AI For Copywriting Megacourse Pt.2

In Part 1 I showed you how AI is fundamentally changing the copywriting game — not by replacing us, but by eliminating the “junior” phase entirely.

You need to wrap your head around the fact that you are no longer a “writer.”

You are an idea person. You are a director. You are this from the very beginning.

And as I explained in the last issue, this is nothing new.

To be honest, I haven’t been a “writer” for over a decade.

Sure, I have laid down more words on a daily basis than Stephen King well exceeding his 2,000-word-per-day habit.

But at no point have I been a “writer.”

The reason I say that is because 95% of my work was always research and idea generation — figuring out the perfect angle that was going to resonate with my potential buyer.

Crafting the persuasive “greased slide” I was going to push them down.

From getting their attention…

To hooking them in by emotionally resonating with them…

To getting them excited about the product / service and how it was going to transform their lives (making them see their future self)…

To making like and trust me by building authority and mirroring…

To overcoming their objections and making the offer a no-brainer through risk-reversal and value stacking.

That was where all the work was.

The writing part? Automatic.

Once I had my plan down, writing was a zombie-like action. Really only 5% of my job. I could do it half awake.

Because in direct response we write like we talk.

We’re not being “pretty” or descriptive.

The words that are flowing out of my fingers may as well just be a transcript of the imaginary conversation I’m having with my prospect at the imaginary bar we’re sitting at in my brain.

I mean good lord — I’m not writing the next great American novel here, I’m just trying to get you excited about butt plugs (or you know…whatever).

And I distinctly remember for YEARS having the following thought…

“I wish I could just plug a chord into my head like the Matrix and just vomit all of my ideas onto the page in an instant.”

Well guess what?

Now we CAN.

When doing research I used to have to Google (ugh).

Let’s go over what that process was like.

I mean let’s really break down the chain of thought…

• BASIC TASK: Sell A Supplement To women 35+ who want to lose weight

• ROLE: Direct response copywriter who is getting paid $10,000 and who will get paid significantly more with the more women I convince to buy this product

Okay — pretty basic.

But now comes the hard part…how do I persuade those women?

Well I have to start asking myself a series of questions.

• Who are they?

• What do they do for work?

• What are their fears?

• Their desires?

• What is the “quiet part” they won’t say out loud (their socially taboo desires like wanting people to find them attractive, for others to be jealous of them, to make their ex-husband / boyfriend jealous, to show off, to be noticed, to post selfies on social media again and make everyone surprised)

To find this out I’d look through a lot of social media posts. Facebook comments. Instagram. TikTok. Reddit. What are they saying? What are they complaining about? What are they talking about?

Then I’d start to put all of these findings into a document for reference.

But that’s just about them, I also have to ask questions like…

• Who are my competitors? (not just other supplements but Ozempic for example)

• What kinds of ads are my competitors running and what promises are they making?

• What are customers’ experiences with those competing products (negative or positive)

In this case I’d create another document and start providing swipes by using “Spy tools” like Atria to see competitors’ ads, see what kind of comments they were getting, check their Amazon reviews (what are the 5 star people saying and the 1 star people saying).

Of course I REALLY need to be familiar with the product

• I need to understand every individual ingredient in the product

• I need to read studies (that are proven) about how it helps reduce cravings, increase feelings of fullness, decrease gut dysbiosis and so on

• How do the ingredients work together in a way that makes it different from all thee other products on the shelf?

By far, this is the biggest shit-filled slop-fest to wade through. Countless generic listicles, blogs, SEO keyword salad, just to get tidbits of half-way reputable information I can put into yet another specialized document to make claims about what these ingredients do (all of which need to be backed up for the Legal and ad compliance department so the FDA or FTC doesn’t sue the bejeezus about of us).

The list of questions goes on and on so I’ll spare you for the moment.

(Trust me this is going somewhere).

Now let’s assume I answered all of the above questions for a product called Provitalean (because I did).

I now need to figure out an angle.

THAT TAKES EVEN MORE GOOGLING.

Now that I’ve got all of this research down I need to begin asking myself generic questions in the hopes that maybe they’ll lead to some bright idea.

• What countries have the lowest overweight / obesity rates?

• What do people in those countries eat?

• Do they exercise more or have better habits and that’s what it is, or can it be literally just the food?

• Is there a connection between the things they eat that keeps them naturally thin and the ingredients in my client’s product?

I mean look — it just goes on and on.

The process of Googling, weeding through the muck, and picking out bits and pieces to compile into a research document that all goes on to begin a “big idea” is arduous to say the least.

This was always the MAJORITY of the job. All of that research and time doing Google-fu just to have an IDEA.

Introducing: Deep Research & Social Media Scraping

Now remember that part about wanting to plug a chord into my head and “vomit” all my ideas / questions out onto the paper?

Now I can.

Let me take you through several deep research prompts and their results using ChatGPT and Claude. I’m going to share the full chats with you so you can see what I did.

Then we’re going to compile this deep research into actionable insight and prompts to create a full sales page, which we will complete by the end of this series.

