YouTube Video Comment & Reply Thread Scraper

Scrape YouTube comments and reply threads as analysis-ready rows. Give it video links, channel video tabs, playlists, or search queries, and get one row per comment or reply with author identity, engagement signals, thread position, and video context. Works on public videos without user cookies, login, or an API key.

What this actor does

• Four source types: individual videos (watch, Shorts, live, embed, youtu.be, bare ID), channel video tabs, playlists, and video search results
• Flat rows by default: every comment and reply is its own row with parentCommentId, rootCommentId, replyLevel, and threadPosition, so CSV, Excel, SQL, and sentiment pipelines can use it directly
• Full reply threads: fetch every page of a thread, cap per comment, or skip replies entirely; maxRepliesPerComment: 0 means all
• Real filters: Top/Newest sort, date cutoff, multiple keywords (ANY match), minimum likes, and only creator-hearted comments
• Honest engagement data: rounded like counts with the original text, advertised reply counts, pinned/hearted/verified flags, and approximate dates with a precision and range
• Video context on every row: title, channel, views, likes, comment count, duration, publish date, and description
• Empty fields are omitted — records only contain values that are actually populated.

Output per comment or reply

Each dataset row is one comment (recordType: "comment", replyLevel: 0) or one reply (recordType: "reply", replyLevel: 1).

Field	Type	Description
recordType	string	comment, reply, or status
commentId	string	YouTube comment identifier
parentCommentId	string	Root comment ID on replies; omitted on top-level comments
rootCommentId	string	Thread root ID (own ID on top-level comments)
replyLevel	integer	0 for comments, 1 for direct replies
threadPosition	integer	1-based order in the emitted feed or reply thread
isReply	boolean	Backward-compatible alias for replyLevel > 0
replyDepth	integer	Backward-compatible alias for replyLevel
text	string	Comment text, with unicode and newlines preserved
authorName	string	Display name
authorHandle	string	@handle when YouTube exposes it
authorChannelId	string	UC... channel ID
authorChannelUrl	string	https://www.youtube.com/@handle or /channel/UC...
authorAvatarUrl	string	Avatar image URL
authorIsVerified	boolean	Author channel is verified
authorIsCreator	boolean	Author is the video's channel owner
authorIsArtist	boolean	Author is marked as a YouTube Music artist
likeCount	integer	Rounded like count (69K becomes 69000)
likeCountText	string	Original formatted count (69K)
replyCount	integer	YouTube's advertised reply count; may exceed visible replies
replyCountText	string	Original reply-count text when YouTube exposes it (963 replies)
repliesFetched	integer	Reply rows actually attached/emitted for this comment (present when replies were fetched)
isPinned	boolean	Comment is pinned
pinnedByText	string	Pinned-by tooltip (Pinned by @channel)
isHearted	boolean	Creator hearted the comment
heartedBy	string	Creator handle from the heart tooltip
heartedByText	string	Normalized creator label for the heart
creatorReplied	boolean	Whether any emitted reply is authored by the video creator (present when replies were fetched)
creatorReplyCount	integer	Number of emitted replies authored by the creator (present when replies were fetched)
isEdited	boolean	Comment was edited, derived from the (edited) label
publishedTimeText	string	Relative publish text exactly as YouTube shows it
publishedAtApprox	string	Midpoint of the inferred date range (ISO-8601)
publishedAtFrom	string	Earliest possible publish time
publishedAtTo	string	Latest possible publish time
publishedAtPrecision	string	second, minute, hour, day, week, month, or year
ageDaysMin	number	Lower bound of comment age in days
ageDaysMax	number	Upper bound of comment age in days
ageDaysApprox	number	Midpoint of the age range
commentUrl	string	Direct link to the comment (...&lc=COMMENT_ID)
videoId	string	Video ID
videoUrl	string	Canonical watch URL
videoTitle	string	Video title
videoChannelId	string	Video channel ID
videoChannelHandle	string	Video channel path (/@handle)
videoChannelName	string	Video channel display name
videoViewCount	integer	View count
videoLikeCount	integer	Video like count (YouTube may round it)
videoCommentCount	integer	Approximate total comments shown by YouTube
videoPublishDate	string	Raw publish date text (Oct 24, 2009)
videoPublishedAt	string	Parsed publish date (2009-10-24)
videoDurationSeconds	integer	Video length in seconds
videoDescription	string	Video description
sourceInput	string	The original URL, handle, or query that produced this video
sourceScope	string	video, channel, playlist, or search
sourceVideoId	string	Video selected from the source
scrapedAt	string	ISO-8601 UTC timestamp of the run
replies	array	Embedded reply rows, only with outputMode: tree or both

Status rows

When emitStatusRows is enabled, blocked/empty/unavailable sources and partial or budget-limited runs add one row with recordType: "status", status, inputUrl, optional video fields, error, and rowsEmitted when known. Status rows are never counted toward maxTotalRows and do not consume the data-row budget. Status values: no_comments, comments_off, unavailable, private, age_restricted, members_only, region_blocked, live_chat_only, blocked, stale_token, transient_empty, parse_empty, partial, budget_reached, input_invalid. Successful comment runs do not add an ok status row.

Dates and counts are approximate

YouTube's public comment payload does not expose exact timestamps. Every date field is derived from the relative label (6 years ago) and comes with a precision plus earliest/latest bounds, so no exact timestamp is invented. Like counts above 999 are rounded by YouTube itself; likeCountText preserves the original value. replyCount is YouTube's advertised count and can be higher than the number of replies the public thread returns.

