Tripadvisor Reviews Scraper: Ratings, Replies, Translations

Give it Tripadvisor hotel, restaurant or attraction pages (or place names) and it returns one JSON row per review: rating, title, original-language text with an optional translation, date, sub-ratings, trip type, the owner reply with its date, and the place's overall rating and review count. $4.50 per 1,000 reviews on the free plan, less on paid plans; you pay only for rows you get.

Why this actor

• No depth limit. A place with 15,000 reviews is read to the end; stars, language, trip type, month and keyword filters are applied by Tripadvisor itself.
• Only new reviews. onlyNewReviews delivers just what earlier runs did not; repeats are not charged.
• Real data or a clear error. A free coverage row per place (stated vs delivered) and a free error row with a code, never a silent empty run.
• Easy switch. The input names of the most used Tripadvisor reviews actor are accepted as they are.

Built by Arcadium Data: public data, dated and agent-ready, never personal data.

Use with AI agents

Returns Tripadvisor reviews (rating, text, date, owner reply, sub-ratings) for hotel, restaurant or attraction URLs or place names. Agents can call it through the Apify MCP server (https://mcp.apify.com, open source as apify/actors-mcp-server) as arcadium/tripadvisor-reviews: the input schema is read from the actor, the result is a dataset of JSON rows.

Input example (minimal):

{
  "startUrls": [
    {
      "url": "https://www.tripadvisor.com/Hotel_Review-g60763-d93450-Reviews-Hyatt_Grand_Central_New_York-New_York_City_New_York.html"
    }
  ],
  "maxReviewsPerPlace": 20
}

What comes back: one review row per review (charged) and, for an input that could not be read, one free error row. Filter on the rowType field. An input that fails is reported, not hidden (includeErrorRows, on by default): the error row carries rowType: "error", a code (invalid_input, not_found, blocked, unavailable, filter_unsupported or no_match), a message and the input it is about. Retry on blocked and unavailable later; do not retry the other codes with the same input.

Cost of the call above: 20 reviews at $4.50 per 1,000 plus the $0.001 start fee is about $0.09 on the free plan; paid plans pay less. Free rows are never charged.

What you get

Field	What it holds
reviewId	Tripadvisor___ plus the platform's own review ID. Stable between runs, so you can dedupe weekly exports.
date	Date the review was written, ISO format.
title, text	What the reviewer wrote, in the language it was written in.
extras.translatedTitle, extras.translatedText, extras.translatedLanguage	Machine translation of a review written in another language, next to the original. Only with translateTo.
extras.originalLanguage	Language the review was written in.
product.rating, product.reviewCount	rating is the average of all review ratings of the place (computed from its star histogram, not Tripadvisor's rounded bubble rating); reviewCount is the total number of reviews. Repeated on every row. Turn off with includePlaceSummary.
ratingOverall	Overall bubble rating, 1 to 5.
ratings	Sub-ratings the reviewer filled in (value, rooms, location, cleanliness, service, sleep quality). Skipped ones are left out.
vendorResponse, extras.ownerResponseDate	The owner or manager reply and the date it was posted, when there is one.
helpfulVotes, language, reviewerVerified	Helpful votes, review language, and whether the review is marked as verified.
productName, productUrl, reviewUrl	The place and where the review lives.
extras.tripType	Business, couples, family, friends or solo.
extras.visitDate	Month of the visit (yyyy-mm).
extras.photoUrls	Photos attached to the review.
extras.totalCount	How many reviews match your filters at that place.
scrapedAt	When the row was collected.

Output example:

{
  "reviewId": "Tripadvisor___123456789",
  "platform": "tripadvisor",
  "productName": "Example Grand Hotel",
  "date": "2026-09-27",
  "title": "Great stay",
  "text": "Quiet room, friendly staff.",
  "ratingOverall": 5,
  "ratings": { "value": 5, "cleanliness": 5 },
  "language": "en",
  "vendorResponse": "Dear [name], thank you for staying with us.",
  "extras": { "tripType": "COUPLES", "visitDate": "2026-08", "placeType": "ACCOMMODATION" }
}

Input

{
  "startUrls": [{ "url": "https://www.tripadvisor.com/Hotel_Review-g60763-d93450-Reviews-Hyatt_Grand_Central_New_York-New_York_City_New_York.html" }],
  "maxReviewsPerPlace": 100
}

Field	What it does
startUrls	Hotel, restaurant or attraction pages on any Tripadvisor domain.
searchQueries	Place names. The best match for each name is scraped. Use URLs when you need an exact place.
maxReviewsPerPlace	Cap per place (default 100, up to 50,000).
sort	MOST_RECENT (default), DETAILED or INSIGHTFUL.
ratings	Only these star ratings, for example ["1", "2"].
languages	Only these review languages, for example ["de"]. Several languages are read as separate passes and merged newest first.
tripTypes, visitMonths	Only these traveler types or months of the visit.
keyword	Only reviews that mention it. Tripadvisor caps keyword results at 1,000.
translateTo	Language code such as en. Reviews in other languages also get a machine translation (Tripadvisor translates reviews from roughly the last two years only).
includePlaceSummary	Adds the place rating and review count to every row (default on, five small extra requests per place).
dateFrom	Only reviews from this date on. Stops early with the newest first sort.
onlyNewReviews, stateKey	Skip reviews that earlier runs with the same key already delivered. They are not charged.
includeErrorRows	One free row per place that could not be read, with the reason.

Coverage and edited reviews

A free coverage row per place says how many reviews the source reports against how many you received, and why reading stopped (your cap, the source's cap, or the end). With onlyNewReviews, a review that was edited after it was delivered comes back once more with extras.changeType set to edited; that row is charged like a new review.

