Create more engaging videos with simple, actionable scriptwriting tips.

This week, we took one of my most commonly used scriptwriting frameworks...

... and applied it to a 6m subscriber kids channel.

To be honest, this is not a niche I've worked in before.

But the results were very interesting.

And they reassured me that this scriptwriting framework is a no-brainer for pretty much any channel (including yours).

So... what's the framework?

Well, the most important part of any YouTube script is structure.

But deciding how to progress from A to B to C is the easy part.

It's how you signpost that structure that matters most.

Regardless of whether you're making an educational or entertainment video, the viewer needs to understand your video's structure in two key places:

• Hook - right at the outset, set the expectation of the video's structure.
• Setups - when a new segment begins, remind the viewer where in the structure they are so they don't lose track.

Now, I've spoken about this before.

But I'd never applied it to kids' entertainment.

Part of me wondered whether an 8-year-old would have the patience to sit through an extra 5-10 seconds of structural explanation.

But once we saw the retention (it stayed above the average for over half the video), not only does it seem like they do...

...but it turns out they were actively more engaged because they understood the video's structure better from the word "go".

Now I don't want to oversell it... this was just one video. We'll keep testing.

But it's an early signal that the exact same framework I've spent 5 years screaming about might truly be niche-agnostic.

i.e. it doesn't matter what niche, sub-niche, format or age-range you're making videos for...

...human brains need structure.

And spending an extra 5 seconds signposting that structure at the start...

...might buy you an extra 50 seconds of their attention later.

That's all for this week.

Speak soon,
George 👋

Losing your audience's trust rarely happens overnight.

So today, I want to show you the scriptwriting mistake you were told would grow your channel faster...

...but which gradually burns the trust you built up with your viewers.

And it comes down to this:

We all know negativity sells.

Negative headlines. Scary tweets. Comment section ragebait.

As human beings, we're drawn to negative things more than positive things.

It's called loss aversion - we're more scared to lose something than we are excited to gain something...

...so we instinctively pay more attention to the things that could impact us negatively to help us avoid them.

And this is exactly why most YouTube videos are framed through a negative lens.

• "Follow these tips, or your portfolio will drop by $10,000."
• "Start this diet, or you're 2x more likely to have a heart attack."
• "Read my newsletter, or your YouTube videos will bomb!" 😉

(Yes, of course, I do this too!)

There's nothing wrong with negative framing.

Until it becomes disingenuous. Or "fake".

"Get this wrong, and your plant will be dead in a week - and you'll have wasted $30 that could have been spent on your kid's clothes!"

The truth is: most YouTubers don't do this deliberately.

But they're so used to framing the stakes of their video through a negative lens that they start to prioritise this fear-based approach over all else.

So on your next video, when you're figuring out how to write the hook, or working on a segment setup...

...start by thinking of the negative frame you could use.

And if it feels authentic... stick with it. Because we know it's effective.

But if it doesn't... consider what the positive frame might look like.

❌ Negative: "Your messy inbox is quietly destroying your productivity." Maybe true! But it's a massive, overused claim that probably can't deliver on the magnitude of its promise.​

✅ Positive: "Clearing your inbox early frees up your mind to tackle the most important tasks without distraction." A more measured, attainable promise.

❌ Negative: "Get this wrong and your dog bolts into traffic, leaving you with a vet bill you can't cover." A real fear, but way too heavy for a video teaching dog training.​

✅ Positive: "Spend just five minutes in the garden practicing this, and your daily walk goes from a stressful to-do to a genuine joy for both of you." Framed around the ease of the practice and the positive outcome.

Because telling your viewers to PANIC in every single video starts to wear thin.

Eventually, you sound like the boy who cried wolf.

And if you're not being authentic, your viewers will start to lose trust.

Genuine negative frame > Genuine positive frame > Inauthentic negative frame

That's all for this week.

Speak soon,
George 👋

If you use AI to help you write YouTube scripts, but you aren't choosing the right AI model...

...the results will be terrible.

Here's why:

While building PlayWriteAI, I've been experimenting with a tonne of different AI models to create the best possible outputs.

