Google Maps Reviews Scraper | Filtered Rows

What you can export

Use this Google Maps reviews scraper to export individual review rows from place URLs, IDs or ranked business searches. Filter by stars, activity-date window and text, and choose the supported review order.

Read the final COVERAGE record for each place. Activity dates can reflect edits; original and absolute reply dates are available only on successfully enriched owner-reply rows.

Try the example input

Use the Console prefill or supply your own place identifiers. Empty input {} starts no source work, exits successfully and logs guidance.

{
  "placeIds": [
    "ChIJ0SnrKjO0RIYRCg9HH9d0gqE"
  ],
  "maxReviews": 50,
  "reviewsSort": "lowestRanking"
}

For a lowest-rated date window, scan newest activity first, then sort the matching reviews. This avoids dropping the date boundary when the requested order is lowest-rated.

{
  "searchStringsArray": ["auto repair"],
  "locationQueries": ["Austin, Texas, USA"],
  "maxPlacesPerSearch": 3,
  "reviewsStartDate": "2026-09-01",
  "reviewsEndDate": "2026-10-06T23:59:59.999Z",
  "filterStars": [1, 2],
  "reviewsSort": "lowestRanking",
  "maxReviews": 0,
  "maxReviewPages": 20,
  "maxRunSeconds": 180
}

URLs and CIDs:

{
  "startUrls": [{"url": "https://www.google.com/maps?cid=11637992854370324234"}],
  "placeIds": ["0x8644b4332aeb29d1:0xa18274d71f470f0a"],
  "maxReviews": 25,
  "personalData": false
}

Both inputs in this example identify the same place; it is scraped once.

All supplied source lists are combined; place IDs are deduplicated within the run. Search locations form a Cartesian product with terms. Google Maps search URLs and share/short links are resolved through anonymous HTTP redirects. A Place ID must resolve to its own business; a similarly named result is never substituted. A place URL with no resolvable identifier is sent to ERRORS; only search URLs expand into ranked searches.

Python API

Install apify-client==3.2.1, set APIFY_API_KEY, and substitute the published Actor identifier below. This example uses the Console prefill and a USD 0.05 charge ceiling:

import os
from datetime import timedelta
from decimal import Decimal
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_API_KEY"])
run = client.actor("ACTOR_ID_OR_USERNAME/google-maps-review-rows").call(
    run_input={
        "placeIds": ["ChIJ0SnrKjO0RIYRCg9HH9d0gqE"],
        "maxReviews": 50,
        "reviewsSort": "lowestRanking",
    },
    memory_mbytes=256,
    run_timeout=timedelta(seconds=300),
    max_total_charge_usd=Decimal("0.05"),
    logger=None,
)
if run is None:
    raise RuntimeError("Actor run did not return a result")
print(run.status)
for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item)
coverage = client.key_value_store(run.default_key_value_store_id).get_record("COVERAGE")
print(coverage)

Inputs

Schema fields use Console prefill rather than source defaults. Runtime fallback values apply only to omitted/null options. maxReviews:0 and maxReviewPages:0 mean unlimited. Empty source lists, filterStars:[], maxPlacesPerSearch:0 or an inverted date window start no work.