Migrating from maxcopell

The input of the other popular Tripadvisor reviews actor works as it is. These field names are accepted and mapped:

Their field	Used as
startUrls	startUrls
maxItemsPerQuery	maxReviewsPerPlace
lastReviewDate (a date, a full ISO datetime, or 3 days, 3 Days ago, -2 weeks, 12 hours)	dateFrom
reviewRatings (ALL_REVIEW_RATINGS or 1 to 5)	ratings
reviewsLanguages (ALL_REVIEW_LANGUAGES or a code)	languages
disableMachineTranslations	false adds an English translation next to the original, true returns originals only
scrapeReviewerInfo, sourceDatasetId	Ignored. Reviewer names and profiles are never returned.

Example, pasted unchanged:

{
  "startUrls": [{ "url": "https://www.tripadvisor.com/Hotel_Review-g60763-d93450-Reviews-Hyatt_Grand_Central_New_York-New_York_City_New_York.html" }],
  "maxItemsPerQuery": 200,
  "lastReviewDate": "2026-09-01",
  "reviewRatings": ["ALL_REVIEW_RATINGS"],
  "reviewsLanguages": ["ALL_REVIEW_LANGUAGES"],
  "disableMachineTranslations": false
}

Output field names are our own (see the table above), so check any downstream mapping once.

Use it from an AI assistant

Paste this into your assistant to run the actor for you:

Use the Apify actor "tripadvisor-reviews" to collect reviews.
Input: {"startUrls":[{"url":"<Tripadvisor hotel, restaurant or attraction URL>"}],"maxReviewsPerPlace":100}
Each dataset row is one review. Key fields: reviewId, date, title, text, ratingOverall, extras.tripType.
Run it, then summarize the main praise and complaints. Mention the review count.

Use it from code

Python:

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("arcadium/tripadvisor-reviews").call(run_input={
    "startUrls": [{"url": "https://www.tripadvisor.com/Hotel_Review-g60763-d93450-Reviews-Hyatt_Grand_Central_New_York-New_York_City_New_York.html"}],
    "maxReviewsPerPlace": 100,
    "ratings": ["1", "2"],
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["date"], item["ratingOverall"], item["title"])

JavaScript:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('arcadium/tripadvisor-reviews').call({
  searchQueries: ['hyatt grand central new york'],
  maxReviewsPerPlace: 100,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.length, 'reviews');

Pricing

You pay per review returned. Failed or empty lookups are not charged.

You pay per review delivered, at the price of your Apify plan. Error rows, coverage rows and repeats in only-new mode are free.

Apify plan	Per 1,000 reviews
Free	$4.50
Starter (Bronze)	$1.75
Scale (Silver)	$1.35
Business (Gold)	$0.80
Platinum	$0.45
Diamond	$0.27

Plus a one-off $0.001 per GB of memory when a run starts.

Set a maximum cost per run in the Apify run options and the actor stops before it goes over.

Related actors and data

• Hotel Reviews Scraper: Tripadvisor, Booking.com, Google
• Booking.com Reviews Scraper: Pros, Cons, Reviewer Country
• Google Hotel Reviews Scraper: Topics, Replies, Partner Sites
• Review Monitor: Negative Review Alerts and Rating Drift
• Hotel demand peaks report: the next 90 nights of hotel rates by city, with spike nights flagged.
• Arcadium Data MCP Server: AI agents can ask the same questions as tools (search_reviews, hotel_prices, city_rate_curve, review_snapshot).

Use cases

• Competitor tracking: export a rival hotel's newest reviews every week and watch what changes.
• Reputation monitoring: pull only the 1 and 2 star reviews of your own places, in every language.
• Sentiment analysis: feed review text, trip type and sub-ratings into your own model or spreadsheet.
• Market research: compare restaurants or attractions in one city side by side.

Limitations

• Public reviews only. Nothing behind a login.
• title and text are always the original. A translation exists only when you set translateTo, and it is Tripadvisor's own machine translation. Tripadvisor does not translate reviews older than about two years, so those rows have no extras.translatedText. If translation stops working during a run, the remaining rows are delivered without it.
• A keyword search returns at most 1,000 reviews, a limit set by Tripadvisor.
• Several languages in one run are read as separate passes and merged newest first, so dateFrom and onlyNewReviews keep their early stop. With another sort the passes are delivered one after another.
• Name search picks the best match only. For hotels with common names, use the URL.
• Sub-ratings exist only where the reviewer filled them in, and the set differs between hotels, restaurants and attractions.
• Tripadvisor changes its site from time to time. The actor adapts on its own in most cases; if a run fails for several places at once, try again later or open an issue.

Privacy

Reviewer names, profile photos, profile links, hometowns, emails and phone numbers are never collected. Names in greetings and sign-offs (for example 'Dear John' or 'Regards, Anna') are replaced with [name]; other names in the text are not detected. The reviewer's own display name and handle are also masked wherever they appear in the text. You get the review and the context around it, not the person.

FAQ

How fresh is the data? It is read when you run the actor, so it matches what the site shows at that moment.

Can I get only new reviews since my last run? Yes. Turn on onlyNewReviews and keep the same stateKey between runs. Reviews you already received are skipped and not charged. The state lives in a key-value store in your Apify account. Pick a new stateKey to start over.

Does the actor store anything in my account? Yes, one small named key-value store, tripadvisor-reviews-state. It caches the identifiers of Tripadvisor's own review queries so the next run starts faster, and it holds the onlyNewReviews state. It contains no review text.

A place returned no reviews. Why? Check that the URL opens a hotel, restaurant or attraction page in your browser. Places without reviews return a free error row.

Something looks wrong in the output. Who do I tell? Open an issue on the actor page with the run link. We read them all.