YouTube comments are one of the few places where people tell you, unprompted, what they liked, what confused them and what they want next. You can mine them without logging in. Open any public video, switch the sort order from Top comments to Newest first, expand the reply threads and read the pinned comment at the top. That already gives you most of what a comment section can offer. For a single video, no tool is needed. For dozens of videos, or for the same video tracked over weeks, you need either the official YouTube Data API or a scraper that returns comments as rows.
The catch is that a comment section is a biased sample. It shows you the people who felt strongly enough to type something, filtered by what the creator chose to hide or pin, and sorted by an algorithm you do not control. It cannot tell you how many viewers agreed in silence, how many left after ten seconds, or who the commenters actually are. The rest of this article covers what each part of the comment section means, how to pull it out at scale and where to stop trusting it.
What a logged-out visitor can and cannot see
Everything in this article assumes a public video with comments enabled. If you are not signed in, you can still read the full comment section, switch between the two sort orders, expand replies and see like counts on individual comments. You can also see the pinned comment, the creator's own replies marked with the channel badge, and whether a comment was edited.
You cannot see comments the creator has held for review or removed. YouTube's own help page on comment moderation explains that creators can hold comments for review, block specific words and hide users from their channel entirely. None of those comments appear to the public. See Manage comments on your channel in YouTube Help. If a product has a known defect and the creator moderates aggressively, the comment section will undercount it.
You also cannot see comments on videos marked as made for kids, because comments are disabled there by policy. You cannot see comments on private or unlisted videos unless you have the link and the owner allowed it. And you cannot see dislikes on the video itself, since YouTube made the dislike count private in 2021. Like counts on individual comments are still public.
Finally, the comment count shown under a video is not always the number you can actually load. Deleted comments, spam filtered after the fact and replies on removed threads can all create a gap between the stated count and what scrolls into view.
Sorting: Top comments versus Newest first
YouTube offers two sort orders in the web interface. Top comments is the default. Newest first is the other. They answer different questions.
Top comments surfaces what the ranking considers engaging. In practice that favours comments with many likes and many replies, comments from the creator, and comments posted early while the video was getting most of its traffic. It is good for finding the dominant reaction to a video. It is bad for finding rare but important feedback, because a single person reporting a bug or a confusing step will sit far down the list.
Newest first is a time series. It shows you how the conversation changed after the video stopped being new. Late comments are often more useful for product work. The early audience is the creator's existing fan base. The late audience arrives through search and recommendations, often with a specific problem in mind, and they are more likely to ask the question you want to hear. If you only read Top comments you will mostly hear the fans.
For content ideas, read Newest first on your own older videos. Questions that keep arriving months later are the questions your existing content does not answer. That is a video brief, written for you by the audience.
Replies: where the actual information lives
A top-level comment is often a reaction. The reply thread is where the details are. Someone writes that the product broke. Three replies down, another person asks which version, and the original commenter answers with the model number and what they tried. That reply is the useful record, and it is collapsed by default.
If you mine comments at scale, make sure you capture replies as separate records with a reference to their parent. Treat a thread as a unit when you read it. The parent gives you the topic, the replies give you the evidence. Counting only top-level comments systematically undercounts technical detail and overcounts one-line reactions.
Replies from the creator deserve their own flag. A creator reply usually means the creator saw the comment as worth answering. It also tells you what the creator already knows. If you are a brand watching a reviewer's video about your product, the creator's replies show you which complaints were acknowledged and how they were framed.
Pinned comments: the creator's editorial choice
A pinned comment sits at the top regardless of sort order. Only the channel owner can pin one, and only one at a time. The YouTube Help page on using comments describes pinning along with hearting and the other creator-side actions.
Pinned comments are useful in two ways. First, they are often corrections or updates. A creator who learned after publishing that a step was wrong will pin the fix. If you are mining a tutorial for product issues, the pinned comment is sometimes the most honest line in the whole section. Second, they tell you what the creator wants the audience to focus on, which is a signal about their priorities and sometimes about sponsorship.