And, as you may know, they're all good at different things.

If we just look at Anthropic (Claude)'s most common models:

• Opus - great at dealing with large amounts of information without forgetting or "hallucinating". Not great at creative writing.
• Sonnet - can't handle large amounts of information, but great at creative writing.
• Haiku - by today's frontier model standards... basically useless. But very cheap comparatively. I use it to "mark" the work of other models.

(And, of course, Fable is the new frontier model, but honestly, for writing, I would avoid it.)

OpenAI (ChatGPT)'s models also have their comparative strengths.

But most people don't even think about which model they're using.

They just assume the default model is probably good enough for whatever they need.

Not only is this not true (per my examples above), but these companies regularly default you to the less powerful models to save themselves cash.

So don't just use whatever model your AI defaults to. Pick the right model for the job.

• Using AI to create structure or analyse large amounts of research? Use Claude Opus or ChatGPT o-series.
• Using AI for creative writing? Use Claude Sonnet (or ChatGPT 5, but it's a bit meh by comparison).

The biggest hack? Simply ask the AI which model would be best for completing the task you've set it.

That's all for this week.

As always, if you have any questions, feel free to reply.

Speak soon,
George 👋

It's been almost two months since I last wrote this newsletter - I hope you're doing well!

The truth is, whenever I have a big project in the works, I prefer to put these on ice.

I thought I might be out of action for 2-3 weeks...

...but working on my scriptwriting app, PlayWriteAI, turned into an absolute monster of a project.

So while I'm not yet ready to return to business as usual here (i.e. newsletter every Friday), I want to share an important reflection I've had during this hiatus:

You can take a break from making content. It's fine 😆

It's beautiful watching YouTube work for you in the background.

Not only did I stop writing my newsletter two months ago, but I also stopped uploading to YouTube.

Yet over the last month, my channel has continued to pull in a continuous stream of viewers...

Two months since my last long-form upload, and views have remained almost 100% consistent.

Of course, the goal is consistency... but you don't have unlimited energy.

It's so easy to get sucked into a content hamster wheel, but it's important to remember that taking a break - especially one in service of a bigger goal - doesn't kill your channel overnight.

In fact, some of my favourite YouTubers deliberately work in seasons.

I'll see weekly uploads from them for two months, then nothing for the following two while they recharge.

Sometimes they semi-disappear for a year (usually if they're writing a book).

But I'm just as happy (if not happier) when they come back.

Energy management is a very real thing, and YouTube is hard work.

So, especially if you're working solo or with a tiny team (🙋🏼‍♂️), give yourself permission to switch your focus when the situation demands it.

Good content keeps growing on YouTube for years. (And it feels bloody amazing.)

Ready to meet your new AI scriptwriting buddy?

I want to make PlayWriteAI the world's best AI scriptwriting app.

I opened up 20 beta testing spots in June, which sold out in 3 minutes (!)

Very soon, it'll be ready for launch. You can join the waitlist if you're interested.​

That's all for this week.

I've learned so much about AI + scriptwriting over the last 3 months.

I'll be sharing more of those insights soon.

As always, if you have any questions, feel free to reply.

Speak soon,
George 👋

I filmed a Q&A last week so I could respond to as many of your recent scriptwriting questions as possible.

So while I'm "taking it easy" on my newsletter for the next few weeks to help me focus on my scriptwriting app...

...I've decided to share my response to three of those questions per week over the next few Fridays.

(Full Q&A releasing on my channel soon!)

How should I adjust my scripts based on my YouTube Studio analytics? I assume AVD is the main metric?

My response:

• I pay almost 0 attention to average view duration (AVD) or even average percentage viewed (APV) because these are so affected by the type/length of video.
• For example, longer videos tend to have higher AVD (yay!) but shorter APV (oh no!)
• But shorter videos tend to have lower AVD (oh no!) but higher APV (yay!).
• In both cases, one metric seems like a "win", but the other seems like a "loss". So looking at those metrics feels like you can't win.
• Retention graphs are, therefore, the main piece of data to pay attention to from a scriptwriting perspective. Pick 3 videos and review the spikes, dips and flat segments from each. I guarantee you'll notice several matching patterns (good and bad), and therefore something to improve in your next script.
• The caveat (and where AVD is useful) is when comparing videos of similar length. But, again, to actually diagnose why the video with the worst AVD was worse... you'd have to look at the retention graph.

