Before you ask AI for better ad copy, read this

I’m working with a client who wanted to launch his own ads using a writing kit we’d put together for him.

The kit gives Claude the context it needs to write for his business: the messaging we’d developed, who the offer is for, the claims we can support, approved copy, and examples of my edits explaining what makes the writing work. It also includes prompts and a workflow for drafting and reviewing new copy.

So instead of starting every chat with a blank slate, he can build on the work we’ve already done.

He used it to create ads and get campaigns running. When we caught up this week, he told me they hadn’t brought in any leads.

Naturally, we wanted to understand why. Was it the message? The people seeing it? Where the ads sent them?

As we looked into it, I realized that AI had helped him do more than write the copy. It had helped choose what to target and where to send people. Decisions about the buyer and the buying journey had become part of building the campaign.

That’s where things got interesting for me. You can give AI a lot of good information about your business and still not give it enough direction about the particular buying decision you’re trying to influence.

Knowing your business isn’t the same as knowing this buyer’s situation

The writing kit wasn’t empty of strategy. It contained the positioning and messaging we’d worked on, and the ads retained useful parts of that work.

But knowing what you can say about your business doesn’t tell you what a particular ICP needs to read at a specific moment in their journey.

An experienced messaging strategist or copywriter should be able to help you make that choice. And there are a few different things we’re looking at when we do.

Audience: who is this for, in what situation? A job title can help you find people, but it doesn’t explain what they’re trying to get done. Two people with the same title might need completely different things from you.

Driver and trigger: what do they want, and why are they looking now? Wanting more control over a process is a driver. A recent failure that makes the current process impossible to ignore might be the trigger. One tells you what matters; the other helps explain why someone might finally act on it.

Awareness: what do they already understand? Someone learning what a category means needs a different conversation from someone comparing providers. If your ad assumes they’ve already chosen a type of solution, you risk skipping the very explanation they need before considering you.

Sophistication: what promises have they heard before? This isn’t about how smart the buyer is. But how familiar are they with the competing claims in your market? If everyone promises quality, speed and support, repeating those benefits gives an experienced buyer very little to work with. You need to explain what you do that makes the promise credible. Sophistication also influences what language you use in your copy.

These aren’t extra fields to fill in so your prompt looks more professional. They change which message you lead with, what you explain, what proof you use, and what you ask the reader to do next.

AI can help you think through all of them. But you have to know to ask for them.

What we found behind the clicks

In the campaign review, some of the visible searches matched the broad topic of the service without showing that the searcher wanted to buy it.

Others were searches for another company’s name. Our ads offered an alternative, but looking up a company doesn’t necessarily mean you’re looking to replace it. You might just be trying to get to its website.

The destinations added another question. Some clicks went to the homepage, some to a booking page, and some to pricing. Those pages asked people to make different decisions. A person still trying to understand the service would have much less explanation and context before a booking request than someone who went through the homepage. Context is everything, remember.

Even the measurement needed a closer look. We found that an inquiry and a completed booking could feed the same conversion event. Those are different steps. And the available analytics didn’t give us a reliable account of where the paid visitors stopped.

So, after digging deeper into why these ads didn’t work, we had concerns about the traffic, the message, the next step and our ability to observe what happened.

If we’d jumped straight to generating more variations, we’d have been changing the copy without knowing which problem we were trying to solve.

What I’m doing before we try again

I’m now working back through the journey, starting with one specific buying situation rather than a massive dump of related keywords.

That means choosing the reason someone would look for this service, deciding which part of the offer answers that need, and making sure the ad and landing page continue the same conversation. I’m also going deeper into how the service actually works, so the copy gives buyers a more specific and useful reason to believe the benefits instead of just repeating them.

Alongside that, we need to distinguish an inquiry from a booking and verify that we can follow the steps we want people to take. A clearer message won’t make an unclear result easier to interpret.

This is still work in progress. We haven’t run the revised approach, and I can’t tell you it’s going to win. But at least we’ll know what we’re trying, why, and what we should look at afterward.

For me, that’s the bigger lesson in this.

I want clients to use my writing kits. I want them to produce useful work without waiting for me to write every sentence. But I also need to help them see where simply producing a draft ends and deliberately planning a campaign begins.

As AI takes on more of the execution, those choices can become less visible. You ask it to help you launch some ads, and before long you have an audience, an angle, a set of headlines and a destination. Everything you need to launch, without necessarily understanding why those things belong together.

Before your next campaign, try explaining the buying situation and the journey in your own words. Who needs this, what triggered this need, what do they need to understand or believe, and why is this the right next step for them?

If you can’t explain that yet, that’s where I’d put the work before asking for another headline.

Discovery

Selling is part of understanding the product

PostHog’s Ryan McCrary makes a good case for more engineers moving into sales. He still gets to solve technical problems, but now he works directly with the people trying to buy and use the product. He can tell them what’s possible, what isn’t, and sometimes make the fix himself, without sending every question back through another team.

What stood out to me was his description of a buyer’s guard coming down when they realize the salesperson actually understands their work. I think good copy should do something similar. Not just use the right terminology, but show that you understand the situation well enough to help someone make progress through it. Selling is problem-solving, and writing the copy shouldn’t separate you from that part of the job.

​Read “More engineers should consider working in sales.”​

Resonance

— Joe Hudson, The New Inner Game: Your Unfair Advantage in the Age of AI