How to analyse YouTube comments with ChatGPT
Why use ChatGPT for YouTube comment analysis?
Comments are the only feedback you get from people who actually watched. Analytics tells you where they left; comments tell you why, in their own words. Inside them:
- Exact questions your audience wants answered in your next video.
- Hidden objections stopping viewers from buying your course or product.
- Recurring points of confusion in your tutorials.
- Genuine testimonials and positive feedback.
However, manually reading through 500 or 2,000 comments across multiple videos is slow and overwhelming. By feeding exported comments into ChatGPT (or Claude), you can extract actionable insights in seconds.
Step 1: exporting your YouTube comments cleanly
Before prompting ChatGPT, you need a clean text format of your comments:
Pasting a YouTube URL into ChatGPT and asking it to analyse the comments does not work, and it fails in the worst way: it returns a confident summary of the handful of comments the page happened to load. You get output shaped like analysis, drawn from two percent of the data.
- Export first. The free exporter takes a video link and returns JSON, which is the right format here. Comment text is full of commas, quote marks and line breaks, and a model reading a CSV by eye will occasionally lose track of where a row ends. The JSON download exists for exactly this.
- Keep the comment text and like count. Author names, channel URLs and timestamps mostly cost tokens and add nothing, and dropping them avoids handing identifiable people to a third party for no reason.
- Paste a batch into ChatGPT or Claude.
Step 2: 5 high-impact ChatGPT prompts for audience mining
Use these copy-and-paste prompt templates to uncover deep audience insights:
The prompts that return usable output name a decision you are about to make. "Analyse these comments" produces a summary that reads well and tells you nothing.
Prompt 1: finding high-demand video ideas
Act as a professional YouTube content strategist. Below is a list of viewer comments from my recent video on [Topic].
Analyze these comments and extract:
1. The top 5 recurring questions viewers are still confused about.
2. 3 specific video titles and angles I should create next to answer these questions.
3. The exact phrasing or terminology viewers use when describing their problems.
Here are the comments:
[PASTE COMMENTS HERE]Prompt 2: extracting sales objections & pain points
I am creating a digital product/course helping my audience with [Topic]. Below are comments from my YouTube channel.
Identify:
1. The top 3 emotional pain points mentioned.
2. Any skepticism or objections expressed regarding solutions in this niche.
3. What free resource, checklist, or template viewers are explicitly asking for.
Here are the comments:
[PASTE COMMENTS HERE]Prompt 3: sentiment & feedback summary
Analyze the following YouTube comments and categorize them into:
- Positive Feedback / Testimonials (Quotes highlighting what worked well)
- Constructive Criticism (Actionable advice for video editing/pacing)
- Unanswered Questions (Queries that need a creator reply)
Provide a 3-bullet summary of overall audience sentiment.
Comments:
[PASTE COMMENTS HERE]Step 3: categorizing viewer sentiment & objections
When ChatGPT processes your comments, organize the output into a 3-column action matrix:
| Viewer Feedback Type | What ChatGPT Found | Action to Take |
|---|---|---|
| Common Question | "How do I connect this to Shopify?" | Create a 2-minute follow-up Short or full tutorial. |
| Resource Request | "Can you share the spreadsheet from minute 4:15?" | Build a lead magnet checklist and set up a ReplyTide keyword trigger. |
| Objection / Skepticism | "Does this work for beginners with $0 budget?" | Address this objection in the first 30 seconds of your next video intro. |
Step 4: check the answer against the data
This is the step everyone skips. Models produce plausible themes nobody actually mentioned, and the output is fluent enough that you will not notice. Insisting on quotes for every claim, as the prompts above do, gives you something to check.
Take two or three quoted comments and confirm they exist. The comment search does this in seconds: paste the video link, search a phrase from the quote, and either it is there or you have caught a fabrication. If one is invented, discard that section rather than the whole answer.
Turning comment insights into automated lead funnels with ReplyTide
Once ChatGPT identifies the exact resource or checklist your audience wants most, don't just leave a static link in the description that 90% of viewers ignore.
Automate the distribution directly in your comments using ReplyTide:
- Create the checklist or template identified by ChatGPT.
- In your next video, say: "A lot of you asked for the step-by-step spreadsheet, drop the word 'SPREADSHEET' in the comments and I'll send it to you!"
- ReplyTide watches for the keyword and posts the link within a minute or two, including while you are asleep. Put the same link in the description so nobody has to comment to reach it.
Common questions
Is it safe to paste YouTube comments into ChatGPT?
The comment text is already public, so pasting it is not disclosing anything new. Strip author names anyway. They add nothing to the analysis, they cost tokens, and there is no reason to feed identifiable people into a third-party tool.
Can ChatGPT read YouTube comments directly from a video URL?
Unless using specialized third-party GPT plugins or web browsing tools that scrape YouTube, standard ChatGPT cannot directly parse YouTube comment sections from a URL alone. Exporting comments to text or CSV first provides the most accurate analysis.
How many comments can I analyse at once?
More than you will need. Current models handle several thousand comments in one prompt. The binding limit is not context size but attention: a model given four thousand comments and a vague question returns a vague summary, so narrow the question rather than the data.