A YouTube comment export is the cheapest content-strategy research available to a channel: free for the first 100 comments of any video, no signup, and downloadable as CSV, Excel or JSON. This guide is about what to do with that file — specifically, how a comment section decides your next ten video titles better than a keyword tool does.
Views tell you what happened. Comments tell you why.
YouTube Analytics is a measurement instrument, and a good one. It will show you a retention curve with a cliff at 2:40 and an impressions-to-click rate that dropped on your last three uploads. What no analytics dashboard can tell you is why people left at 2:40 — whether the explanation lost them, the music was too loud, or they got the answer and had no reason to stay. Only the comments hold the reasoning, written by people who were not asked and had nothing to gain.
That is also why comments beat surveys for this. A survey asks a question you already thought of. A comment section tells you which question you failed to think of, which is the whole point of doing research at all.
The four strategy questions a comment file answers
Not every question belongs in this data. These four do, and they map directly onto decisions:
- What did I fail to explain? Any question asked more than twice is a gap in the video. Three of them is a follow-up video.
- What do they want next? Explicit requests — "part two", "do this with X", "can you cover Y" — are a content calendar you did not have to brainstorm.
- Who is actually watching? Comment language, the software or gear people mention in passing, and the level of the questions tell you whether your audience is the one you are writing for.
- What do they call it? The words viewers use for your topic, when they differ from yours, are title and search words. YouTube's own help on comments and their moderation is worth reading alongside this, because it explains why some of those words will be missing from the public list.
How we know this
We operate the exporter, so we see YouTube comment data at volume and in comparison with other platforms. Three observations that change how you should read a file.
YouTube threads have a long tail. Unlike short-form platforms, where a comment section is effectively finished within 72 hours, YouTube videos keep accumulating comments for months — a tutorial from last year is still collecting "this still works in 2026" replies. That makes an export a dated snapshot far more literally than on TikTok, and it makes re-exporting before a quarterly review worth the thirty seconds.
The row count will not match the badge. The number under the video counts held, filtered and since-deleted comments; the public list does not serve them. That gap is moderation, not a broken export, and it is why we show rows rather than trying to reconcile with the badge.
Replies carry the disagreement. On YouTube specifically, top-level comments skew towards praise and jokes while the substantive corrections, the "actually this broke for me" and the platform-specific workarounds live several replies deep. A strategy review that reads only top-level rows systematically misses the problems.
The caps and column names quoted here are the live product values. Anything about YouTube's own rules links to Google's documentation rather than our summary of it.
Pulling the file
Paste the video link into the YouTube comment exporter. Watch links, youtu.be links and Shorts links all work, and there is no Google account, API key or extension involved. The first 100 comments of any video are free; larger videos need a one-time payment, never a subscription.
Each row carries author, username, text, likes, replies, created_at, language, is_pinned, the comment id, the reply-to id and the avatar URL. Take CSV or Excel if you intend to sort and read; take JSON if something downstream is going to parse it.
If you would rather build ongoing ingestion yourself, the YouTube Data API commentThreads endpoint is the documented route — it needs a Google Cloud project, a quota you will exhaust faster than you expect, and code. For a one-off strategy review across videos you may not own, it is the wrong amount of work.
Sort by likes first, always
Chronological order is how the data arrives and is almost never how you should read it. A YouTube comment with 4,000 likes has been ratified by 4,000 people who did not have to type anything; a comment with none is one person's opinion. Sort descending on the likes column and read the top fifty rows before you do anything clever.
Two adjustments. Exclude pinned comments from the ranking — the is_pinned flag marks comments the channel elevated, so their likes measure placement rather than agreement. And check the dates on your top rows: on an old video, the most-liked comments are often from the first week and may describe a version of the thing that no longer exists.
Mining the file for titles
This is the highest-return use and it takes about twenty minutes per video.
Filter the text column for rows containing a question mark. Read them, and write each distinct question down as a sentence in the viewer's words, not yours. You now have a list ordered by how often each appeared. The top three are video titles — literally, near-verbatim, because a title phrased the way viewers ask the question matches how they search.
Then filter for request language ("please do", "can you", "part 2", "next video"). Those are a backlog. Anything appearing across several of your uploads has been requested by your audience repeatedly and ignored by you repeatedly, which is a strange position to defend at a content meeting.
Reading a competitor's comment section
The same file on someone else's video is a free gap analysis. You are looking for two things only: the questions their video did not answer, and the complaints about their approach. Both are briefs for content you can make better, and neither requires guessing at their strategy — their audience has written it down.
The practical method is to take the three highest-performing videos on the topic, export each, and pool the question lists. Anything asked under all three is a gap in the topic's coverage as a whole, not a flaw in one creator's video. That is the strongest kind of content opportunity this data produces.
Making it a monthly habit
One export is a project. A routine is a strategy. The version that holds up over time is unglamorous: on the first of the month, export the comment sections of everything published in the last 30 days, keep the files in one folder named by video, and answer the same three questions each time — what was asked repeatedly, what was requested, and what vocabulary appeared that was not in the titles.
Keep the old files. The value compounds precisely because you can compare: a question that was asked once in March and thirty times in September is an audience shifting under you, and no single export would ever show it.
Where this data misleads
Commenters are not viewers. A tiny, unrepresentative and more opinionated fraction of your audience types anything at all, and they skew towards the people who felt something strongly enough to act. Treat comment themes as hypotheses about the silent majority, not measurements of it — then test them the only way that counts, by publishing and watching retention.
Moderation also shapes what you can see. Held comments, blocked words and spam classification remove rows before you ever get them, and on a controversial video that removal is not random. Finally, note the legal frame: the Ninth Circuit's hiQ v. LinkedIn decision put public-data collection outside the CFAA, while display names in your file are personal data under the GDPR for EU and UK commenters. Ordinary audience research is fine; keeping the file forever, or turning it into an outreach list, is not the same activity.
Download YouTube comments free →
Related reading: how to analyse YouTube comments, exporting YouTube comments to Excel, downloading YouTube Shorts comments.