AI can be an incredible copywriting partner.

But there's a big difference between using AI to make your writing better and asking AI to do all your thinking for you.

Used well, AI can help you critique your writing, identify weak spots, sharpen your message, repurpose content, and become a better copywriter.

Used poorly?

You can end up with generic AI-generated content—or worse, a self-created echo chamber where your AI keeps building on its own ideas and telling you everything is great.

So, how do you use AI for copywriting without stripping the soul from your content?

I follow four basic rules.

But first, let's talk about the mistake that inspired this post.

The AI copywriting mistake that made me think

Recently, one of our Max members made an interesting mistake:

She accidentally put an old post into Star Path—and to her surprise, it said:

"Great job, you nailed it."

That's not what she was expecting.

So she sent me a great question about copywriting with AI:

How do I double check that my AI isn't falling prey to "garbage in, garbage out"—or becoming a self-created echo chamber over time?

It's an excellent question.

Because AI can analyze, critique, brainstorm, edit, and write.

But AI is fundamentally working from patterns—and what you give it, what you ask it to do, and how you define success massively influence what you get back.

That's why good AI copywriting starts before you write the prompt.

Before using AI for copywriting, answer these 3 questions

When I'm co-creating with AI, I try to be laser-sharp about three things:

1. What's my goal?

What am I actually trying to accomplish with this piece of writing?

GOAL

2. What original information or thinking does the AI need?

Have I done enough prep to appropriately prime the AI?

FEEDER CONTENT

3. What does success look like?

How will I—or the AI—know whether the writing is actually good?

SUCCESS CRITERIA

If you can't answer those questions...

...you're probably not ready to pull in your AI copywriting partner.

At least, not without creating a lot of the slop I see on Instagram.

This is the framework I come back to again and again:

Goal → Feeder Content → Success Criteria

Now let's talk about how I actually put it into practice.

4 rules for using AI as a copywriter

1. Always feed AI original human content

When it comes to new ideas...

Humans > AI.

AI is exceptionally good at identifying and manipulating patterns in information. That's precisely why I don't want it doing all of my original thinking for me.

Instead of asking:

Write me a post about why entrepreneurs should trust themselves.

I'll usually write the post myself first.

Yes, you'll have to actually use your brain 🤦

This is a good thing.

Human brains are surprising.

We make strange connections. We have lived experiences. We notice tiny details. We change our minds. We say things in ways another person wouldn't have thought to say them.

And that's what makes written content interesting.

So give AI your stories, opinions, arguments, observations, research, examples, rough drafts, transcripts, customer language—whatever original material you have.

Then let AI help you make that material better.

Why human-generated content matters in the age of AI

There's also an interesting body of research emerging around what happens when AI-generated material continually feeds more AI.

In a 2024 paper published in Nature, researchers found that generative AI models trained recursively on model-generated data can experience "model collapse": over generations, models begin losing information about the original underlying data distribution and the quality and diversity of their outputs deteriorate.

The researchers specifically concluded that maintaining access to original, human-produced data is important. Their research concerns the training of AI models, rather than what happens inside an individual copywriting chat—but I think there's a useful lesson here for writers, too:

Don't let AI-generated material become a substitute for original human input.

The more AI I use, the more protective I become of the human part of my creative process.

2. Ask AI for an audit—not just an opinion

One of my favorite ways to use AI for copywriting is to ask for an audit.

An audit is different from:

What do you think of this?

That's vague.

And vague prompts often produce vague—and overly agreeable—answers.

Instead, make your ask highly specific.

Use words like:

Audit. Critique. Review. Identify weaknesses.

Invite criticism.

Even better, tell the AI exactly what it's auditing for.

For example:

Audit this email for conversion. My goal is to get readers to register for my free event. Tell me what's working, where I'm losing momentum, and give me the three highest-impact changes you'd make.

Now AI has a job.

And importantly, you're using AI as a copywriting coach rather than a ghostwriter.

That's one of the biggest differences I've found between AI that makes my writing more generic and AI that actually makes me a better writer.

3. Start fresh AI chats frequently

This is where the "echo chamber" problem comes in.

If you've been copywriting in one enormous AI thread for weeks or months, that conversation can accumulate:

• your original writing
• AI's suggestions
• your revisions
• AI's revisions of those revisions
• more AI suggestions
• more AI-generated copy based on earlier AI-generated copy

Eventually, it can become difficult to distinguish the original signal from all the accumulated noise.

So I prefer to save the useful context and frequently start fresh.

For a long-running project, I'll save curated information about the brand, audience, goals, voice, product, or project to Star Path's memory.

Then I'll open fresh conversations for individual tasks.

This gives me the best of both worlds:

the right context + a clean slate.

Now, this is practical copywriting advice - which is not the same thing as AI "model collapse." Research on model collapse looks at what happens when AI models themselves are continuously trained on synthetic data—not whether your personal AI conversation gets worse after a particular number of messages.

But the underlying principle holds:

Keep bringing fresh data back into the system.

Don't just endlessly feed AI more AI.

4. Tell AI what "great" actually means

Want to avoid hearing:

"Great job, you nailed it!"

Tell your AI what great means.

This is one of the most overlooked parts of AI prompting.

AI can't read your mind.

It doesn't already know whether you're optimizing an email for conversions, emotional intimacy, clarity, SEO, humor, authority, retention, or something else entirely.

So I tie my copywriting prompts to a specific goal.

For example:

"I want a high opt-in rate to my event from this email."

Or:

"I want to create a deep sense of emotional connection."

