Most creators decide what to sell by introspection, then find out at launch whether they guessed right. This is a use case for doing it the other way round: a free YouTube comment download, read properly, decided what a channel built, what it called it and roughly what it could charge — before a single lesson was recorded. If you want the general strategy version, YouTube comment export for content strategy covers it. This post is about the revenue.
Your audience already wrote the product brief
The channel in question had about 40,000 subscribers, taught spreadsheet automation, and had spent two years planning a course it never made because it could not decide on the scope. Every version of the outline was a guess about which parts were hard.
The comments already knew. Under one 90-second video about a lookup formula sat 340 comments, and after an hour with the file in Excel the outline stopped being a matter of opinion. Eleven separate people had asked the same question about matching two columns that do not quite match — a question the video did not address and no other video on the channel answered. That question became module one, and module one became the thing people quoted back when they bought.
This is the whole argument for reading comments before building: a comment is unpaid, unprompted evidence written by someone with no stake in flattering you, at a moment when nothing was for sale. That is a different and better class of information than anything a survey or a poll returns.
Picking the videos worth mining
Not every video's comments are useful for this. The ones that are share a shape: the video solves a narrow problem, and the comments are full of people describing a slightly different version of that problem.
Skip your most viral video. Its comments are dominated by people who arrived from recommendations, found it entertaining and will never buy anything — high volume, low intent, and the reason so many channels mistake reach for demand. Choose instead the videos with a high comment-to-view ratio and long comments. A video with a tenth of the views and twice the questions is worth more here, because typing a long specific question is an effort signal.
Then add two or three videos from other channels on the same topic. Paste each link into the YouTube comment exporter and take Excel for the ones you intend to read by hand. Your own comments tell you what your audience trusts you for; someone else's comments tell you which problems the field as a whole has left unsolved, which is where an unmet need actually lives.
Four phrases that mean someone would pay
Reading 800 pooled rows top to bottom is a waste of an afternoon. Four text searches do most of the work, and anyone building a comment-mining habit should learn them in this order:
- "how do I" / "how can I" — the unsolved-problem search. Every distinct completion is a candidate lesson. Frequency ranks your outline for you.
- "I tried" / "I've been using" — the alternatives search. This is what you are competing against, in the customer's own words: a template someone downloaded, a plugin, a colleague who does it for them, or three hours a week of manual work. You cannot price a product without knowing this.
- "still" / "every time" / "keeps" — the pain-frequency search. "Still can't get this to work" and "every time I reopen the file" describe a recurring cost, and recurring costs are what people pay to remove. One-off annoyances are not products.
- "part 2" / "full course" / "would pay" / "is there a" — the explicit-demand search. The smallest set and the most quotable. Screenshot these; they are what silences your own doubt at week three of building.
Tag each hit in a new column with the theme it belongs to, then count themes. The channel's count came out at 11 rows for the mismatched-columns problem, 9 for cleaning imported data, 7 for automation that breaks when a sheet is renamed, and a long tail of ones. Three modules, ranked by evidence, decided in an hour.
Naming it in their words, not yours
The channel's working title used the word "automation". Nobody in 800 comment rows used that word. What they wrote was "fix", "clean up", "stop breaking", and — repeatedly — "without VLOOKUP".
Product names and sales headlines built from vocabulary that appears in the file outperform names built in your own head, for the obvious reason: the words people use for their problem are the words they recognise, search and click. Pull the twenty most frequent non-trivial words from the text column, discard the platform noise ("video", "thanks", "subscribed"), and what remains is a naming shortlist you did not have to invent. It doubles as the language for your thumbnails and your comment analysis going forward.
Deciding what to charge, with evidence instead of nerve
Comments do not contain prices, and anyone who tells you a comment file reveals willingness to pay is selling something. What the file does contain is the two things a price is actually built from.
The first is the alternative. If the "I tried" rows are full of free templates, you are priced against free and you need to be obviously, demonstrably different. If they are full of "I pay someone on Fiverr to do this each month" or "we bought a plugin", the anchor is much higher and pricing timidly costs you real money.
