September 28, 2026
YouTube comment mining is the fastest way to turn an existing audience into a backlog of validated video ideas. Instead of guessing what viewers want, you read what they are already asking, complaining about, and requesting under the videos that dominate your niche. In about one hour of focused work, you can extract 50 evergreen ideas, categorize them, and rank them by demand.
This method works especially well for documentary channels, where a single well-chosen topic can carry a 20-30 minute video for years. The comments under top-performing documentaries are a live focus group: they reveal technical doubts, comparison requests, myths the audience believes, and stories they explicitly want told. You just need a repeatable process to capture and score them.
Below is the exact workflow: which channels to pick, which YouTube filters to use, how to export manually to a sheet, how to sort comments into four buckets, and how to score each idea with a simple template. No paid tools required beyond a spreadsheet.
Keyword tools show you search volume. Comments show you intent plus emotion. In documentary niches, search volume often misleads because viewers phrase their curiosity as full questions, not keywords. A tool might show "ancient rome" at 200k monthly searches, but the comments tell you exactly which angle is missing: "Why did nobody explain how Roman concrete actually worked?"
Three practical advantages of comment mining:
The tradeoff is manual labor. Expect 45-70 minutes of real work for 50 ideas, and expect roughly 30-40% of raw comments to be useless (spam, insults, off-topic). That failure rate is normal. Plan for it.
You want channels that (a) publish in your language, (b) have videos with 500k+ views, and (c) get at least 200 comments per popular video. If a channel has fewer comments, skip it. Sparse comment sections produce weak signals.
For a faceless documentary niche, public examples of channels with dense comment sections include Kings and Generals, Real Engineering, Fall of Civilizations, Lemmino, and MagnatesMedia. Pick five in your specific sub-niche (e.g., ancient history, engineering failures, business collapses) rather than five generalists.
How to find them fast:
That gives you 15 videos to mine. Fifteen is enough. More than that and you'll spend hours without better output.
YouTube's built-in comment sorting is underrated. You don't need a scraper for this step. Open the comment section of each video and use these two filters in order:
Then use the browser's find function (Ctrl+F or Cmd+F) inside the page and search for question marks and question words: why, how, what if, did, was, is it true, can you. This instantly filters thousands of comments down to the ones with actual questions.
Practical rule: capture only comments that contain a question, a comparison ("X vs Y"), a correction ("actually, this is wrong because..."), or an explicit request ("please make a video about..."). Everything else is noise.
If you capture these, your 50-idea list becomes 30 real ideas and 20 wastes of time. Be strict.
You don't need an API or a paid scraper. Copy-paste works and takes about 10 minutes for 15 videos. Build a sheet with these columns:
| Column | What to enter | Example |
|---|---|---|
| A: Channel | Source channel name | Fall of Civilizations |
| B: Video | Source video title | The Bronze Age Collapse |
| C: Comment (verbatim) | Copy the comment exactly | "Why did nobody cover the Sea Peoples' origins?" |
| D: Likes | Number of likes on the comment | 412 |
| E: Bucket | One of 4 categories (see next section) | Requested story |
| F: Frequency | How many times a similar comment appears | 7 |
| G: Search intent | High / Medium / Low | High |
| H: Score | Calculated (see template) | 8.4 |
Column F is the most important and the most skipped. Frequency is what separates a one-off opinion from real demand. If you see the same question under three different channels, mark frequency 3+. That's a signal.
Every useful comment falls into one of four buckets. This is what turns 200 raw comments into a structured idea list.
The viewer doesn't understand a mechanism, a cause, or a process. Example: "How did they actually move those stones without wheels?" These become explainer videos with a clear, single-question hook.
The viewer wants two things contrasted. Example: "Roman roads vs modern highways — which lasted longer?" These perform well because they force a structured, visual video.
The viewer states something false with confidence, or asks if something is true. Example: "Did Vikings really wear horned helmets?" Debunk videos get high CTR because they promise a correction.
Direct requests for a topic. Example: "Please cover the Hanseatic League." These are the easiest to validate because the audience is literally asking.
Target distribution for 50 ideas: roughly 15 technical doubts, 10 comparisons, 10 myths, 15 requested stories. Adjust to your niche, but keep all four buckets represented. A list with only "requested stories" is fragile — you're dependent on what viewers happen to mention.
Not all 50 ideas deserve production. Score each one with a simple formula. This template is copy-ready:
| Factor | Weight | How to score (1-10) |
|---|---|---|
| Frequency | 35% | 1 = once, 5 = 3-4 times, 10 = 5+ times |
| Search intent | 25% | 10 = clear question, 5 = vague curiosity, 1 = opinion |
| Evergreen potential | 20% | 10 = timeless, 5 = 2-3 year relevance, 1 = trend |
| Production feasibility | 20% | 10 = easy visuals, 5 = moderate, 1 = requires rare footage |
Formula: Score = (Frequency × 0.35) + (Intent × 0.25) + (Evergreen × 0.20) + (Feasibility × 0.20)
Sort descending. Anything above 7.5 goes into your immediate production queue. 6.0-7.4 is your backup list. Below 6.0, archive it — don't delete, because audience interest shifts.
Realistic output: from 200 captured comments, expect roughly 50 usable ideas, of which 12-18 will score above 7.5. That's your next 3-4 months of content.
Time honesty: if you try to do this for 10 channels instead of 5, you'll spend 3+ hours and get maybe 15% more ideas. Diminishing returns kick in fast after channel five.
Roughly 150-250 raw comments. From that pool, 40-50% will be usable after filtering, and about 50 will survive as scored ideas. If you're getting fewer, your source videos likely have thin comment sections or you're filtering too aggressively.
You can use the official YouTube Data API for comment retrieval within quota limits, but for 15 videos the manual copy-paste method is faster to set up and avoids quota issues. Paid scrapers exist but rarely justify the cost for a one-hour task.
Start with "Top comments" to catch the most-liked questions and myths, then switch to "Newest first" to catch current audience interest. Combining both gives you a balanced signal. Use Ctrl+F with question words to filter fast.
Ask: will this question still make sense in three years? Questions about historical events, scientific mechanisms, and comparisons between long-standing things are evergreen. Questions about a specific 2026 event, a specific creator, or a specific trend are not.
No — it complements it. Comments give you intent and phrasing; keyword tools give you volume and competition data. The strongest workflow is to mine comments first, then validate the top 10 ideas with a keyword tool before production.
1. Frequency beats volume. A question asked 7 times across 3 channels is worth more than a single viral comment. Always track frequency in your sheet.
2. The four buckets keep you balanced. Technical doubts, comparisons, myths, and requested stories each serve a different audience mood. A list with all four is more resilient than one dominated by a single type.
3. Score before you produce. The template above takes 15 minutes and saves weeks of wasted production. Anything below 6.0 should wait.
If you'd rather skip the production bottleneck entirely, AI-powered documentary production services like VAATIK can turn a scored idea list into finished 20-30 minute videos, so your hour of comment mining becomes a quarter of published content instead of a backlog. Whether you produce in-house or with VAATIK, the mining method above is what determines whether those videos find an audience.
See How VAATIK Can Run Your Channel → Partner Program → €200 + 10% Recurring