Or:

"I want to inspire people to take action on XYZ."

Or:

"I want to make it super clear what this product is, who it's for, and why their lives will be better if they have it."

Now there's a bar.

AI can't read your mind, and it doesn't know your priorities—so set expectations and hold HIGH standards.

For yourself and your AI co-creator.

The biggest AI writing mistake: not knowing when to stop

There's one more rule I think is arguably more important than all the others.

You need to know when to stop.

AI—as a pattern matcher—will always have something to say...for better or worse.

Ask it to audit your copy?

It can find another thing to change.

Ask for five ideas?

It can give you five more.

Ask whether your headline could be stronger?

Of course it could suggest another headline.

This means you're the only one who can discern when something is done.

That's part of the job AI can't take away from you.

If you're asking for audits, something in your content will always get audited.

If you're asking for ideas, AI will always give you another one.

So trust your gut—and know when to say:

Stop. It's done.

Otherwise you risk ending up in an echo chamber filled with garbage. 😉

10 AI copywriting prompts I actually recommend

The best AI copywriting prompts aren't:

"Write me an email about X."

Or:

"What do you think of this post?"

Instead, give AI original material + a goal + clear success criteria.

Here are some prompts you can copy and adapt.

AI copywriting prompt #1: Audit your writing

Audit this [email/post/landing page] for [GOAL]. Tell me what's working, what isn't, and give me 3 specific things I can do to improve it.

AI copywriting prompt #2: Fix something that feels off

I feel like this [email/post/page] is [too long / too dry / confusing / not landing with my intended audience]. Tell me how to improve it. Give me 3 actionable changes and explain why each matters.

AI copywriting prompt #3: Polish without replacing your voice

Here's a draft I wrote. Keeping my voice & flow, audit it for [GOAL] and tweak sections as needed. Preserve my original ideas, personality, and phrasing wherever possible.

AI copywriting prompt #4: Critique for conversion

Audit this copy for conversion. My goal is [DESIRED ACTION]. Identify anywhere I'm losing momentum, creating confusion, or failing to give the reader a compelling reason to act. Prioritize the 3 changes most likely to improve conversion.

AI copywriting prompt #5: Audit for emotional resonance

Audit this piece for emotional resonance with [AUDIENCE]. Where does it feel specific, human, and emotionally compelling? Where does it become generic or detached? Give me specific revisions without changing my overall voice.

AI copywriting prompt #6: Make something shorter

I like this draft, but I think it's too long. Identify anything repetitive, unnecessary, or over-explained. Tell me what you'd cut before rewriting anything.

This one is important.

I frequently ask AI what to cut before asking it to actually cut it.

That lets me make the editorial decision instead of automatically accepting every change.

AI copywriting prompt #7: Find what's unclear

Read this as someone who knows very little about [TOPIC/PRODUCT]. Identify anything that would be confusing, ambiguous, or require knowledge the reader may not have. Don't rewrite it yet—just show me where clarity breaks down.

AI copywriting prompt #8: Check whether the hook delivers

Read the opening of this piece, then tell me what you expect the rest of it to deliver. Now read the full piece. Does it fulfill that promise? If not, show me where the expectation and the actual content diverge.

AI copywriting prompt #9: Repurpose your original writing

Here's an [email/transcript/post] I wrote. Turn it into an SEO-optimized blog post. Keep my voice, ideas, examples, and overall flow, but add or restructure what's needed for search discoverability and readability.

Notice the sequence here.

Human creates → AI repurposes.

Not:

AI creates → AI repurposes → AI creates something else from the AI repurposing.

AI copywriting prompt #10: Ask AI to challenge you

Critique the core argument I'm making here. What assumptions am I making? What would a smart skeptic push back on? Where am I overstating the evidence or making a claim I haven't sufficiently supported? Don't rewrite my argument—help me make my thinking stronger.

This is one of my favorite ways to break out of the AI echo chamber.

A simple AI copywriting workflow

If you want an easy process to remember, here's mine:

1. Think. Start with your own ideas, expertise, stories, opinions, research, or observations.

2. Write. Create the rough version yourself. It doesn't have to be beautiful.

3. Define the goal. Tell AI what the piece is supposed to accomplish.

4. Audit. Ask AI to critique the writing against that goal.

5. Edit. Use the feedback you agree with. Ignore what you don't.

6. Stop. You—not the AI—decide when it's finished.

That's it.

AI isn't replacing my writing process.

It's sitting inside my writing process.

How I used AI to write this article

Here's the funny part.

The original email that became this article—and 99% of my content in general—was written without AI.

The place I used it?

To cut some stuff.

Because you're busy, and you don't need to read four sentences when two will suffice.

Here's the prompt I used:

Give this email a quick audit for efficacy. My goal is to educate people on how to use AI more effectively for copywriting, and I want them to walk away with simple & actionable things they can do immediately. I really like it, but I'm concerned it's a bit long. What can I cut?

Notice what's in there:

Goal: Educate people about using AI effectively for copywriting.

Success criteria: They should leave with simple, actionable things they can do immediately.

My own judgment: I already like the piece, but suspect it's too long.

AI's job: Audit and identify cuts.

That's what I mean by using AI as a copywriting partner rather than outsourcing your brain to it.

If you want to try this approach yourself, bring something you've already written into Star Path.

Give the Oracle a clear goal.

Ask for an audit.

Invite criticism.

Then decide for yourself which advice makes your work better.

Because the best AI copywriting still starts—and ends—with a human.