Field	Type	Behavior
startUrls	array	Place, CID, share/short-link or Maps search URLs. All supplied source lists are combined; duplicate place IDs are scraped once.
placeIds	array	Public Place IDs, 0x feature IDs or decimal CIDs. An unknown identifier must resolve to its own place.
searchStringsArray	array	Search every term at each locationQueries entry. Results follow Google's ranked search pages.
locationQueries	array	Free-text locations. An empty list searches each term without an added location.
searchCoordinates	array	Optional [longitude, latitude] camera for term searches. This is a search bias, not a geographic census or boundary.
maxPlacesPerSearch	integer	Runtime fallback 20. Zero starts no work.
maxReviews	integer	Runtime fallback 0 means unlimited. A positive cap produces PARTIAL coverage when it truncates matches.
reviewsSort	string	Runtime fallback newest. Star/date-filtered highest or lowest scans newest then sorts the matched window.
filterStars	array	Integers 1–5. Omitted means all five ratings; an empty array starts no work.
reviewsStartDate	string	Inclusive activity timestamp boundary: ISO date/time or duration such as 30 days. Date-only means midnight UTC; months are 30 days and years 365 days.
reviewsEndDate	string	Inclusive activity timestamp boundary. A date-only value includes midnight UTC, not the whole calendar day. Inverted windows start no work.
reviewsFilterString	string	Case-insensitive substring in full original or translated text. Leading and trailing spaces are significant. Not Google's semantic review search.
reviewsOrigin	string	Filter the origin label returned by Google. Partner origins are source-dependent and not yet platform-proven. Runtime fallback all.
language	string	Requested language, runtime fallback en. Original text and separately returned translations are retained.
countryCode	string	Two-letter search/review country bias, runtime fallback us.
personalData	boolean	Runtime fallback true. False clears reviewer identity/profile fields and reviewUrl; review ID, text and owner reply remain.
maxReviewPages	integer	Runtime fallback 0 means unlimited. A positive page cap marks uncompleted scans PARTIAL.
maxRunSeconds	integer	Runtime fallback 0 disables the budget. The clock begins when the source client opens; includes retries. Exhaustion marks interrupted places PARTIAL.
maxRetriesPerRequest	integer	Runtime fallback 3; includes the first attempt. Blocks, transport failures and malformed answers rotate to fresh sessions.
proxyConfiguration	object	Omitted means Apify datacenter proxy. useApifyProxy:false uses direct egress. Residential is used only when explicitly configured.

Review rows

The dataset contains only recordType:"review" artifacts. Each row carries the place identity/context and these fields:

• Place: placeId, fid, cid, title, address, neighborhood, street, city, postalCode, state, countryCode, categoryName, categories, location, totalScore, reviewsCount, phone, website, description, price, imageUrl, url.
• Review: reviewId, reviewUrl, name, text, textTranslated, publishAt, publishedAtDate, activityTimestampMs, timestampKind, likesCount, reviewerId, reviewerUrl, reviewerPhotoUrl, reviewerNumberOfReviews, reviewerNumberOfPhotos, isLocalGuide, reviewOrigin, stars, responseFromOwnerText, responseFromOwnerTextTranslated, responseFromOwnerPublishedAge, responseFromOwnerDate, responseFromOwnerTimestampUs, originalPublishedAtDate, originalTimestampUs, editedAtDate, editedTimestampUs, reviewImages, reviewImageUrls, reviewContext, reviewDetailedRating, visitedIn, originalLanguage.
• Run context: language, translatedLanguage, searchString, sourceUrl, scrapedAt, scanCoverage, coverageReason.

Optional source fields are null or empty collections when absent. Photos are public review image URLs; no image binary is downloaded. reviewImages preserves each source photo's imageUrl, photoId, caption, aspectRatio (dimensionless source ratio), thumbnailUrl and relative publishedAge when available. The Search feed does not supply dimensions, coordinates or an absolute upload epoch for these photos. Full original and translated review and owner texts are normalized once from source HTML: <br> becomes a newline and HTML entities decode once; nested literal spellings such as < remain <. Text filters match the normalized review content. The separate owner-reply age remains a relative source string. Context and detailed ratings preserve the labels returned in the requested language.