Treat them as a separate field. Do not let a pinned comment inflate your sentiment counts as if it were an ordinary top-ranked reaction. It got its position by decision, not by likes.
Turning comments into product feedback
The practical workflow is simple. Collect comments and replies as rows. Keep the text, author display name, publish time, like count, whether it is a reply, the parent id, and whether the creator posted it or pinned it. Then do three passes.
First, filter for questions. Anything ending in a question mark, or starting with how, why, does, can, where. Questions are requests for content and reports of confusion in the same breath. Group them by theme and count the themes, not the individual comments.
Second, filter for comparisons. Comments that mention another product or an earlier version are where positioning feedback lives. People rarely say a product is good or bad in isolation. They say it is better or worse than the thing they had before.
Third, read the long comments. Length correlates with effort, and effort correlates with information. A comment over a few hundred characters is usually someone explaining a real experience. Sort by length and read the top of that list by hand.
For content ideas, repeat the same passes across your own channel and across channels in your niche. Questions that appear on multiple creators' videos and never get a dedicated answer are gaps in the market.
What a comment section cannot tell you
Be clear with yourself about what this data is not.
It is not a survey. Commenters self-select, and the people who comment are not representative of the people who watch. Strong opinions are over-represented. Satisfied viewers who got what they came for mostly leave without typing.
It is not complete. Moderation, spam filtering and deletions remove an unknown share of what was posted. You cannot measure that share from the outside.
It is not verified. Display names are chosen by the user. You cannot tell whether a commenter owns the product, works for a competitor, or is the creator's relative. Like counts on comments can be influenced by timing and by who the creator replies to.
It is not a measure of reach. A video with ten thousand views and fifty comments tells you about fifty people. Retention, click-through and watch time live in YouTube Analytics, which only the channel owner can see.
It is not stable. Comments can be edited and deleted after you collect them. Sort order changes as new likes arrive. Two collections a week apart will differ, and that is normal, not an error.
If you need the audience's opinion rather than the commenters' opinion, you need a different instrument. A survey, a support ticket export or your own analytics will answer questions a comment section cannot.
Free ways to get the same data, and when they are better
For one video, read it in the browser. Switch to Newest first, expand the replies, and copy what matters. This is faster than any tool for a handful of videos and it costs nothing.
For your own channel, use YouTube Studio. The comments tab lets you filter by unanswered, by whether you have responded, by subscriber count and by whether a comment contains a question. That is more than any public scraper can see, because it includes held and reviewed comments.
For many videos, use the official YouTube Data API. The commentThreads endpoint returns top-level comments with their replies, and the comments endpoint returns the rest of a thread. The API reference is at developers.google.com/youtube/v3/docs/commentThreads. It is free within a daily quota, it is documented, it is stable, and it is the right choice if you are comfortable writing a short script and your volume fits inside the quota. The API also respects the same visibility rules as the public page, so it will not show you moderated comments on channels you do not own.
The API becomes awkward in three cases. When your volume exceeds the daily quota and you do not want to apply for more. When you need the data in a spreadsheet today and do not want to set up a Google Cloud project. And when you want fields the API presents differently, such as the pinned state, which is visible on the page but not exposed as a simple flag in the comment resource.
When a scraper makes sense
A scraper is a convenience, not a source of hidden data. It sees what a logged-out browser sees, nothing more. Its value is that it hands you the comments and replies as structured rows for a list of video URLs, with the parent reference and the author and the timestamps already filled in, without quota planning or code. If that saves you an afternoon, it is worth paying per result. If you have one video or your own channel, use the browser or Studio and keep the money.
If you want comments and replies as a dataset for a list of videos, including the sort order, reply threads and creator flags described above, the YouTube Comments Scraper on this site returns them as rows you can filter and export, priced per comment delivered.
Want the signal instead of the raw filings? Get a free report preview. Prefer the tool to the write-up? Browse all data feeds or connect the free MCP server.