Big Idea

The argument in one line.

Connecting a frontier reasoning model to a live YouTube keyword tool turns topic selection into a repeatable, evidence-backed process that works whether a channel has years of view data or none at all.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You run a YouTube channel for a business or coaching practice and are guessing at topics instead of testing them against real search data.
Your channel has been live for years without traction, or you're starting from zero and don't know where to point your first videos.
You work in a small or specific niche and worry there isn't enough search volume to justify making videos at all.

SKIP IF…

You already have a working keyword-research workflow and a channel with consistent view growth.
Your content depends on trends and personality rather than search-driven topics, like comedy or vlogging.

TL;DR

The full version, fast.

Picking the right YouTube topic is usually the difference between a channel that grows and one that stalls, and most creators solve it by feeling instead of data. This breakdown tests a new AI model against a live keyword-research tool on three real channels: a plateaued 12-year channel, a small business channel struggling for traction, and a niche channel with almost no search volume. A single dictated prompt, connected to real search and competitor data, produces ten ranked topics with search numbers attached. Wrapping the same request in a structured project with explicit rules for what a channel needs at each stage sharpens those same ideas into specific sequels and objection-by-objection variations instead of generic suggestions.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 00:51

01 · Cold open: the real bottleneck

Names topic and title selection, not scripting or camera skill, as the biggest reason business channels stall, and introduces ChatGPT's new Astra model as a candidate fix.

00:51 – 02:59

02 · What Astra is, and the "prompt engineering is dead" claim

Explains agentic AI (goal-driven, not instruction-driven) and addresses the objection that AI-assisted research is a race to the bottom.

02:59 – 03:50

03 · The setup: two struggling channels, one live data source

Introduces the two test channels, a 12-year plateaued channel and a brand-new one, and explains that vidIQ is the tool supplying real YouTube data.

03:50 – 04:57

04 · Sponsor break: vidIQ connects the model to real YouTube data

vidIQ sponsor read: a $1 trial offer and how to install the vidIQ-to-ChatGPT MCP connector.

04:57 – 05:15

05 · Configuring ChatGPT Astra's Work mode

Selecting the GPT-6 Astra model and an effort level inside ChatGPT's Work mode before running the first case study.

05:15 – 09:06

06 · Case study 1: a tattoo artist's first 10 video ideas

A dictated prompt about a tattoo-artist client kicks off a 17-minute vidIQ research pass that returns ten ranked topics with monthly search volume and competitor evidence attached.

09:06 – 11:24

07 · Case study 2, round 1: a sales coach's generic pass

A plateaued sales-coach channel gets a simple, minimally-guided prompt; the model spots a 35x view outlier and recommends a sequel plus nine more titles.

11:24 – 12:41

08 · Building the Expert Topic Research Kit project

Introduces a packaged ChatGPT project holding the host's own rules for what kind of topic a channel needs at each stage of its YouTube journey.

12:41 – 14:09

09 · Case study 2, round 2: the project sharpens the same lead into sequels

Re-running the same sales-coach request through the structured project turns the single sequel idea into a full set of objection-by-objection video variations.

14:09 – 16:40

10 · Case study 3: a niche merch consultant's evidence-backed topics

A much smaller, harder-to-search niche (branded client gifts) still yields several defensible topics by cross-referencing the client's best video against verified search phrases.

16:40 – 16:45

11 · Close: get the free kit

Direct-to-camera close pointing viewers to the free Expert Topic Research Kit and a related masterclass.

Atomic Insights

Lines worth screenshotting.

Picking the right topics and titles matters more than scripting or camera skill, and it's the step most struggling channels skip.
An agentic AI model works from a stated goal instead of a detailed instruction, shifting the skill from writing prompts to supplying a real data connection.
A one-off spike in views, 35 times a channel's normal average, is strong evidence for what to make more of, not a fluke to ignore.
Reusable rules packaged into a structured AI project produce sharper, more specific results than typing the same request freeform each time.
A single proven video format can be multiplied by mapping it against every close variant of the same situation, in this case every common sales objection.
Search volume alone doesn't make a topic good; pairing verified demand with an angle that already worked for a similar channel is what makes it defensible.
A counterintuitive fact, like a bulk discount raising total cost, creates a built-in hook because it contradicts what a viewer assumes is true.
For a small or niche channel, the fix for low search volume isn't abandoning research, it's cross-referencing your best-performing video against verified search phrases.
The most useful AI-generated topic suggestions explain not just a title, but exactly what should happen inside the video and why.
Letting a research pass run for 15 to 20 minutes against live keyword data produces meaningfully different output than a fast, single-shot answer.

