Facebook Reviews Scraper

Paste public Facebook Page URLs and get their Recommendations back, one row per review: who wrote it, whether they recommend the Page, the review text, the exact date, the reaction and comment counts, the recommendation tags Facebook attaches, and any photos the reviewer posted with it.

You never sign in and you never paste a cookie. See About logging in for what the run actually does.

• Many Pages in one run. Each review is its own flat row, with the Page it belongs to on it.
• Positive and negative are read from the recommendation marker itself rather than guessed from the wording, so the flag stays right on a Page that renders in another language.
• Every row also carries the Page-level summary: what percentage of readers recommend it and how many reviews there are in total.
• Date and positive/negative filters are free. Anything filtered out is dropped before it's charged.
• Empty input returns one labelled sample row, free.

Price

$0.35 per 1,000 reviews. Plus $0.0005 per run as the platform start fee, billed per gigabyte of run memory, so exactly that on the default 1 GB.

No volume tiers, no minimum spend, no subscription.

Reviews	Total
100	$0.0355
1,000	$0.3505
10,000	$3.5005
100,000	$35.0005

What gets charged

• One review-scraped event per review row written to the dataset. Nothing else is metered per row.
• Free: the sample row an empty run returns, and every diagnostic row. They all carry "charged": false.
• Reviews dropped by the date filter or the positive/negative filter are never charged.
• A Page that has switched Recommendations off returns one free diagnostic row, not an invoice.
• A run that finds nothing costs the start fee and nothing else.
• Rows never leave the dataset without a charge and are never charged without a row. The billed event is a named one, so nothing is quietly attached to apify-default-dataset-item.

Input

{
  "startUrls": [
    "https://www.facebook.com/Dominos",
    "https://www.facebook.com/KFC"
  ],
  "resultsLimit": 10,
  "recommendationType": "all"
}

Field	What it does
startUrls	Public Facebook Page URLs, one per line. A bare handle (Dominos), a Reviews-tab URL, a /pg/ URL, a /p/ URL and a profile.php?id= URL all work. Up to 50 Pages per run.
resultsLimit	How many recent reviews to take from each Page, newest first. Default 10, maximum 200. Keep it low while testing, since you pay per review.
recommendationType	all (default), positive to keep only "recommends", or negative to keep only "doesn't recommend". Filtering is free.
onlyReviewsNewerThan	Drop reviews older than this. Accepts 2026-08-01, a full ISO timestamp, an epoch, or a relative window like 7 days. Free.
onlyReviewsOlderThan	Drop reviews newer than this. Same formats. Also free.
proxyUrls	Leave empty. Only for callers who want traffic to leave through proxy servers they already pay for, as http://user:pass@host:port.
sessionCookies	Leave empty unless you're running heavily. See About logging in.

Run it with empty input and you get one labelled sample row, free.

Output

One row per review. This is a real row from a real run:

{
  "ok": true,
  "charged": true,
  "recordType": "review",
  "inputUrl": "https://www.facebook.com/Dominos",
  "pageName": "Domino's Pizza",
  "pageId": "100064795448235",
  "pageUrl": "https://www.facebook.com/Dominos/",
  "pageIsVerified": true,
  "pageProfilePicture": "https://lookaside.fbsbx.com/lookaside/crawler/media/?media_id=100064795448235",
  "pageRecommendPercent": 30,
  "pageReviewsCount": 12661,
  "pageRatingSummary": "30% recommend (12,661 Reviews)",
  "reviewId": "10164211035477496",
  "reviewUrl": "https://www.facebook.com/randyfgarciaLV/posts/pfbid05YabRqiEtRu7e4Lqm1uDpYtjWhzTAEJj4DCemp6BywqxnkyjDAh9aFXCKMKE3Aujl",
  "isRecommended": true,
  "recommendation": "positive",
  "text": "Nice place. I would come back for more of that sweet pizza anytime.",
  "textLength": 67,
  "tags": ["Fast delivery", "Child-friendly"],
  "date": "2026-08-09T09:19:05.000Z",
  "timestamp": 1786267145,
  "reviewerName": "Randy F Garcia",
  "reviewerId": "562052495",
  "reviewerUrl": "https://www.facebook.com/randyfgarciaLV",
  "reviewerProfilePicture": "https://lookaside.fbsbx.com/lookaside/crawler/media/?media_id=562052495",
  "likesCount": 0,
  "commentsCount": 1,
  "sharesCount": 0,
  "photos": [],
  "photoCount": 0,
  "scrapedAt": "2026-08-16T13:33:50.062Z"
}

Field notes

• isRecommended is true when the reviewer recommends the Page and false when they don't.
• recommendation is the same fact as a string, "positive" or "negative", for spreadsheets and pivot tables.
• pageRecommendPercent is the Page-level figure Facebook shows in place of an average rating: the share of reviewers who recommend it. Null on a Page Facebook labels "Not yet rated".
• pageReviewsCount is how many reviews the Page has in total, which is almost always far more than one run fetches.
• tags are the chips Facebook offers a reviewer alongside the text, like "Fast delivery" or "Poor presentation". Empty on most reviews, because most reviewers pick none.
• photos are direct links to images and video stills attached to the review. Empty on the majority. Roughly one review in sixteen carries media.
• likesCount counts reactions on the review itself, not on the Page. Most reviews sit at zero.
• reviewerId is an opaque pfbid… identifier for most profiles and a numeric one for older accounts. Both come back as strings, exactly as the platform returns them.
• reviewUrl points at the reviewer's own profile, because a review is a post there. It's the canonical permalink for that review.
• date is normalised to ISO 8601 UTC. timestamp is the same instant in epoch seconds.

