YouTube Comment Scraper

Read YouTube comment threads as a corpus. Start with YouTube post or video URLs; each returned comment keeps comment, comment identifiers, text, authors, and handles.

What it returns

Each row keeps the YouTube source record beside the fields needed to use it. The opening set is whatIFound, text, comment, url, videoId, commentId, parentCommentId, replyLevel, author, username, authorChannelId, and authorUrl. The complete schema is declared before the run, and dataset views keep related fields together without changing the underlying row.

Captured row

{
  "commentId": "Ugzge340dBgB75hWBm54AaABAg",
  "videoId": "dQw4w9WgXcQ",
  "likeCount": 304000,
  "replyCount": 963
}

Input

YouTube Comment Scraper accepts source URLs. Run controls stay in the same form.

Field	What it controls	Starting value
startUrls	Paste exact YouTube URLs, one per line.	["https://www.youtube.com/watch?v=dQw4w9WgXcQ"]
maxItems	Stop after this many dataset rows.	100
includeReplies	Keep replies in the returned record.	true
maxSeconds	Stop after this many seconds and keep completed rows.	240

Example input

{
  "startUrls": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
  ],
  "maxItems": 3,
  "includeReplies": true,
  "maxSeconds": 240
}

No field is required. Start with the filled example, then replace only the target values needed for the job. Run controls can stay at their starting values for the first collection.

Dataset fields

Field	Type
whatIFound	string
text	string
comment	string
url	string
videoId	string
commentId	string
parentCommentId	string
replyLevel	integer
author	string
username	string
authorChannelId	string
authorUrl	string
authorAvatar	string
authorVerified	boolean
authorIsCreator	boolean
publishedAt	string
publishedTimeText	string
likes	string

Dataset views

Views are working surfaces for review and export. They select and order fields while leaving the stored row unchanged.

View	Opening fields
overview	whatIFound, text, pinnedText, comment, commentId, parentCommentId, likes, and likeCount
answer	author, username, url, and whatIFound
content	author, username, url, text, and pinnedText
engagement	author, username, url, comment, commentId, parentCommentId, likes, and likeCount
identity	author, username, authorChannelId, authorVerified, authorIsCreator, and videoId
timing	author, username, url, publishedAt, publishedTimeText, scrapedAt, sourceUrl, and route

Output and exports

Output	Type	Destination
results	string	{{links.apiDefaultDatasetUrl}}/items
json	string	{{links.apiDefaultDatasetUrl}}/items?clean=true&format=json
csv	string	{{links.apiDefaultDatasetUrl}}/items?clean=true&format=csv
excel	string	{{links.apiDefaultDatasetUrl}}/items?clean=true&format=xlsx
jsonl	string	{{links.apiDefaultDatasetUrl}}/items?clean=true&format=jsonl

Completed rows are available in the Apify dataset as JSON, CSV, Excel, and JSONL exports. The run output also carries the declared links above for API clients and automations.

Pricing

$3.45 per 1,000 dataset items on the Free plan. Other Apify plans use the rates shown in the Pricing tab.

Console, API, schedules, and MCP

Runs can begin in Apify Console, from a saved task, or through the Actor API. A schedule can reuse the same input, and a run-finished webhook can pass the dataset or run ID to the next system.

POST https://api.apify.com/v2/acts/5N0e72dyBufNBTtT5/runs
GET  https://api.apify.com/v2/datasets/{datasetId}/items

For MCP selection, use YouTube Comment Scraper. Its machine entry carries the same description, input field names, no-required-field contract, output types, dataset fields, views, and pricing facts as this document.

Saved tasks

Twenty saved-task products cover distinct lookup, comparison, research, operations, automation, and export jobs:

• YouTube Comment reply count core record view: buyer-job; opens overview.
• YouTube Comment finding and authors finding brief: buyer-job; opens answer.
• YouTube Comment content handles copy review: buyer-job; opens content.
• YouTube Comment comments response readout: buyer-job; opens engagement.
• YouTube Comment authors author channel ID identity: buyer-job; opens identity.
• YouTube Comment publication dates timing timeline: buyer-job; opens timing.
• YouTube Comment reply count exact reply count core: buyer-job; opens overview.
• YouTube Comment answer finding result handoff: buyer-job; opens answer.
• YouTube Comment text content brief: buyer-job; opens content.
• YouTube Comment ID and parent comment ID audience sizing: buyer-job; opens engagement.
• YouTube Comment name and ID match: buyer-job; opens identity.
• YouTube Comment published time text collection times timing: buyer-job; opens timing.
• YouTube Comment reply count full record: buyer-job; opens overview.
• YouTube Comment finding answer decision file: buyer-job; opens answer.
• YouTube Comment pinned text handles content: buyer-job; opens content.
• YouTube Comment response parent comment ID rating view: buyer-job; opens engagement.
• YouTube Comment authors handoff: buyer-job; opens identity.
• YouTube Comment publication and collection dates: buyer-job; opens timing.
• YouTube Comment reply detail handoff: buyer-job; opens overview.
• YouTube Comment finding links answer: buyer-job; opens answer.

