Simulated customers made my copy stronger. Here’s how.

Welcome to this week's issue of Unpacking Meaning. If you received this from a friend and enjoy it, subscribe here.

What if your next round of customer research didn’t involve a single customer?

That’s the future being painted by a new wave of AI-native research platforms — ones that don’t just automate surveys or analyze reviews, but simulate entire societies of agents who think, talk, shop, and even complain like your customers.

I read a piece this week from a16z about this exact trend, and while it focused on product development and startup ops, I couldn’t help thinking:

What happens when we apply this same tech to crafting messaging and copy?

Context: The rise of synthetic research

The piece outlines how generative agents are making market research faster, cheaper, and always-on.

Instead of slow, one-and-done surveys, you can now query an AI-modeled population of, say, French Gen Z skincare shoppers. Watch them react to influencer content. See what they post in synthetic social feeds. Observe how they respond to new pricing models or packaging designs.

It’s wild. But most of the conversation so far is focused on what to build.

What if we asked:

How do we use this to better understand people — so we can speak to them more clearly, empathetically, and persuasively?

My take: Synthetic research shouldn’t be a shortcut, but a way to sharpen your gut.

Most marketers and copywriters I know don’t even use AI this way yet. They might summarize interviews or cluster survey data, but they rarely use AI to simulate human reactions, let alone pressure-test messaging with synthetic versions of their target audience.

I’ve started doing exactly that.

Sometimes I even simulate my client’s voice — their logic, their tone, their product convictions — and run my copy through that model to preempt feedback before I ever hit send.

It doesn’t replace the work. It makes it sharper.

But only if you use it intentionally.

That’s why I put together a quick framework to help you think through when synthetic research is actually useful in messaging work, how to apply it like a strategist (not just a prompt jockey), and where it might mislead you if you’re not careful.

Because if you’re just plugging prompts into GPT and hoping for clarity… you’re missing the point.

The 3 F’s of synthetic research (for messaging)

Here’s how I decide when and how to use synthetic research responsibly in my work and how you can, too.

☑️ 1. Fidelity

How close is this to real, human experience?

You get out what you put in. A raw GPT chat pretending to be your ICP is just mimicry. But an agent seeded with real interviews, VOC, CRM notes, behavior patterns can get you very close to resonance.

Quick note: In B2C, lower fidelity might be “good enough.” AI is trained on endless consumer content — social/Reddit posts, product reviews, lifestyle takes. But B2B requires depth. You’re speaking to decision-makers within decision-making units, which means you deal with layered incentives and hidden contexts. That data often lives behind closed doors.

☑️ 2. Function

What are you trying to decide?

Synthetic research is great for:

Prioritizing messaging angles
Spotting objections you might’ve missed
Testing emotional or logical resonance (with 70-80% accuracy) before putting copy in front of a human

It’s not great for:

Defining your narrative or category POV
Exploring deep emotional territory
Making high-risk calls without real-world input

Use it to accelerate thinking, not replace it.

☑️ 3. Feedback

Will this ever be tested outside a sandbox?

This is the part most people skip. Even if you’re using synthetic research well — with high fidelity and the right intent — you still need validation.

Run it by your sales team.

Test it in an A/B email.

Share it with a real customer.

If there’s no feedback loop, there’s no learning.

Bottom line: AI can absolutely help you craft sharper, more resonant messaging — if you treat it not as a replacement for empathy, but as a rehearsal space for it.

It’s a place to probe ideas, anticipate reactions, and test angles. But the real work still happens in the trenches: in the customer calls, the heatmaps, the validation, the rewrites.

That’s how synthetic research becomes a competitive edge. Not because it gives you perfect answers. But because it forces you to ask better questions.

Let me know: What would it take for you to trust a synthetic persona with shaping your next message?

DISCOVERY

🎙️ Episode 35 of The Message-Market Fit podcast is out!

I had an great chat with Hugo Alves, co-founder and Chief Product Officer at Synthetic Users. This is part 1 of a 2-part mini-series on synthetic research that you won't want to miss! Here's what you'll learn:

How AI can effectively simulate human behavior and decision-making
How large language models can role-play specific personas for product testing
The technology behind creating diverse synthetic users with realistic personalities
How to incorporate human biases and emotional responses into AI personas
The surprising effectiveness of synthetic research for product development
How different commercial AI models compare when simulating human behavior

And way way more.

Check it out here. And if you find it valuable, would you consider subscribing and leaving a rating? 🙏

🧠 New case study: How Growth Shop fixed their messaging to scale faster

When Mark Patchett came to us, Growth Shop had everything except a clear, unified message.

Their team knew what they did. But they couldn’t say it consistently. And that made hiring, onboarding, and scaling feel harder than it needed to be.

Here are a few things we did together that you can steal:

✅ Turn internal intuition (+ research) into a clear external message
✅ Build a Value Prop Canvas that actually gets used (for onboarding, AI prompts, partner handoffs)
✅ Translate fuzzy positioning into a pitch deck that sells

We didn't just write “copy", but reduced friction across the entire business.

📄 Read the full case study write up →

📽️ Or watch the video here →