Facebook Comments Scraper (Reaction Breakdown & Engagement Data)

Scrape comments and replies from public Facebook posts and videos, then get a per-comment reaction-type breakdown (LIKE/LOVE/HAHA/WOW/SAD/ANGRY/CARE) and locally-computed engagement analytics — engagement rate per comment, a top-commenter leaderboard, and comment velocity since the post was published. Built for social listening, engagement research, and Facebook comment-data export/API workflows that need more than a raw like count.

Features

• Bulk URLs — process multiple Facebook post/video permalinks in one run.
• Comments + first-page replies — top-level comments and their first page of nested replies, tagged type/isChild/parentId so parent and reply rows are both easy to filter and easy to reassemble.
• Reaction-type breakdown — raw reaction ids + counts per comment/reply, plus a best-effort id → name map. Unknown ids stay name: null rather than a guessed label.
• Engagement analytics — engagementScore / engagementRatePercent per row, a per-post summary (total comments/replies, best-effort post date, comment velocity per hour), and a run-level top-commenter leaderboard ranked by total engagement.
• Absolute or relative date filter — keep only comments newer than a fixed date or a relative window (2 weeks, 3 months, …).
• Honest truncation flag — every row carries commentsTruncated so it's always clear when a thread had more comments/replies than were fetched.
• Real-time output — rows land in the Output tab as soon as each post is processed, no waiting for the whole run to finish.

Use cases

• Social-listening dashboards that need reaction sentiment, not just a like count.
• Brand/campaign post monitoring — who is most engaged, and how fast a post is picking up comments.
• Research and moderation datasets needing per-comment reaction breakdowns.
• Feeding a downstream sentiment/AI pipeline with real comment text plus its engagement context.

Input

Field	Type	Default	Description
startUrls	array (required)	—	Facebook post/video permalink URLs, one per line
commentSettings.maxComments	integer	10	Max top-level comments per post
commentSettings.maxRepliesPerComment	integer	2	Max replies kept per top-level comment
commentSettings.commentsSortOrder	enum	all	most_relevant, newest, or all (original order)
onlyCommentsNewerThan	datepicker (absolute or relative)	—	Keep only comments/replies newer than this date/window
reactionEngagementSettings.includeReactionBreakdown	boolean	true	Include the per-comment reaction-type breakdown
reactionEngagementSettings.computeEngagementAnalytics	boolean	true	Include engagement score/rate, post summary, and the leaderboard
reactionEngagementSettings.topCommentersLimit	integer	10	How many ranked rows the leaderboard keeps
proxyConfiguration	object	informational only	The Actor always routes through Apify residential proxy regardless of this field

Output

Each comment/reply is one dataset row (type: "comment" or "reply"):

• facebookUrl, commentUrl, id, feedbackId, date, text
• profilePicture, profileId, profileName, profileUrl (when available)
• likesCount, commentsCount, reply (nested replies, kept for backward compatibility), threadingDepth, facebookId, inputUrl
• type (comment/reply), isChild, parentId, commentsTruncated
• reactions.raw — [{id, name, count}, ...] (raw reaction ids always present; name is a best-effort LIKE/LOVE/HAHA/WOW/SAD/ANGRY/CARE map, null for unmapped ids)
• reactions.byType — {LIKE: n, LOVE: n, ...} summed from the mapped raw entries
• reactions.totalFromBreakdown, reactions.reactorsCountReduced — cross-check totals
• engagementScore, engagementRatePercent — this row's share of the post's total engagement

Once per post, a type: "postEngagementSummary" row is added with totalComments, totalReplies, totalEngagement, postDate (best-effort, null when it can't be reliably determined), postDateCoverage, and commentVelocityPerHour.

At the end of the run, up to topCommentersLimit type: "commenterLeaderboard" rows are added, ranked by totalEngagement across the whole run.

An error field (with full_url/url_key) is added instead when a post could not be processed.

How it works

The Actor loads each post's own permalink page (through Apify's residential proxy) and parses comments — including their first page of replies — directly out of that page's server-rendered data. This is the same single page fetch needed to identify the post, so no extra network round-trip is spent per comment or per reply page. Because this reads only the first page of replies Facebook renders server-side, very long reply threads are naturally partial — this is disclosed honestly via commentsTruncated rather than hidden.

Reaction-type mapping

Facebook represents each reaction type as an opaque numeric id rather than a readable name at the comment level. This Actor ships the raw ids and counts always, plus a best-effort mapping to LIKE/LOVE/HAHA/WOW/SAD/ANGRY/CARE built from publicly known, long-stable Facebook reaction ids. An id this Actor doesn't recognize is reported with name: null — never a guessed name.

Cautions

• Data from publicly available posts only.
• Comply with Facebook's Terms of Service and applicable privacy laws.
• Use responsibly.