How to export YouTube comments for every video on a topic

The job: you sell a product, or plan to, and the review videos about it carry thousands of comments from buyers and doubters. Search for the videos people actually watch, then read the comments under each one into a single sheet. You end with one CSV where each row is a comment with its video, text, author, likes, reply count and date, ready for tagging by hand or for a model to group into complaints and praise.

No Google Cloud project, API key or quota request is involved. Each call is an HTTPS request with your key.

Step by step

Step 1: find the videos

Call YouTube video search with query and sort: relevance (the default), latest or most_viewed. One call returns about 20 videos for $0.004, each with id, title, url, the channel in author and metrics.views. Search leaves metrics.comments null, so choose videos by views.

curl -G "https://api.okdata.online/v1/youtube/videos/search" \
  -H "Authorization: Bearer $OKDATA_API_KEY" \
  --data-urlencode 'query=robot vacuum review' \
  -d sort=most_viewed \
  -d fields=id,title,author.name,metrics.views

Step 2: page through the comments

Pass a video's id to YouTube video comments. sort takes top (the default, YouTube's Top comments order) or latest (newest first). Use top to read what most viewers saw, latest to follow a video that is still collecting comments.

One call returns about 20 comments for $0.002. Each has id, text, author (name and username, the handle), likes, replies (how many replies sit under it) and published_at. For the next page, send page.next_cursor back as cursor, unchanged; stop when page.has_more is false.

curl -G "https://api.okdata.online/v1/youtube/videos/comments/list" \
  -H "Authorization: Bearer $OKDATA_API_KEY" \
  -d id=h4Lw0pZk3sE -d sort=top

Step 3: put it together in Python

The script keeps the four most viewed videos and reads MAX_PAGES pages under each. Every page is its own call, so the page cap is also your spending cap.

import csv
import os
import requests

API = "https://api.okdata.online"
HEADERS = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
VIDEOS = 4
MAX_PAGES = 2  # comment pages per video; each page is one call


def get(path, **params):
    body = requests.get(API + path, headers=HEADERS, params=params, timeout=60).json()
    if body["code"] != "ok":
        raise RuntimeError(f"{body['code']}: {body['message']} (charged: {body['charged']})")
    return body


found = get("/v1/youtube/videos/search", query="robot vacuum review", sort="most_viewed")["data"]["items"]
videos = sorted(found, key=lambda v: v["metrics"]["views"] or 0, reverse=True)[:VIDEOS]

with open("youtube_comments.csv", "w", newline="", encoding="utf-8-sig") as f:
    out = csv.writer(f)
    out.writerow(["video_id", "video_title", "comment_id", "author", "likes", "replies", "published_at", "text"])
    for v in videos:
        cursor = None
        for _ in range(MAX_PAGES):
            body = get("/v1/youtube/videos/comments/list", id=v["id"], sort="top", **({"cursor": cursor} if cursor else {}))
            for c in body["data"]["items"]:
                out.writerow([v["id"], v["title"], c["id"], c["author"]["name"], c["likes"], c["replies"], c["published_at"], c["text"]])
            if not body["page"]["has_more"]:
                break
            cursor = body["page"]["next_cursor"]

utf-8-sig writes a byte-order mark so Excel shows emoji and non-English text correctly. To go deeper, raise MAX_PAGES: ten pages under each of the four videos is 40 comment calls, $0.08 on top of the search.

Good to know

• Top-level comments only. You get each comment's reply count, not the replies themselves; parent_id is always null here.
• Dates are approximate. published_at is worked out from labels such as "2 months ago", so older comments carry a rounded date.
• No commenter location. YouTube does not show it, so ip_location is null and named in meta.unavailable.
• Removed videos are free. A video that no longer exists returns not_found and costs nothing.
• The official quota. Google's quota calculator gives a new project 100 search.list calls a day in their own bucket and 10,000 units a day for the other methods, reset at midnight Pacific time. commentThreads.list costs 1 unit and returns up to 100 threads per call. If you already have a Google project and mostly read comments, the official API is the cheaper choice; the search quota is where a topic-wide job usually runs out first.
• Retries without paying twice. Send an Idempotency-Key header with each call. The same request with the same key within 24 hours returns the stored answer with meta.replayed: true and no new charge.

For a single video's comments in a spreadsheet, see export YouTube comments to CSV. When the daily search limit is the problem, see the YouTube search quota answer.

{
  "endpoint": "youtube.videos.search",
  "params": {
    "query": "claude code tutorial"
  },
  "idempotency_key": "<a new UUID>"
}