My videos involve an element of "discovery", so how can I write my payoffs before I film when I don't know what's going to happen?

My response:

• This is very common on entertainment channels, and even educational channels like "review" channels that test products "live".
• In these cases, focus on writing just the hook and segment setups. These are the parts that rely on your actual curiosity as the YouTuber, and which help you plan the structure before you film.
• For example, in a video reviewing a pair of bluetooth headphones, one of my setups might be something like: "Next I wanted to test the [headphone name]'s noise cancelling. But honestly... I was going in with pretty low expectations after their demo at CES."
• Writing just the hook and setups before filming frees you up to act naturally during filming, while still establishing a clear structure to guide you.
• Then, edit brutally after filming to remove rambling, and write additional voiceover to tie things together.

Is scriptwriting dying?

My response:

• This question intrigued me, but despite the rapid advancement of AI, I don't think it is at all.
• I still get asked for writer recommendations as frequently as ever... that truly hasn't changed in the last 5 years.
• But something has changed. When I started doing this in 2021, I grew my personal brand quickly because there weren't many people doing YouTube-specific writing-for-hire. I also stood out because I had worked for Ali Abdaal.
• But today, the market has been absolutely flooded with YouTube scriptwriters (and Ali Abdaal has hired half the known universe). Therefore, the supply has increased more rapidly than the demand, which makes it harder to stand out.
• My advice - learn to use AI "responsibly". Ergo, don't entirely outsource your scriptwriting to AI and expect clients to love it. Work with the AI to write faster. Then share what you're learning publicly (YouTube, X, personal newsletter) so when new clients come around, they don't have to read 10-page scripts you wrote to assess your quality...
• ...they can simply scroll your socials and see that you think deeply and carefully about this stuff.

That's all for this week.

Speak soon,
George 👋

In this week's video, I shared the exact 5 documents you need to turn Claude from a forgetful assistant into a reliable scriptwriting partner.

Watch Here!

Don't have time to watch? Here's the short version...

The problem with using AI to write YouTube scripts isn't that it's not good enough yet - it's because you're not giving it enough context.

And the easiest way to do that is to create a dedicated "Project" for scriptwriting, and to upload the exact context it will need to help you brainstorm and write faster.

Having written my last 8 videos collaboratively with AI, and spent the last 3 months building my new AI scriptwriting app...

...I've distilled the most important pieces of context Claude needs into 5 simple documents.

​Click here to get instant access to the documents.​

Before all else, the very first step is to head over to the top left of Claude's main chat area, go to projects, and create a new project.

Done? Now, to the prompts...!

1. The Anti-Slop Prompt

Avoid the irritating default words and structures AI produces with my "Anti-Slop Prompt" - a detailed list with explicit instructions telling Claude to stop sounding like the dang robot he is.

2. The Training Doc

This training document teaches Claude the most important scriptwriting principles it'll need to help write retention-optimised scripts.

3. The "Process Prompts" Doc

This Process Prompts document contains the "instructions" you're going to add to your project, so it understands what its job is (to help you write YouTube scripts.

4. The Style Guide

The Process Prompts Doc also contains pre-loaded instructions to guide you through creating a comprehensive style guide in under 5 minutes.

5. The Audience Avatar

Likewise, the Process Prompts Doc contains similar instructions for the creation of an audience avatar.

So... ready to write your next script with Claude's help?

​Click here to get instant access to the documents.​

That's all for this week.

Speak soon,
George 👋

P.S. I'm building a complete AI Scriptwriting app to help you brainstorm and write full YouTube scripts half the time. Wanna test it early? Join the waitlist.

Today, we'll cover the small change that turned a dead video into my second fastest-growing.

How I revived a dead video (steal this):

After getting a good response to some AI content here on my newsletter, I decided to make two YouTube videos about Claude AI.