Example (selected fields):

{
  "recordType": "review",
  "placeId": "ChIJ0SnrKjO0RIYRCg9HH9d0gqE",
  "title": "Example business",
  "reviewId": "example-review",
  "stars": 2,
  "text": "Example original review",
  "textTranslated": null,
  "publishedAtDate": "2026-09-19T00:00:00Z",
  "activityTimestampMs": 1789776000000,
  "timestampKind": "activity",
  "reviewImageUrls": [],
  "scanCoverage": "PARTIAL",
  "coverageReason": "review_limit"
}

Timestamp meaning: publishedAtDate is the absolute activity time supplied by this feed, mirrored as epoch milliseconds in activityTimestampMs. For edited reviews it may be the edit time. It is not a separately verified original publication timestamp. Date filters use this activity time. publishAt and responseFromOwnerPublishedAge preserve relative source strings. Rows with owner replies request the anonymous review permalink detail by review ID. responseFromOwnerDate and responseFromOwnerTimestampUs carry the absolute original reply time; originalPublishedAtDate/originalTimestampUs carry the original review time, and editedAtDate/editedTimestampUs carry the activity time only when it differs from the original. These epochs use microseconds; ISO strings retain that precision. These fields remain null on rows without owner replies. No date is inferred from relative text. A blocked or unreadable detail retains the review and relative age, leaves these fields null and records PARTIAL/owner_reply_enrichment_failed in COVERAGE when the scan was otherwise complete, plus the reason in ERRORS. An existing scan-limit or failure reason is preserved.

Fractional date boundaries require a clock with seconds, such as 2026-09-19T00:00:00.5Z. Fractions after only hours or minutes are rejected rather than treated as fractional seconds. Lower boundaries remain inclusive across pages; completion by date cutoff requires a timestamp strictly below the boundary.

Coverage and failures

Read the final COVERAGE record via the Actor output link. Each entry includes place ID, title, source, published review count, requested date boundaries, status/reason, unique reviews read, matching reviews, rows delivered, pages read and oldest activity time. Source failures before the count/window is known may have fewer context fields.

• COMPLETE means the observed newest stream crossed strictly below the inclusive lower date boundary, or a terminal page yielded at least the published count. It applies to the requested window/filters at that read time, not to deleted, hidden or later reviews.
• PARTIAL means a review/page cap, time budget, charge limit, source error, repeated pagination, ordering failure, storage failure, owner-reply enrichment failure or terminal count shortfall prevented that proof.
• scanCoverage on dataset rows describes the scan before delivery. The final KVS status is authoritative because a later billing/storage stop can truncate delivery.
• Google’s count and feed may differ or change during a scan. A count shortfall remains PARTIAL. Valid zero-review or zero-match answers are successful runs.
• Google relevant order is retained for mostRelevant. It must page to the terminal count for a complete filtered result; date cutoff is used only on newest streams.
• Star/date-filtered highest/lowest scans the full available window before choosing a capped subset, using a temporary disk-backed store. This can read many more reviews than it delivers.

Read ERRORS for failed inputs and interrupted source reads. Errors are not billable rows. Invalid input is recorded in ERRORS and exits with guidance before any source attempt. A run in which every attempted source fails and no review is delivered is marked FAILED. Partial runs with delivered rows can succeed; inspect COVERAGE instead of treating the run status as a completeness guarantee.

Billing and egress

Billing uses Apify pay-per-event dataset delivery: one delivered review row is the unit. A charge-limit preflight starts no source work, and a limit reached during delivery stops the run before the next place. Errors and COVERAGE entries are key-value records, not review events. Pricing is set by the publisher; check the Actor's current pricing before running.

Apify datacenter proxy is the runtime fallback. Explicit direct egress is supported. Blocks, temporary HTTP failures and unreadable responses retry with fresh cookie-free sessions, up to the configured attempt count. Residential proxy groups are used only if explicitly configured. No browser, account login or Google API key is required.

Current limits

Search follows Google's ranked pages and a camera bias; it is not a geographic census. Dense-area tiling, polygons, category/place filters, website contact crawling, place galleries, review updates/change feeds and saved-dataset sorting are outside this review-only Actor.

The six origin selectors filter labels in the returned feed; this Actor has no proof of fetching every hotel's partner-review feed. Original/edit and absolute reply dates are available only on successfully enriched owner-reply rows. Photo dimensions, coordinates and absolute upload times are unavailable from the Search review payload. A measured 30-day service success rate is not yet available.