Takeaway

How to pick topics with data, not feelings

TOPIC RESEARCH METHOD

Connecting a reasoning model to a real search-data tool turns topic research from a gut call into a ranked, evidence-backed list, whether a channel has years of data or none at all.

01Cold open: the real bottleneck

Topic and title selection, not scripting or camera skill, decides most of whether a video works.
A channel that guesses topics from feelings instead of data is the most common reason it never gains traction.

02What Astra is, and the "prompt engineering is dead" claim

An agentic AI model works from a stated goal rather than step-by-step instructions, shifting the needed skill from prompt-writing to supplying a real data connection.
Treat any claim that prompt engineering is now dead with caution; the model still needs a real data source to be useful, not just a bigger goal statement.

03The setup: two struggling channels, one live data source

A topic-research method should work the same way on a plateaued channel with years of data and a brand-new channel with none.
An AI model can only surface real patterns when it's connected to a tool with actual audience and competitor data behind it; without that it's guessing with better grammar.

06Case study 1: a tattoo artist's first 10 video ideas

A single prompt describing the business, the audience, and the goal can produce ten ranked video ideas, each backed by a monthly search-volume number.
Letting research run for 15 to 20 minutes against a live keyword tool turns a topic list from a hunch into something with search demand and competitor evidence behind each idea.
The strongest ideas pair a real search term with an angle that already worked for a similar, smaller channel.

07Case study 2, round 1: a sales coach's generic pass

Feeding a model a channel's own video performance lets it spot which past video was a genuine outlier and recommend building more videos in that exact format.
A generic prompt with almost no added rules can still produce reasonable, evidence-based ideas, which shows the data connection matters more than an elaborate prompt.

08Building the Expert Topic Research Kit project

Packaging your own rules for what a channel needs at each stage of growth into a reusable project turns one-off AI prompts into a repeatable system.
A structured project with explicit criteria produces sharper, more specific output than a single freeform request, even when the underlying facts fed to the model don't change.

09Case study 2, round 2: the project sharpens the same lead into sequels

With added rules, a model can turn one proven format into a systematic set of variations covering every major version of the same underlying situation.
Turning one winning topic into a deliberate set of parallel versions, one per common objection or variant, is a repeatable way to multiply it.

10Case study 3: a niche merch consultant's evidence-backed topics

For a niche with genuinely low search volume, cross-referencing what worked for your own best video against verified search phrases still turns up defensible topics.
A counterintuitive claim works as a video hook because it creates a small piece of drama around an otherwise ordinary business fact.
Good topic research doesn't stop at a title; it should also outline what the video should cover and why that angle matches a real buying decision.

Glossary

Terms worth knowing.

Agentic AI: An AI system given a goal rather than step-by-step instructions, which plans and executes the intermediate steps itself using connected tools.
MCP connector: A plug-in style connection between a chat model and an external tool, letting the model pull that tool's live data into a conversation.
Outlier score: A multiplier showing how far a single video's views exceed a channel's own typical performance, used to flag which past video is worth building a sequel around.
Effort level: A setting on some AI models that trades response speed and computing cost for how thoroughly the model reasons before answering.
vidIQ: A third-party YouTube analytics and keyword-research tool that surfaces search volume, competitor performance, and channel-level view data.

Resources

Things they pointed at.

03:50toolvidIQ ↗

16:30productExpert Topic Research Kit ↗

16:40productHow to Attract Unlimited Clients From YouTube (masterclass) ↗

Quotables

Lines you could clip.

00:15

“the thing that trips up my students more than anything else is just picking the right topics and titles that their customers care enough to watch”

names the core problem in one breath, works as a hook with zero setup→ TikTok hook↗ Tweet quote

07:51

“it worked for 17 minutes and 21 seconds, and that's a big brain working for that long”

concrete, slightly funny specificity about how long the AI worked→ IG reel cold open↗ Tweet quote

11:20

“these are pretty well thought out and they're based on evidence again”

sums up the entire value proposition in one clean line→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphor

ChatGPT's latest model Astra just solved the hardest part of YouTube, especially for business channels. So what do you think the hardest part of YouTube is? Is it the scripting or the editing or getting really good on camera?

Actually, the thing that trips up my students more than anything else is just picking the right topics and titles that their customers care enough to watch. That is the part of YouTube that's not only the hardest, but it's the part that if you can get right, you're like 80 % of the way there.