I’m also going to provide you with 20 Deep Research Prompts you can use (courtesy of my internal team at one of the marketing companies I work with).

Let’s start with the research…

Audience Research / Social Media Scraping:

Here is the chat with the results: https://chatgpt.com/share/688092a5-be50-800e-a5d9-d9341cabd9cc

This is with o3-Pro Deep Research.

Then the follow up question asking GPT to create a psychographic profile was done selecting the o3 model for a complex, but faster result.

The result is something that — previously — would have taken me about 6 hours of total research to achieve and nowhere near the same depth.

Whenever I am writing copy, I always have to remind myself of what the UNDERLYING desire of the audience is. These are “primal desires.”

And as you can see, the psychographic profile helps tremendously with this.

There’s how she feels inside and what she wants to project to the world.

There’s always the REAL reason people do things and the reason they tell everybody else.

These are the parts of ourselves most people do not even admit are real or that exist.

In fact, many of us are so afraid of our true, “dark” shadow self desires we will not even admit internally to them.

For example “I want to be a successful surgeon so more women will want to fuck me.”

Ouch. Don’t say THAT out loud. Many won’t even admit it to themselves.

But very few — if any — men will strive to be the top of their field (whatever it is) without the background driver being to have the greatest amount of optionality with the most amount of beautiful women.

Think back to the famous old John Caples letter from 1927.

What is Caples REALLY selling here?

It’s a course that promises people that it can teach them how to play any instrument.

But what is he SELLING?

He is selling the feeling of being admired…

Of being impressive…

Of being desirable…

Of people envying you…

Whoever buys this product will rationalize it to themselves and other people that they “have always wanted to learn an instrument” maybe taking up a new, productive hobby. They love music — whatever.

The real reason is what John Caples was able to “tickle” inside their primal psychology without overtly pointing it out (because it makes people feel bad) and that is everyone will be jealous of how cool and desirable you are.

So when conducting audience research, this is what you’re trying to get to the heart of…

So that is what we did in this deep research.

The great thing about GPT and/or Claude (it also has a fantastic deep research function) is that you can then ask questions about the research, ask it to expand, maybe spitball some ideas.

Next up is REDDIT ANSWERS:

Whenever I want to find personal stories from the audience I’m researching and exploring this is my go-to.

Let’s say that I want to know what high-powered, type-a personality women are struggling with regarding stubborn weight gain over 40.

For example…

Reddit is a goldmine of insight.

Often I find that comments / threads from users can be their very own ads with even slight editing.

And this is a sentiment I’ve seen surface over and over again. “I used to be able to eat whatever I want, but now I’m gaining a bunch of weight” or “I’ve changed nothing, but now I’m gaining weight.”

So using deep research and Reddit answers we’ve honed in on a few potential audiences.

But really it seems to come down to women over 40 who lead busy, professional lives who are suddenly ballooning up despite making zero changes and they feel like their body is betraying them.

Now that I know who I want to targe and which “primal desires” to tickle.

The next question is — why are they gaining stubborn weight after 40 despite changing nothing and can the ingredients in Provitalean address this?

So we do more deep research.

Here is the result — 6,000+ words of research all cited and backed up.

https://chatgpt.com/share/68809ee6-c358-800e-94cc-00517e553914

So far in the span of about 30 minutes we have identified our audience, figured out their most primal desires, collected a list of stories from them about their struggles (that we can use) and have conducted in-depth research on how Provitalean’s ingredients is CLINICALLY shown to solve their problems with stubborn weight gain by improving their gut health.

This research alone would have taken me several days.

In Part 3 I’ll show you how we can take all of this research to create a project both in ChatGPT and Claude using “Context Engineering” but suffice it to say — you can clearly see the benefit of using AI for high level research.

Prompt Engineering 101

I want to start this section by saying there’s lots of “proven” ways people talk about prompt engineering. Some people are very structured about it and others — like me — are pretty loose and less structured.

In my day-to-day work with AI conducting deep research, I prompt by basically stream-of-thought dictating / delegation.

I have a series of ideas and starting points and I write them out to Claude or GPT the same way I would if I were giving the task to a junior copywriter or researcher.

I am as detailed as possible.

And as long as you are being as detailed as possible and you are clearly articulating well-thought out ideas and directions you will get better and better outputs.

But most people are too generic in their prompting.

1. Zero-Shot Prompts (Lazy Briefs)

This is when you give AI zero examples. Just "Write me a Facebook ad for jeans."

The result? Hot garbage.

I tested this with The Perfect Jean (a client). Asked for a simple 30-second script with zero context. The AI wrote: "Over 50,000 women have already made the switch!"

...For a men's jean company.

That's what zero-shot gets you. Use it to see how bad AI can be, then use those terrible outputs as "what NOT to do" examples.

2. One-Shot Prompts (Swipe & Deploy)

Give AI one winning example. It'll create variations that mirror the original's style, tone, and structure.

