How to export YouTube comments to CSV

Extensions, the official API and its quota, and how to pull comments for a whole channel without babysitting it.

The short answer

For one video, an extension or a copy and paste will do. For a channel, or for anything you need to repeat, use something that pages through properly and writes straight to a file. The interesting part is what happens in between, because that is where most exports quietly go wrong.

Route 1: extensions and web tools

Plenty of them exist and they do work on a single popular video. Their limits show up fast: many stop after the first page of comments, most skip replies, and the free ones tend to cap the number of rows exactly where your video gets interesting. If the export is a one-off and the video is small, take this route and move on.

Route 2: the official YouTube Data API

This is the correct route in the sense that it is official, documented and permitted. It gives you comments with their replies, author, like count and timestamps, and it pages cleanly.

The catch is the quota. Every Google Cloud project gets a daily allowance, and comment requests eat into it alongside everything else that project does. Exporting a channel with a long comment history is exactly the kind of job that hits the ceiling, and when it does, everything else on that project stops too. You can request more, which is a form and a wait.

The second catch is plumbing. OAuth or an API key, paging tokens, retries, and a schema you have to flatten yourself before it becomes a CSV.

Route 3: a dataset that hands you rows

If what you want is a file with one row per comment, skip the plumbing. Our comments dataset is billed per comment listed, at $0.35 per 1,000, which puts a busy video with 20,000 comments at seven cents.

Getting a whole channel

Comments live on videos, so a channel export is two steps: list the videos, then pull comments for that list. Listing videos is billed separately and cheaply, per video listed, at $0.35 per 1,000. In practice the video list is the cheap part and the comments are the volume.

What to decide before you export

• Replies or not. Top-level comments only is a much smaller and often more useful dataset than every reply thread.
• Sort order. Top comments and newest comments give you genuinely different samples. If you are doing sentiment work, newest avoids the popularity bias.
• Deleted and held comments. They will not be there. Your totals will not match the count shown under the video, and that is normal.
• Personal data. Comment authors are people. If you store this, you are processing personal data and the usual rules apply.

The datasets behind this