And ChatGPT 6, otherwise known as Astra, it came out last week, and right away it was blowing past every industry benchmark, even outdoing Claude Fable in terms of logic tasks. So I was like, I wonder if this can just take over this part of YouTube for people and tell them exactly what types of videos they should be making that are just going to guarantee them lots of views from the right viewers.

So I tried it, and honestly, I am kind of blown away by how good it is. really think we're there now. So in the rest of this video I'm just going to show you how you can use ChatGPT Astra's really big brain to work on your channel whether or not you've even started on YouTube yet.

It works either way. And this works especially well if you're the kind of channel where you're really niched down and there might not be a lot of search volume for what you're trying to make videos about. And even if there is search volume, you probably have a lot of competition out there.

So this really is just going to help you kind of find your openings and your opportunities. So I want to talk really quick about what Astra is and like what people are saying about it. So basically people are saying that it is the most agentic AI we've had yet, meaning you can just give it a goal rather than like telling it exactly what to do.

You give it a goal and then it's going to work out how to get there on its own. And people are also going so far as to say prompt engineering is now dead, right? You can just tell it what you want it to get to and it's going to figure all that out.

So you don't really have to be a master prompter anymore to get it to work. So that's a pretty big claim and we will get to that a little later on. And I do want to address something before we go any further.

So whenever I do a video like this that kind of talks about using AI to come up with topics or titles or help you out in any way, I always get the comments from people saying, well, isn't this just a race to the bottom, right? If AI is helping us do all this stuff, isn't it at all just going to feel the same? Is it going to be AI slop?

Here's the deal. You always have to do research anyway, right? If you're going to be starting on YouTube, the last thing you want to do is just go on feelings and like what you want to make videos about.

That's a recipe for disaster. And that's why most channels never really make it because they're making the wrong videos. So as long as we have to do research anyway, you can either do it with your own head and you can agonize over it for a long time.

And trust me, I've seen a lot of people do it that way. they kind of always fumble the ball. So this is just a much more scientific way.

It actually crunches the data. AI is way better. at coming at seeing patterns than we as human beings are especially now with astros so is it a race to the bottom not even close this is your chance to kind of rise up to the top if you do it right okay so what i'm planning on doing in this video is i'm going to put this to the test on two different channels that you guys submitted so one is a channel that's been struggling for about 12 years never got any traction and the other one is just a brand new channel i think there's a couple of videos from like three years ago but it's basically a clean slate so that's the set Now we need to understand how this is gonna actually work in practice.

So think of ChatGBT as the one who crunches all the data, right? It finds the patterns, it knows what to look for, it gives you your suggestions. But it has to get...

It's data from somewhere. It can't just make it up. That would be the worst thing ever.

So what we have to do is actually connect it to another tool called vidIQ. Now, vidIQ is actually connected into the back end of YouTube. So it can see, if you have a channel already, it can see all of your numbers.

It can see most of your competitors' numbers, their titles, what's in the videos. It can see all the transcripts. So this is where the data comes from.

This is how it's going to know. what is working, what's not working. So, and it did work out a deal for you guys.

You can start today for a dollar for your first month. So you can try it. You can cancel it after that if you want to.

But once you're set up inside of there, what you need to do is get the connector that you're going to use to connect to ChatGPT. It's really easy, I promise. You're just going to go down to your name down here and you're going to go up to MCP.

You're going to choose ChatGPT and then just make sure you're logged into ChatGPT. And if you are, you're just going to click this button right here and then it make sure you're all set up on ChatGPT.

It's a one -click install. So then I'm just going to click try in chat and the first thing we want to do is we want to make sure that we are choosing the right model. We want to be using Astra here.

So we're going to choose GPT -6 Astra and then you can choose your own effort level. I find the ultra at the very top that does run through the credits the fastest so maybe you want to back it down to max but if you've got the credits and you don't mind spending it might work a little better for you on the far end here all right so our first case study is a guy named delvirus he's a tattoo artist so i would just go here and i would click on the little microphone it's a lot easier just to dictate this rather than type everything out.

It lets you talk all about your business, all about your clients, and you're not gonna censor yourself the way that you would if you were actually typing it out. So I would just click here and I would say something like, I need your help in coming up with the first 10 YouTube videos for a tattoo artist. He wants to.

Talk about things like how to prep people for their first tattoos, things they need to know. Use the vidIQ connector, and I want you to really figure out what's working in the tattoo space right now on YouTube. What has search volume that we could go for, especially for a new channel, where they would actually be able to rank for these terms.

