Writing Better Marketing Copy with Claude
Most people ask Claude for copy once, get something generic, and give up. Treating it as a single request is the main reason the output feels flat. A one-line prompt like "write me an ad" gives Claude almost nothing to work with, so it fills the gaps with the safest, most generic choices it can make.
Copy improves through the same back-and-forth you'd have with an editor.
Tip
Treat the first draft as a starting point, not a final answer. The real improvement happens in the follow-up messages, not in trying to perfect the first prompt.
Why the brief matters more than the prompt wording
People spend a lot of effort trying to phrase the perfect one-shot prompt. That effort is usually better spent on the brief itself. Compare these two requests:
Weak brief: "Write ad copy for our new plan."
Strong brief: "Write ad copy for our Pro plan, targeting freelance designers who are currently on the free plan and hitting the project limit. Goal: get them to upgrade. Tone: confident, not pushy. Format: three headline options, each under 10 words, plus one 40-word body."
The weak version could describe almost any product. The strong version tells Claude exactly who's reading, what they already know, what needs to change in their mind, and what shape the output should take. That's the difference between copy that sounds like a template and copy that sounds like it was written for a specific reader.
A simple process that works
- 1
Start with a specific brief
Include the audience, the goal of the piece, and any constraints like length, channel, or tone. A vague prompt like "write me an ad" produces vague copy, no matter how good the underlying model is.
- 2
Give direct feedback
Respond the way you would to a person: "shorter," "less formal," "lead with the benefit, not the feature." Claude keeps the context of what you already discussed, so there's no need to start over.
- 3
Ask for variations
Having three headline options to compare is usually more useful than one Claude is confident about. Comparing options also makes it much easier to tell Claude what's actually working, since you're pointing at real examples instead of describing an abstract preference.
A good brief usually covers:
Who the audience is, specifically, not just "our customers"
What the piece needs to achieve, in terms of the reader's next action
- Length, channel, and tone constraints
Here's an example of a brief with enough detail for Claude to work from:
“
Write three headline options for a product launch email. Audience: existing customers who haven't upgraded in a year. Goal: get them to try the new plan. Tone: direct, not salesy. Keep each headline under 8 words.
”Giving feedback that actually improves the next draft
Vague feedback produces vague revisions. "Make it better" doesn't give Claude anything concrete to change, so the second draft often just rearranges the same ideas. Specific feedback, tied to an actual line of the draft, gets you somewhere.
Common mistake
Giving feedback like "I don't love it" without saying what specifically is off. Claude can't read your reaction, only your words. Point at the exact sentence or word that isn't landing.
A useful pattern is to name the problem, then name the direction. "This headline focuses on our feature, not their problem. Rewrite it starting from what's frustrating about the free plan's limits." That single sentence gives Claude a clear target instead of a vague mood to guess at.
Inside Claude Tutorial
Feedback like this gets easier with practice.
If this workflow made sense to you, you'll probably enjoy learning dozens of similar patterns inside the app, built the same way: a short lesson, then a real chance to practice it.
Adapting one piece of copy for multiple channels
Once a piece of copy is working, marketers often need it in several formats: a full email, a shorter social post, an ad headline. Rewriting each one from scratch wastes the work you already did getting the core message right.
Instead, ask Claude to adapt the same message across formats in one conversation, so the core idea stays consistent even as the length and tone shift per channel.
“
Here's the email copy above. Turn it into a 40-word LinkedIn post and a single ad headline under 8 words, keeping the same core message: try the new plan before the free-plan limits become a real problem.
”The goal isn't three unrelated pieces of copy. It's one message, resized for where it needs to live.
Knowing when to iterate versus when to start over
Iteration works when the core idea is right and the execution needs work: wrong tone, too long, weak opening line. It stops working when the core angle itself is wrong. No amount of line editing fixes a headline built around the wrong benefit.
Common mistake
Spending five rounds of feedback trying to fix a draft that was never going to work, instead of stepping back and asking for a different angle entirely. If the third round of feedback is still about the fundamental approach, not the wording, that's a sign to restart with a clearer brief instead of continuing to patch it.
A useful test: if your feedback is about word choice or length, keep iterating. If your feedback is about what the piece is even trying to say, go back to the brief.
“
This draft isn't working because it leads with our feature list instead of the customer's actual problem. Forget the draft above. New angle: start from the fact that switching tools is annoying, and make the case that this migration takes under ten minutes.
”Testing more than one approach for high-stakes copy
For anything with real reach, a homepage headline, a launch email, a paid ad, it's worth asking Claude for genuinely different angles, not just variations on the same idea. Three headlines that all lead with speed aren't really three options.
- One version that leads with the core benefit
One version that leads with the problem it solves
One version that leads with social proof or a specific number
“
Give me three genuinely different headline approaches for this launch, not three variations of the same idea: one that leads with the core benefit, one that leads with the problem we solve, and one that leads with a specific number or result.
”Continue reading
- Getting Started with Claude: the fundamentals behind this whole process, context, iteration, and reusable instructions, in case you want the underlying habits, not just the marketing-specific version.