Video 1 started ok:

Video 2 bombed:

Although I had expected a slower start, seeing as I hadn't made any videos about AI before...

...I was a little disappointed.

But then, something changed.

After 10 days, I noticed Video 1 was starting to pick up views.

Not from browse, but from search.

And the search term that was driving the most traffic was:

"claude script writing"

So I went back to Video 2, which was still absolutely bombing.

And I changed its name from: AI Is Ruining Your Scripts. Here's Why.

To: How I Use Claude For YouTube Script Writing (2026)

(I deliberately kept the space between "script" and "writing", even though this is not how I would usually spell it.)

Almost immediately, Video 2 started picking up views.

And it literally hasn't stopped since:

Meanwhile, Video 1 is still flying:

These two videos are now growing my channel (and my email list) faster than any others.

So, while the goal is to grow your channel through browse, not search...

...you've got to take your opportunities where you can.

So for an easy win:

1. See if you have any videos growing consistently through search.
2. See which search terms are causing that growth (reach tab).
3. Re-title an underperforming video in the same content bucket using those search terms.

Let me know how you get on.

That's all for this week.

Any questions? You can to reply to this email and I'll get back to you.

Speak soon,
George 👋

Two weeks ago, I wrote about running little experiments on your channel.

This week, I wanted to show you a couple I'm running on my own to give you some additional inspiration.

1/ End-screens

One experiment I ran recently found that in any videos where I use "reveal cards" like this...

...my end-screen CTR was higher when I never turned around the final card.

The idea being that, by turning around that final card, it signals to the viewer the last piece of value has been delivered, so they're less likely to listen to my end-screen pitch.

But one person commented that it felt a bit annoying lol. And fair enough!

So right now I'm running a new end-screen experiment.

Instead of simply not revealing the card, I do reveal it, before quickly adding in a "secret" card - just as I'm starting my end-screen spiel - used to represent the lessons my viewers will learn if they click to watch the next video...

First attempt: 13.1% end-screen CTR vs my 12.5% average.

​Yesterday's video used the same technique, so I'll assess the data when I get it.

I usually like to test new things 2-3 times before deciding whether to adopt them or not.

But this wasn't the most surprising experiment I ran recently...

2/ Thumbnails

I recently shared a video of me generating this thumbnail in just a couple of minutes using 1of10's thumbnail generator.

​Get 1of10 with 20% OFF​

So I decided to test this entirely generated thumbnail against two in my usual style.

Now, normally, my thumbnail A/B tests come back pretty indecisive - literally 0.5%-1% difference.

But the 1of10-generated thumbnail produced some very surprising results... 👀

Test 1:

Test 2:

The 1of10 thumbnail absolutely smashed it.

Ok... it was only a 2/10, but I'll take it! 😉👇

I plan to test some other 1of10-generated thumbnails in the weeks to come, to see if this is a viable part of my strategy going forward.

Thanks to Richard and the team for supporting my newsletter since 2024!

Ready to start generating thumbnails from scratch and finding winning idea formats for your channel?

​Get 1of10 with 20% OFF​

That's all for this week.

Any questions about these experiments? Or any you've run recently?

You can to reply to this email and I'll get back to you.

Speak soon,
George 👋

The biggest problem with using AI to write scripts?

People can tell.

AI has those vocal "ticks" that completely wreck a first draft.

Things like:

• Negative parallelism: “It’s not X, it’s Y”
• Triple countdowns: "Not X. Not Y. Just Z."
• Self-posed rhetorical questions: "The result? Devastating."

I often start a YouTube script by brain-dumping my thoughts into a voice note...

...only to have the AI reformat my thinking into these irritating patterns.

So I asked Claude to do some deep research on the most complained-about AI-sounding patterns across the entire Internet...

...so I can train it to AVOID those phrases at all costs.

​The full list is yours for free - click here!​

Drop the list into your...

• Project Instructions (Claude)
• Custom Instructions (ChatGPT)
• Gem Custom Instructions (Gemeni)

...and let me know whether your AI output improves.

That's all for this week.

Any questions? You can to reply to this email and I'll get back to you.

Speak soon,
George 👋