews	Omitted	Up to 20,000 minimal snapshot records.
maxRunSeconds	180	60–240 seconds, including a 25-second collection reserve for delivery. Use a platform timeout of at least 300 seconds.
requestTimeoutSeconds	20	3–30 seconds per bounded source request including safe retries and robots checks.

Supported codes: ae ar at au bd be br ca ch cl cn co cz de dk eg es fi fr gb gr hk hu id ie il in it jp ke kr lk mx my ng nl no nz pe ph pk pl pt ro sa se sg th tr tw ua us vn za.

Scopes are processed in app order, then country order, with pages from most recent onward. A small overall result or charge cap can stop before later countries or apps. Inspect coverage rather than assuming every supplied scope was visited. since and filters apply within the fetched recent window; they do not extend it or request historical backfill.

Coverage, snapshots and rollups

OUTPUT.complete means all requested bounded windows were inspected without source errors and every planned matching row was delivered. It never means the app's full review history was collected. Each app-country reports pages fetched, source/invalid/duplicate entry counts, planned/delivered reviews, requestedWindowComplete, recentFeedExhausted and a stop reason. configured_page_limit is the requested boundary, while feed_exhausted means a short page was observed at that time.

The snapshot carries forward earlier records and updates only rows actually delivered. It does not advance filtered, skipped or undelivered reviews. At 20,000 records, additional new review identities are skipped without charge, snapshot_limit is reported, and existing review updates remain eligible. The Actor never silently evicts baseline hashes; rotate or partition your baseline intentionally when that bound is reached. Changing a baseline's scope can make previously unknown reviews appear new.

versionCountryRollups groups the delivered sample by app, country and version. It includes the sample size, average rating, counts for each star rating and change counts. These values are affected by your filters, result caps and current feed window. They are not the app's overall store rating or a representative survey of all customers.

Source failures and malformed entries are available in OUTPUT.errors without a review event charge. A fully failed source collection fails the run; a valid empty match succeeds. Partial results are retained and marked partial. The Actor does not infer deleted reviews from missing entries, since Apple's recent window moves and can omit older or edited reviews.

Billing and recovery

Each delivered default-dataset row triggers the single apify-default-dataset-item event. There is no second custom charge. The run's remaining charge allowance limits both collection and delivery; a zero allowance fetches no feeds. Examples: 3 delivered rows cost $0.0015, 50 cost $0.025, and 200 cost $0.10 in Actor events. Platform charges outside the Actor's event price, if applicable to your account, are shown by Apify.

Results are frozen before delivery. Within the same run's resurrection, the Actor compares saved output with that frozen plan and native event counts, then resumes without reposting confirmed rows. A lost write response leaves a pending checkpoint; an unresolved append fails safely instead of risking a repeated billable write. A new run with changesOnly: false intentionally returns and charges its selected reviews again. Use the snapshot workflow to avoid repeatedly purchasing unchanged reviews.

Source and limitations

• Only public Apple customer-review RSS JSON feeds are requested. There is no App Store Connect login or private review access.
• Source URLs use the fixed itunes.apple.com host and explicit cc country selection. Supplied App Store URLs are parsed for IDs, never fetched directly. Links embedded in source data are never followed.
• Robots policy is respected. DNS must resolve only to public destinations; HTTPS, redirects, response sizes, request counts and execution time are bounded. No deployment credential is sent to Apple.
• Apple exposes a recent window of at most 10 pages and 50 reviews per page for this endpoint. It may rate-limit, change formats, return fewer reviews or omit older edits. HTTP 429/503 stops further same-origin collection in that run.
• Live pagination can change while a run is in progress. Duplicate review identities are not delivered twice; contradictory content or repeated pages mark coverage incomplete.
• Reviews are preserved as user-written text and are untrusted content. No reviewer names or author IDs are included. The Actor does not verify claims made in reviews.
• A 20 MiB encoded plan bound, a 20,000-entry snapshot bound and per-field size limits keep processing finite. A limit or malformed field is reported rather than silently billed as a complete review.

Local verification

The project includes deterministic fixtures for country provenance, parser shape changes, malformed/empty responses, filters, pagination, edits, snapshot limits, deadlines and retry safety. Runtime failure injection exercises the pinned Apify client's actual dataset write path before commit, after commit with a lost response, and on restart with pending rows visible or unavailable.

uv run python -B -m unittest discover -s tests -v
uv run python -B tools/local_smoke.py --cases /path/to/private/cloud-cases.json --case full
uv run python -B tools/validate_cloud.py /path/to/private/run /path/to/private/cloud-cases.json full /path/to/private/source-evidence

Raw live feed evidence remains in ignored private storage. The source validator independently compares rating, title, text, version and UTC timestamp against Apple's original entries; dataset schema validation alone is not treated as output correctness.