The second is frequency. A problem described as happening every week is worth many multiples of one described as happening once. The channel's pain rows were overwhelmingly weekly, and against a recurring several-hours-a-week alternative the product was clearly underpriced at the figure originally planned. It launched at roughly twice that and the objection rate did not move.
The number itself still comes from the market, not the file. What the file buys you is the confidence to test a higher one.
The presell: replies before a single lesson exists
With the outline and the name settled, the fastest validation is to go back to the people whose comments produced it. Not a mass message — YouTube has no such thing, and the equivalent on other platforms is how accounts get restricted — but a reply, in the thread, to the specific question that was asked, ending with a genuine note that you are making a thing about exactly this and a link to a waitlist.
Forty-one replies, written one at a time over two evenings, produced 180 waitlist signups, because the replies are public and the lurkers who never typed anything read them too. This is the mechanic most creators miss: a reply to one person is a broadcast to everyone who had the same question and stayed silent. The likes column is your guide to which questions those are — a question with 200 likes is 200 people waiting.
Only then does building start, and with the waitlist's replies in hand the first module changes shape at least once. That is the presell working as intended.
Launch week: the file becomes the sales page
The unglamorous payoff is that the same export is a copy bank. Your sales page needs a problem statement, an objection list and an FAQ, and inventing all three is why sales pages take a week and still sound like nobody.
The problem statement is the most-liked question, near-verbatim. The objection list is the replies buried in the threads — "I tried that and it broke when the source file moved" is an objection you must answer above the fold, and it was written by a customer, not a copywriter. The FAQ is the eleven distinct completions of "how do I", each answered in one sentence, which also happens to be the section that earns the page citations in AI answers and search results.
Keep the file until the page is written, then delete it. Comment rows carry display names, and that makes them personal data under the GDPR for EU and UK commenters. Reading public comments is lawful — the Ninth Circuit's decision in hiQ v. LinkedIn put public-data collection outside the CFAA — but research is not a licence to keep a name list indefinitely, and it is certainly not consent to email anyone.
Why we can vouch for this method
We run the exporter, which means we see YouTube comment data at volume and next to every other platform, and three observations from operating it shaped the advice above.
YouTube comment sections have by far the longest useful life of anything we export. Short-form platforms finish within roughly 72 hours; YouTube tutorials keep collecting "does this still work in 2026" comments for years. For product research that is a gift — a two-year-old video's comment section is a longitudinal record of a problem that has not been solved — and it is why re-exporting before a launch is worth thirty seconds.
Replies carry the disagreement. Across the YouTube files we see, top-level comments skew to praise and jokes while corrections and failure reports sit in replies. Any research pass that reads only top-level rows will produce a product with an objection list it has never seen, which is the expensive kind of mistake.
The row count will not match the badge under the video, and that difference is moderation rather than a defect. We report rows served, because that is the list the platform actually gives anyone.
Caps and column names quoted here are the live product values. For YouTube's own rules on comments and moderation, Google's comments documentation and the commentThreads API reference are the sources, not our paraphrase of them.
The failure mode: demand that evaporates at checkout
The honest limit of this method is that commenters are a small, self-selected and unusually vocal slice of an audience, and "I would pay for this" is free to type. Some themes are loud because a few people are loud.
Three guards against it. Require a theme to appear across several videos, including at least one you did not make, before it becomes a module. Weight by likes rather than raw count, because a like is a second person agreeing at zero social cost. And never skip the presell step — a waitlist that converts is the only evidence that turns a comment theme into revenue, and it costs two evenings of replies to get.
Treat comment themes as hypotheses with unusually good provenance. Then charge money and find out.
Do it with one video this week
Take the video on your channel with the most questions per view, export its comments, and run the four searches. If eleven people have asked you the same thing, your next product is already named. The first 100 comments of any video are free with no signup; larger threads need a one-time payment, never a subscription, and pricing has the detail.
Download YouTube comments free →
Related reading: YouTube comment export for content strategy, exporting YouTube comments to Google Sheets, turning Instagram comments into DM profit.