Blog · · Benatar Brands
How to find your customers' real questions in YouTube comments with AI
Your customers are already asking questions in public. They ask them under the YouTube videos they watch before they buy: product reviews, how-to videos, comparisons. Reading those comments by hand takes hours, so most owners never do it. An AI assistant can do the reading and hand you a short list of the questions that keep coming up, in your customers' own words.
This post shows how to set that up with Grok Bot and a connected YouTube research tool, and where the method stops being reliable.
Why YouTube comments are worth reading
YouTube is where a large share of your buyers already spend time. Pew Research Center's social media fact sheet, from a survey of U.S. adults conducted January 30 to June 18, 2026, shows YouTube use at 95% of adults ages 18 to 29, 91% of ages 30 to 49, 85% of ages 50 to 64 and 65% of those 65 and older. YouTube and Facebook are the most widely used platforms in the survey.
Comments are useful because nobody wrote them for you. A comment like "does this work on hard water?" under a review video is a question a real person had at the moment they were deciding. That is better raw material than a guess in a planning meeting.
It also lines up with how Google describes good content. Its guide to creating helpful, people-first content asks whether you have an audience that would find the content useful if they came directly to you, and whether the content shows first-hand expertise. Answering the questions your buyers actually ask, from your own experience, fits both.
What the assistant can actually do
With the TubeAlfred YouTube connector added, Grok Bot can:
- Search YouTube for a query and page through results
- Pull YouTube's search autocomplete suggestions for a phrase
- Fetch the comments on a video, page by page, and the replies under a comment
- Fetch a video's transcript when one is available
It reads public videos only. It does not post, reply, like or change anything on YouTube, and it cannot see private or members-only content.
Step 1: Pick the videos your buyers watch
Start with what your customer would search before buying. For a small bakery-supply shop, that might be "best stand mixer for bread dough" or "how to keep sourdough starter alive". Ask the assistant to search each phrase, pull the autocomplete suggestions for it, and list the top videos with their titles and channels.
Pick five to ten videos yourself from that list. Prefer recent videos with real discussion over the single most-viewed one. Leave out your own videos for now; you want to hear from people who are not already your audience.
Step 2: Read the comments and pull out questions
Ask the assistant to fetch the comments on each chosen video, including the replies under the most-discussed ones, and to pull out every comment that asks a question or describes a problem. Have it keep the exact wording and the video it came from.
Then ask it to group the questions by theme and count how many comments fall in each group. Counting matters. One loud comment is an anecdote; the same question under four different videos is a pattern.
A useful instruction is: "Quote the comments, do not paraphrase them, and do not add questions that are not in the comments." That keeps the list honest.
Step 3: Check what the videos already answer
Comments often ask something the video skipped. Ask the assistant to read each video's transcript and mark, for every question group, whether the video answered it. The gaps, questions people asked that the video did not answer, are the most useful part of the result. They are what a blog post, an FAQ entry or a product page section can cover better.
Step 4: Turn it into a short list
The output should fit on one screen. For example:
- Question group: in the commenters' words, with two or three quotes
- How often: number of comments, and across how many videos
- Answered in the videos: yes, partly or no
- Where we could answer it: a product page, an FAQ, a blog post or an email
You decide what to write. The assistant's job is to collect and sort, not to choose your content plan.
Step 5: Make it a monthly routine
Questions change with seasons and new products. Grok Bot can run this as a scheduled routine, for example on the first Monday of each month, using the same search phrases. It leaves the new list in chat and notes which question groups are new since last month. Nothing gets posted anywhere; you read it and act on what is worth acting on.
Limits
Comments are not a survey. People who comment are a self-selected group, and which comments you see first depends on how YouTube orders them, so counts tell you what is common in the comments, not what all your customers think. Some videos have comments turned off or no transcript, and those simply drop out of the list. Spam and jokes need filtering, so read the quotes before trusting a group.
Use the list as a starting point for your own judgment. If a question group is large and none of the videos answer it, that is a strong sign it is worth a page on your site.
If you also want to watch what is being talked about on X in your niche, see our guide to tracking what is trending on X.
Setting up research routines like this is part of our embedded AI partnership. Book a free session and we will pick your first search phrases and run the first list with you.
Related service: embedded AI partnership.
Get our notes on AI and how businesses run in your inbox: subscribe to The Beyond Brief.