Integrations

Use the dataset API from any HTTP client, export rows to a spreadsheet, or send the run ID through an Apify webhook. Saved tasks give schedules and automation tools a stable input without changing the Actor contract.

Related products

• YouTube Video Scraper
• YouTube Data Export Scraper
• YouTube Transcript Scraper
• Google Maps Review Scraper
• Amazon Review Scraper

When a run needs attention

• No rows: Open the target in a browser, check spelling and source visibility, then retry the saved example before widening the input.
• A field is empty: Check the field beside its source URL. A missing source value stays empty instead of being replaced with a guess.
• A target fails: Keep successful targets in the dataset, then retry only the affected input.
• An automation cannot find results: Read the dataset ID from the run and request its items endpoint directly.

FAQ

What do I get back from one run?

One row per comment with 28 declared fields, opening on whatIFound, text, comment and url. The schema is published before the run, so you know the shape before you spend anything.

Do I need a youtube account or login?

No. The run works from the youtube sources you supply in the input. Nothing is posted, changed or accessed on your behalf.

What does a run cost?

The current rate is shown on the Pricing tab and is charged per row you receive, so a run that finds nothing costs close to nothing. Cap the run with the item limit when you want a predictable ceiling.

Can I try it before committing budget?

Yes. Cap the run with the item limit in the input and inspect the first rows. The cap is enforced before charging, so a trial run stays a trial.

What do I put in the input?

The staged input is already usable: startUrls, maxItems and includeReplies. Replace the staged target with your own list when you are ready to run for real.

Are any fields required?

No field is required. Every input carries a working default, so the Actor can be started as-is and refined afterwards.

Can I schedule this to run on its own?

Yes. Save the input as an Apify task and attach a schedule. Keep separate tasks when different teams need different targets or delivery paths.

Do I need to configure proxies?

No. Network access is handled inside the Actor and needs no proxy configuration from you.

How fresh is the data?

Every row is collected during the run you start, not served from a cache. Re-run the same input whenever you need the current state of a youtube comment.

Can I use the results commercially?

The Actor collects publicly accessible youtube information. You remain responsible for how you use it, including any privacy or contractual obligations that apply to your business.

How do I compare two runs?

Keep whatIFound and text as your join key and diff the exports. The identity fields stay stable across runs, which is what makes a comparison meaningful.

What happens if a source fails mid-run?

The run continues through the remaining sources and finishes with what it collected. Partial results are still written to the dataset rather than discarded.

Can I limit how long a run takes?

Yes. The maxItems input caps the run. Use it when you need a predictable cost and a predictable finish time.

How do I report a problem?

Open an issue on the Actor with the run ID, the input you used and the field or row that needs attention. The run ID lets the exact execution be inspected.

How is this different from YouTube Video Scraper?

YouTube Comment Scraper answers one job: Read YouTube comment threads as a corpus.. YouTube Video Scraper covers a different question on the same platform. Run both when you need both sides.

How do I read the output without scrolling through JSON?

Open the overview view on the Output tab. 6 views ship with the Actor (overview, answer, content, engagement, identity and timing), each grouping the fields that belong to one question.

How do I get the data into my own tools?

Export the dataset as JSON, CSV, Excel or XML, call the dataset API directly, or attach a run-finished webhook and collect the dataset reference as soon as the run ends.

Can an agent or LLM call this?

Yes. YouTube Comment Scraper is exposed over MCP with the same description, no-required-field input contract and output types shown here, so an agent can select and call it without a human in the loop.

Why is a value empty on some rows?

youtube does not expose every field on every comment. An absent value stays empty rather than being filled with a guess, so a row never invents a fact it did not receive.

A run returned fewer rows than I expected. Why?

The usual causes are a narrow source list, an item cap still set low, or a source that genuinely holds less than expected. Widen the input or raise the cap and run again.

Support

Use the Actor issue form for product questions, broken source routes, schema mismatches, and feedback. Include the smallest input that reproduces the problem. That is enough to locate the run and its dataset without sharing an entire working list.

Use this Actor only for data you are allowed to collect. Follow source terms, privacy law, and your own retention policy.