Real rows carry "charged": true. Sample rows carry "_sample": true, diagnostic rows carry "_diagnostic": true and an errorCode. Neither is billed.

How it works

It reads the Page's public Recommendations tab the way an indexing crawler does and parses the structured data the page already ships. Nothing is rendered and no headless browser starts.

Reviews past the first render are followed through the tab's own paging cursor, three at a time on the same connection, so depth costs time rather than extra page loads. Requests leave through a rotating pool of addresses, so a per-address rate limit is answered by moving instead of waiting.

Positive and negative come from the structured marker Facebook attaches to each recommendation rather than from the English sentence.

About logging in

You never supply a Facebook account and you never paste a cookie. Reads go out logged out, and that's what answers a normal run.

When Facebook answers with a login interstitial on address after address, the run continues with a Facebook session that ships inside the Actor. That session is shared by everyone running this Actor, and Facebook rate-limits per account rather than per caller, so under heavy load it's a small shared budget.

If you'd rather that fallback used your own account, paste your cookie into sessionCookies as c_user=<value>; xs=<value>, one line per account. Treat those values like a password. Anyone holding them can act as that account, and Facebook may sign the session out or restrict an account it decides is automated.

Recommendations, not stars

Facebook retired 1-5 star ratings for Pages in 2018. What a Page has now is Recommendations: each reader either recommends it or doesn't, and writes free text underneath. There's no per-review star value stored anywhere, so this Actor doesn't publish one. A stars: 4 column here would be a number somebody made up.

What you get instead:

• isRecommended, a clean boolean per review, plus recommendation as "positive" or "negative".
• pageRecommendPercent, the Page-level share of reviewers who recommend it. This is the figure Facebook itself shows in place of an average.
• pageReviewsCount, the total number of reviews the Page has.

If you need a star-shaped number for an existing dashboard, pageRecommendPercent / 20 gives you a 0-5 scale from a real figure rather than an invented one.

What people use it for

• Reputation monitoring for a brand with many location Pages. Schedule the same list daily, filter on recommendation == "negative", route the new ones to whoever answers complaints.
• Competitive research. Pull recommendations for a set of Pages in one run and compare pageRecommendPercent and the language people actually use.
• Feeding a review widget or testimonials page with real, attributed, dated quotes instead of hand-copied ones.
• Support triage. Negative reviews arrive with a permalink and the reviewer's name, so a reply can be written straight from the row.
• Product analytics. tags plus the free text is a labelled dataset of what customers praise and complain about.
• Due diligence on a supplier, franchise or venue. Total review count, percentage recommending, and the twenty most recent opinions in one run.

Reading the output

• Real rows carry "charged": true and "recordType": "review". One billed event each.
• The sample row carries "_sample": true and "charged": false. There's exactly one, only when the input had no Page URLs.
• Diagnostic rows carry "_diagnostic": true, "charged": false and an errorCode: REVIEWS_DISABLED when the Page has no public Recommendations tab, NO_REVIEWS when the tab exists but is empty or everything fell outside your filters, NOT_FOUND when no Page lives at that URL, BAD_INPUT when the URL isn't a Facebook Page, BLOCKED when Facebook answered with a login interstitial from every address tried, NETWORK when it couldn't be reached, TIME_BUDGET when time ran out. Each carries a plain-English details and the inputUrl it belongs to.

Filter on charged == true and you have exactly the rows you paid for.

Limits

• Reviews are optional on Facebook. A Page owner can switch Recommendations off, and a lot of large brand Pages have. Those Pages return a free REVIEWS_DISABLED row rather than reviews. It's a property of the Page, not a failure of the run, and nothing is charged for it.
• There's no 1-5 star rating to return.
• Only what a logged-out reader can see is returned. Reviews restricted by the reviewer's own privacy settings are invisible without an account.
• Reviews come newest first, and the depth available per Page is finite. Asking for 200 on a Page that has published forty publicly returns forty.
• Reaction, comment and share counts are a snapshot at read time.
• Most reviews carry no photos and no tags, because most reviewers add neither. Those fields are an empty array rather than null.
• Photo links are Facebook CDN URLs. The stable lookaside form keeps working. The signed scontent form expires after some hours, so download anything you need to keep rather than storing the link.
• A reviewer profile URL is only published when Facebook publishes one. Where it doesn't, the link is derived from the review permalink, which points at the same profile.
• Hard ceilings: 50 Pages per run and 200 reviews per Page.
• Facebook rate-limits this surface per address. The run answers by moving to another address, but a very large single run can still take a few minutes.

Questions

Do I need a Facebook account, a cookie, or a token?

No. You supply nothing. Everything comes from what Facebook serves a logged-out reader, which is also why reviews a reviewer has made private don't appear.

Why does one of my Pages return no reviews?

Almost always because that Page has Recommendations switched off. It's a setting the Page owner controls, and many big brand Pages disable it. You get one free REVIEWS_DISABLED row and no charge for that Page.

Where is the star rating?

It doesn't exist any more. Facebook replaced 1-5 stars for Pages in 2018 with recommend / doesn't recommend, so every row carries isRecommended instead, and pageRecommendPercent takes the place of an average.

Can I get only the bad reviews?

Yes. Set recommendationType to negative. The filter runs before billing, so the positive reviews that get dropped cost nothing.

Will the run fail if something goes wrong?

No. A blocked, empty or missing Page produces an uncharged diagnostic row and the run still finishes as succeeded. A failed run would still bill the start fee, which would mean paying to be told something went wrong.

Can I run it on a schedule?

Yes. Nothing is held between runs. Use reviewId to work out which rows are new since last time.