This is your go-to for iterating on winners. Got an ad that's crushing it? Feed it to AI with a one-shot prompt and get 20 variations in seconds.

But it's still limited—like having a junior who can only copy, not create.

3. Many-Shot Prompts (Patterning).

Feed AI 5-10 examples. Now it starts recognizing patterns across all your winners.

This is where AI begins to "get" your brand voice. It picks up on subtle patterns you might not even consciously notice.

Golden rule: 5-10 examples is the sweet spot. More than 20 is overkill.

4. Chain Of Thoughts (Strategic Brief)

This is where you stop treating AI like a vending machine and start treating it like a strategist.

Instead of "Write me a headline," you say:

"First, identify the core psychological pain point. Then, analyze how our current ads address this. Next, brainstorm new angles we haven't tried. Finally, write 10 headlines using the most promising angle."

You're teaching AI to think like you think.

5. Decision Tree (The A/B Test Machine)

Tell AI to create multiple versions for different scenarios, then pick the best.

"Write 5 different scripts—one for each of these pain points. Then analyze which would perform best for men 25-35 who are style-conscious but hate shopping.

6. Reverse Engineering (Prompt Hacking)

Feed AI your best performing ads and say: "Create a prompt that would generate ads like these."

Then use THAT prompt for all future ads.

It's like having AI psychoanalyze your winners and create the perfect brief for more.

The Best Prompts Have 4 Essential Components

1. Identity (Who AI Should Be)

Stop letting AI be generic. Give it a personality, expertise, and perspective.

Bad: "Write a Facebook ad"

Good: "You're a direct response copywriter who studied under Eugene Schwartz. You've written ads that generated $50M+ in revenue. You believe in entering the conversation already happening in the prospect's mind."

Pro tip: Tell AI to mimic, not create. "You're an expert at writing in the EXACT style of the examples provided" beats "Be creative" every time.

2. Task (The Specific Job)

Here's where most copywriters screw up. They ask AI to give them answers instead of helping them GET to answers.

Instead of: "Write 3 ad scripts"

Try: "Interview me about my product until you understand it better than I do. Then we'll write scripts together."

AI as co-creator > AI as servant.

3. Context (The Intelligence Briefing)

This is your unfair advantage.

Include:

• Winning ad examples (with explanations of WHY they won)

• Customer language (from reviews, Reddit, support tickets)

• Brand voice guidelines

• Psychological triggers that work for your audience

• Complete transcripts from YouTube videos about your topic

The more context you provide, the less "AI-sounding" your copy becomes.

4. Output Structure (The Deliverable Format)

Don't let AI choose how to structure its response. TELL it exactly what you want.

"Rank these from most to least likely to convert" "Group into 'Ready to Run,' 'Needs Editing,' and 'Trash'" "Give me 3 versions: Safe, Edgy, and Unhinged" "Rate each headline 1-10 and explain why"

Context Is King

Before you actually begin prompting to write any kind of copy you should do what copywriters have ALWAYS done before writing copy — and that’s to conduct all of the research first and identify an angle and unique mechanism.

This is why you should START with a series of deep research queries like I did at the beginning.

Then organize all that deep research into a series of documents, then upload it to a specialized Claude / GPT “Project” with specific instructions (identity, task, goal).

THEN and only then should you begin prompting to create copy.

That’s what we’re going to be going over in Part 3. I am going to take all of the deep research I have conducted for this mock project and organize it into context documents (you call this “Context Engineering” as some people do, back when I had to do this without AI I just called them my “ammo docs” or “ammunition” for writing my copy).

What AI Tools Are Best And Should You Pay?

My four favorites…

1. ChatGPT

2. Claude

3. Poppy

4. Reddit Answers

I also sometimes use Gemini.

The two I use the absolute most are…

• ChatGPT

• Claude

And yes I pay. I pay $300 a month for ChatGPT Pro and $100 a month for Claude.

But that’s because I conduct TONS of deep research and heavily use them so I need the best versions and I need tons of usage ability.

They more than pay for themselves and my job would not be possible without them at this point.

But for you? You’d probably be good with the $20/mo versions of both.

I NEVER use ChatGPT for writing.

But I use it for pretty much everything else research and analysis related.

By far, hands down, across every AI tool I’ve used nothing matches Claude’s copywriting abilities.

But Claude is not as good as GPT at research (although I’m increasingly liking their Deep Research function).

This could change in the future, but my sweet spot is conducting the lion’s share of research on GPT and building context documents.

Then creating a project in GPT and Claude with those reference context documents.

Then using Claude to work on copy while querying those context document for specific insight on GPT.

So for example to GPT, “In context document X I remember something about 80% of women quit Ozempic, can you explain why” then it will bring up the reason for example, then I’ll back to Claude and use that stat and be like “Now in this section, we’re going to overcome the objection of ‘why don’t I use Ozempic’ here’s the research on why most women quit and gain all the weight back.”

Anyway, we’ll get more into that later. And by the end of this series we’ll have a full sales page written.

20 Deep Research Prompts