YouTube Comments Scraper (no login, text + likes, monitor)

Get the top-level comments of any YouTube video: text, like count, reply count, pin/heart status and a relative + estimated publish date. No YouTube account, no cookies, no API key. Runs the same Innertube endpoint the official Android app uses.

Try it: the prefilled example returns 19 rows in about 5 seconds and costs about $0.010. Results cost $0.50 per 1,000.

What you get

Real rows from the example dataset (video dQw4w9WgXcQ):

text	likeCount	replyCount	publishedText	isPinned / isHearted
"can confirm: he never gave us up"	318000	962	1 year ago	true / true
"I Rick rolled my entire French class because we had to study a city in France…"	101000	548	6 years ago (edited)	false / false
"Gonna flag this for nudity so I can rick roll the YouTube staff"	566000	700	6 years ago	false / false

Full row: id, url, videoId, videoUrl, text, likeCount, replyCount, publishedText, publishedAt, isPinned, isHearted, isReply, query, scrapedAt.

Not collected, ever: commenter display name, channel id/handle, avatar image, verification/creator badges. Only the comment content and its counters.

Use cases

• Read sentiment and reactions on your own or a competitor's video without opening YouTube.
• Track how many likes a pinned comment collects over time (monitor mode).
• Feed comment text into a summariser or moderation pipeline.

Try it in 10 seconds

Hit Start/Try it — the input already works: videoUrls: ["https://www.youtube.com/watch?v=dQw4w9WgXcQ"], maxCommentsPerVideo: 20, maxItems: 20, nothing required.

To track a video's comment section over time: save the task, set Mode to monitor, and put it on a schedule (Apify → Schedules → cron 0 */6 * * * every 6 hours). Each run then returns only comments that are new since the previous check.

Related actors

• Reddit Scraper (subreddit posts, search, comments, no login) — the same text-and-counters shape for Reddit posts and comment threads.
• Threads Scraper — Posts, Profiles, Replies & Search, No Login — reply feeds on Threads instead of YouTube.
• Facebook Reels Scraper (page reels, views, MP4, no login) — the video and view-count side of the same content, on Facebook.

How it works

1. Entry — the video's Innertube next response carries the comment section's continuation token, if the video has comments open at all.
2. Pages — each continuation page returns up to ~20 comments with their like/reply counters, pin status and whether the creator hearted them; the actor walks continuations until maxCommentsPerVideo is reached, a page repeats the same ids, or comments run out.
3. Comments turned off (private uploads, some made-for-kids videos) or an invalid video id come back as a normal empty result for that video — not a block, not a charge. If every proxy tier YouTube answers with a wall for a video, the run stops and reports a block instead of silently returning nothing.

Input

Field	Meaning	Default
videoUrls	Video URLs (watch?v=, youtu.be/…, /shorts/…) or bare 11-character ids	required
maxCommentsPerVideo	Stop fetching comments for one video after this many	100
sort	top or newest — see Limits, only top is actually served	top
maxItems	Stop the whole run after this many comment rows	200
mode	scrape or monitor (only new comments since last run)	scrape
monitorStateId, webhookUrl, telegramBotToken, telegramChatId	Monitor-mode state key and alert targets	empty

Pricing

Event	Price
result	$0.0005 per comment ($0.50 per 1,000)
monitor-check	$0.005 per monitor run
change	$0.0005 per new comment

Charged only for comments actually pushed. Residential proxy traffic is billed by Apify on top of these prices when YouTube walls the datacenter tier.

Found it useful? A short review on the Store page helps other people find this actor and tells us what to improve. If something is wrong, open an issue on the actor page — issues are answered within a day.

Why this actor

• No login, no cookies, no API key or quota to manage.
• Comment counters (likes, replies) alongside the text in one row — most comment scrapers on the Store return text only.
• A run that finds nothing pushes nothing and charges no result events; the run summary explains why.

Limits

• Top-level comments only. Replies are not expanded in this version; isReply is always false.
• Sort order. YouTube's Android client (the one this actor reads) does not expose the "Newest first" sort chips in its response, only the default "Top comments" order. sort: "newest" is accepted on input for forward compatibility but every run currently returns the same top-comments order.
• publishedAt is an estimate, computed from YouTube's own relative text ("1 year ago") at scrape time, not an exact timestamp.
• Live/upcoming streams and members-only comments are not specifically handled and may return fewer rows than expected.

FAQ

Does it need a YouTube account or cookies? No — every request is made as a logged-out Android app client.

Why is there no commenter name or avatar? By design: this actor collects comment content and counters only, never person-level identity of individual commenters.

What happens when comments are off for a video? No rows are pushed for that video and nothing is charged; the run continues with the rest of videoUrls.

Changelog

• 0.1: initial release — top-level comments with text, counters, pin/heart status and an estimated publish date; monitor mode; walled videos are reported as a block instead of an empty dataset.