YouTube comments are a goldmine for audience research, product feedback and sentiment analysis. Getting them out at scale is annoying, though: HTML scrapers break or get rate-limited, and writing your own API pagination for comment threads and replies takes an afternoon.
Here's the approach I use: the official YouTube Data API v3, wrapped in a ready-made actor so there's no code to maintain.
Why the official API
- It doesn't get blocked: no proxies, no CAPTCHAs, and no breakage when YouTube changes its HTML.
-
It's cheap on quota: one
commentThreadscall returns up to 100 comments for 1 quota unit. In one test, about 28,500 comments used under 400 units of the free 10,000-unit daily quota. - The data is clean: author, author channel, likes, reply count, publish and update dates.
The tool
YouTube Comments Scraper takes:
- video, Shorts or youtu.be links
-
channels (
@handle); it collects comments from their latest N videos - options: include replies (with a per-comment cap), filter by keywords, minimum likes or date, and sort by top or newest
Every row also includes the video title, channel, views and total comment count, so a multi-video export stays easy to analyze in Sheets or pandas.
Python example
from apify_client import ApifyClient
import pandas as pd
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("datacraft-tools/youtube-comments").call(run_input={
"channels": ["@mkbhd"],
"latestVideosPerChannel": 3,
"maxCommentsPerVideo": 1000,
"includeReplies": True,
"maxRepliesPerComment": 20,
"keywords": ["battery", "price"],
})
df = pd.DataFrame(client.dataset(run["defaultDatasetId"]).list_items().items)
print(df.groupby("videoTitle")["likes"].describe())
Ideas for what to do with the data
- Feed comments to an LLM: "Summarize the top 10 complaints about this product."
- Track how sentiment changes from one video to the next.
- Mine questions from comments for your next video or FAQ.
- Pick a fair giveaway winner from comments that contain a keyword.
Pricing is pay-as-you-go: $0.70 per 1,000 comments plus $0.01 per run.
I built this actor. If you need a feature, tell me in the comments.
Top comments (0)