And then balance that out with titles that actually are a little spicy, that have something to say beyond just the search term. There and we can see the vidIQ connector the little logo is right there. So then at this point I would just let that run and We're gonna see what it comes back with now is a good time to talk about usage credit So you're basically going to be going through credits in two places one on chat GPT itself and two with vidIQ It's all kind of a credits based system.

So you pay for ChatGPT, you get a certain number of credits to use a week. And I won't lie, Astra does use them pretty quickly.

So this isn't the kind of thing you would want to do every day, but it's the kind of thing you could do once a week or you could do once a quarter just to give yourself like a bunch of topics. You can ask for as many as you want. And then when we get to the vidIQ side.

It depends what plan you have there too. I did go through a lot of credits yesterday when I was testing this. So you just want to be mindful of that.

But if it's like a once a month thing that you're doing, it's not the end of the world. It's a good use of your credits. So from here on out, we can just see what it's doing.

It's pulling up all of these searches within vidIQ. I'm just gonna let that run, and then we'll see what it comes back with. Okay, so it worked for 17 minutes and 21 seconds, and that's a big brain working for that long.

So we'll see what it comes back with. We can see all these searches it's done with vidIQ. We'll just get to the good stuff though in the end.

Let's see what we've got. So it says, I build this channel around helping people spot problems before they become permanent tattoos. Cool.

So it's a lot of background it's giving you too. So it's telling you what's on its mind, why it's kind of going where it's going. So you would want to read through all of that.

And it says, these are the 10 I'd launch with roughly in this order. All right, cool. So let's go.

We got... tattoo placement mistakes you only notice too late. And the search demand is the word tattoo placement.

It gets about 4 ,000 searches a month. And then it gives a little more detail here than how fine line tattoos age, the photos you need to see. Cool.

It's got a search demand behind that. And it's also got evidence of other videos that have done well. So it's kind of basing it on that too, which is really smart.

We've got another eight tattoo styles, one idea. What should you choose? How much do tattoos cost?

I priced the same idea three ways. So already, I would say these are pretty good. It's pretty solid choices with some good solid thinking behind each one.

But now let's go to our next case study that was submitted. So this is Christine Harrington. She's a sales coach.

And if we look at some of her videos, we can see her views are not... that great. And for something like this, a sales coach isn't that niched down.

There should be a little bit more views on videos like this. So we have 26 views, 78 views, 76 views. And she's been at it for 12 years.

So let's see if we can get her out of whatever rut she's in with some more thought out topics. Because if we look at these, we have the mental game of sales, how to win every time. Forcing a buyer to buy is not a sales strategy.

Here's what will happen. So I feel like some of these are just a little too kind of abstract or this one might be a little too specific and it might feel like it gives the whole thing away. So let's go ahead and we'll go back to ChatGPT.

And again, I'm just gonna dictate. I'm just gonna click on the microphone button. Help me make the best list of 10 YouTube videos for my friend, Christine.

She's a sales coach. She mostly deals with... people who are solopreneurs who haven't gotten into sales yet, they've never made sales for themselves, or they are sales professionals themselves who are kind of struggling with that.

I'm gonna paste in a link to her YouTube profile so you can look at her videos and you can see what's done well for her, what hasn't done well for her, and take a look at her competition and see what's working for them, see whatever keywords. that have some good search volume that she could use. But basically, she's been struggling with this for a while.

So we wanna make sure we get her a win and her next 10 videos are gonna perform higher than her average. And then I will just paste in her YouTube link. And again, we wanna make sure Astra Max is selected and we're gonna click on the button to send it off.

Okay, so it's done. It worked for 15 minutes, 44 seconds. And we can see all these.

searches again. So let's get to the good stuff, what it's recommending. So I would build these next 10 around specific moments where she loses a sale and show her handling those moments.

Okay. Her channel already supports this direction. Cool.

So it's giving all the cool background. And then here are the 10 titles. What to say when a prospect says I'm not interested, live role play.

So substantive sequel. So basically her one video, what to say when someone says I'm not interested, got way more views than her normal videos did.

It got 35 times her normal views. So it picked up on that and said, maybe we make a sequel to that. So that's awesome.

Stop being ghosted after sales calls.

based on this guy Ryan Dorns stopped being ghosted after a client sales call, which was an outlier for him as well. So these are pretty well thought out and they're based on evidence again. So I like all of these, but I have to wonder, we gave it so few guardrails and so few things to look for.