Input

Field	Type	Default	Description
videoUrls	array	[]	Video URLs or bare IDs (watch, youtu.be, Shorts, embed, live, v/)
channelUrls	array	[]	Channel video tabs: /@handle, /channel/UC..., /c/name, /user/name, @handle, or UC...
playlistUrls	array	[]	Playlist URLs or bare playlist IDs
searchQueries	array	[]	Search terms or YouTube search-result URLs; video results only
maxVideosPerSource	integer	20	Video discovery cap per channel/playlist/search; 0 = all
maxComments	integer	100	Top-level comments to examine per video; 0 = all. Filtered-out comments still count toward this scan cap
includeReplies	boolean	true	Fetch reply threads
maxRepliesPerComment	integer	50	Replies per thread; 0 = all until the token is exhausted
maxReplyDepth	integer	1	0 = top-level only, 1 = direct replies
maxTotalRows	integer	5000	Global comment + reply budget; 0 = unlimited
sortBy	string	top	top or newest
commentsSince	string	–	ISO date/datetime or relative duration (7d, 24h); forces newest
keywords	array	[]	Case-insensitive ANY-match filter on comment/reply text
minLikes	integer	0	Drop rows below this rounded like count
onlyCreatorHearted	boolean	false	Keep only creator-hearted rows
outputMode	string	flat	flat, tree, or both
emitStatusRows	boolean	true	Emit recordType: "status" rows for blocked/empty/invalid sources

At least one of videoUrls, channelUrls, playlistUrls, or searchQueries is required.

Example: one video with replies

{
  "videoUrls": ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"],
  "maxComments": 50,
  "includeReplies": true,
  "maxRepliesPerComment": 20,
  "sortBy": "top",
  "maxTotalRows": 500
}

Example: recent comments from a channel

{
  "channelUrls": ["https://www.youtube.com/@RickAstleyYT"],
  "maxVideosPerSource": 10,
  "sortBy": "newest",
  "commentsSince": "7d",
  "maxComments": 100,
  "minLikes": 5,
  "maxTotalRows": 1000
}

Example: search with creator-hearted comments

{
  "searchQueries": ["product launch review"],
  "maxVideosPerSource": 5,
  "onlyCreatorHearted": true,
  "maxRepliesPerComment": 0,
  "maxTotalRows": 2000
}

Example: backward-compatible tree output

{
  "videoUrls": ["https://www.youtube.com/watch?v=jNQXAC9IVRw"],
  "maxComments": 20,
  "includeReplies": true,
  "maxRepliesPerComment": 50,
  "outputMode": "tree"
}

Use cases

• Sentiment and topic analysis — export flat comment and reply rows directly into a notebook or BI tool
• Creator engagement audits — measure which comments the creator pinned, hearted, or replied to
• Audience research — mine questions and recurring themes across a channel's recent uploads or a playlist
• Moderation datasets — collect long reply threads with parent/root IDs for context-aware classifiers
• Campaign monitoring — pull the last few days of comments from several videos with a date cutoff and minimum-like floor
• Competitive research — compare comment volume and engagement on the same topic across search results

FAQ

Do I need a YouTube API key or cookies?

No. The actor reads publicly available comment data without login, user cookies, or an API key. It sends YouTube's standard consent cookie so the EU consent page does not replace the video page.

Why are some comments missing?

YouTube's "Top comments" feed exposes a relevance-ordered subset (roughly 800–1,200 comments on very large videos). Use sortBy: "newest" for complete date windows and deeper coverage. The actor also continues with the newest feed when the top feed is exhausted, and maxComments, maxTotalRows, and your filters can all stop a run earlier than the source total.

Are replies complete?

The actor follows every reply continuation until the thread token is exhausted when maxRepliesPerComment: 0. YouTube can hide deleted, spam, or moderated replies, so the advertised replyCount may be higher than the rows returned. repliesFetched tells you exactly how many were emitted for a comment.

Why are timestamps approximate?

The public comment payload only contains relative labels such as "2 days ago". The actor converts each label into a date range with a precision (hour, day, week, month, year) and exposes the earliest and latest possible time. It never claims an exact timestamp.

Why do like counts end in round numbers?

YouTube exposes formatted counts (1.2K, 69K, 4.8M). likeCount is the rounded integer and likeCountText preserves the original string.

What does a date cutoff do?

commentsSince forces the newest-first feed, keeps pinned comments, and stops paginating at the first non-pinned comment older than the cutoff. Because dates are approximate, a comment whose range overlaps the cutoff is kept.

What happens if a video is private, age-restricted, or has comments off?

The actor classifies the page and, when emitStatusRows is enabled, writes one status row explaining the outcome. It does not try to bypass login, age, or region restrictions.

What is the difference between flat, tree, and both?

flat emits one row per comment and reply. tree emits parent comment rows with a replies[] array (useful for older integrations). both emits flat reply rows and also attaches replies[] to each parent.

Do I need to configure a proxy?

No. The actor chooses a working connection to YouTube automatically. Just fill in the input and run it.

Is this affiliated with YouTube?

No. This is an independent tool and is not affiliated with, endorsed by, or sponsored by YouTube or Google.

Limitations

• Exact comment timestamps and exact like counts are not available in the public payload.
• Nested replies beyond direct level 1 are not exposed in the observed public payload; maxReplyDepth above 1 is reserved for future support.
• The top-comments feed is capped by YouTube; use newest-first plus a date cutoff for exhaustive windows.
• Age-restricted, private, members-only, and region-blocked videos are reported as statuses, not bypassed.
• replyCount is advertised, not guaranteed; use repliesFetched for the delivered count.