Personally, I know what goes into making a good YouTube topic. Like there's a bunch of different factors. So what I did is I made an actual ChatGPT project with all of my rules for the road, basically, of what kinds of videos you need at what stage in your YouTube journey.

So I packaged up that project and all of the instructions. It's called the Expert Topic Research Kit. So get that below and all the instructions are in there.

We're gonna use that now to see. with all of that extra context, can we get even better topics and titles picked up? And by better, I just mean, are they going to have a better chance of working at whatever stage these channels are in?

So let's go ahead and go back to ChatGPT and we'll go over to projects. So you'll install this on your own. All the instructions are there.

It's really easy. And I'm just gonna pick a YouTube topics test. That's the one I've been using.

And now I'm going to prompt it the same way. I'm just going to dictate it. I would do it the same way.

I would tell it about the person. The only difference now is it knows all the rules to come up with like the best possible topic. So I'll just do that again right here for Christine.

All right, so here's what it came back with this time. It says, I'd build three videos around a promising format on Christine's channel, a familiar prospect statement followed by exactly what to say back. So this is.

Going on her, you know what to say when someone says I'm not interested remember that was her her big hit on her channel the other version said like make a sequel to that essentially this is Even smarter because what it's saying to do is it's basically saying make different versions of that So what to say when someone says I need to think about it What to say when someone says it's too expensive What to say when someone says, just send me some information.

So these are like really common sales call objections and she's just gonna apply what worked to all of those, which super smart. And again, we've got all the context of like why it's suggesting these. So then we have how to write cold emails without begging for 15 minutes.

40 years in sales, how I'd build a pipeline from scratch. Stop getting ghosted, fix your sales follow -up process. So yeah, I would say this is a huge step up from what it was just giving us originally, but that's for a case where there's plenty of things to talk about, right?

There's a lot of salespeople out there. It's easy to make a lot of different videos for it. What if you were in a niche that is just way, way smaller and you're really struggling for topics that anyone would be searching for and anyone would, you know, have an appetite for.

So now I want to talk about one of my students, Barbara. She sells something way more specific and niche down. Basically, she consults with people how to come up with the best possible branded merch for their business as client gifts, to get clients, to keep clients, that kind of stuff.

And her topics have been a little scattered. So how to turn your brand into a product people actually want, the junk drawer test every merch buyer needs. So these probably aren't getting a lot of search volume as judged by her view counts at least.

So I think this might be a huge help for Barbara to finally find the right topics. So I just used the same project within ChatGPT and had a conversation with it about Barbara, what she does, who her clients are, and this is what I got back. So the first one, seven corporate gift ideas that don't look like corporate gifts.

I think that's great. And then the phrase corporate gift ideas. has verified search demand so that's awesome so again this is based on evidence it's based on research and data not just feelings the bulk discount that makes client gifts cost more hmm that's interesting right you wouldn't think a discount would make something cost more so it's got some like drama behind it which is good and it's also by the way saying what you could make the video about so it's not just like saying a topic in a title it's giving really good advice like what exactly might go in a video like this.

Then we've got company swag I'd pay for and what I'd leave behind. I love that. And then, yeah, company swag.

has verified search demand and represents a real purchase Barbara can help someone make. So these are really solid topics that it would have been really hard to come up with before we got this upgrade in ChatGBT. And if you want to get that expert research kit, it's for free right here.

So just click there and I'll send it right over to you. It's really the best way to make this work for you in literally less than 20 minutes. So get that and I'll see you in our next video.

The Hook

The bait, then the rug-pull.

The video opens by naming the one thing that quietly kills most business YouTube channels: not the script, not the camera work, but picking topics nobody is searching for. It then puts a new AI model to that exact test, live, on three real channels.

Frameworks

Named ideas worth stealing.

11:37concept

Expert Topic Research Kit

A packaged ChatGPT project holding the host's own rules for what kind of video topic a channel needs at each stage of its YouTube journey, so every future prompt inherits that judgment automatically.

Steal forany repeatable creative-decision prompt, like headlines, hooks, or offers, where the same judgment call gets made over and over

CTA Breakdown

How they asked for the click.

VERBAL ASK

16:30link

“get that expert research kit, it's for free right here... I'll see you in our next video”

Soft, single CTA delivered direct to camera in the final 15 seconds, pointing to the free kit and vidIQ trial links that were already dropped earlier, not a hard sell.

MENTIONED ON CAMERA

03:50toolvidIQ ↗

16:30productExpert Topic Research Kit ↗

16:40productHow to Attract Unlimited Clients From YouTube (masterclass) ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.