Claude vs ChatGPT for Copywriting: A Working Copywriter's Verdict

By Frederik Baun, conversion copywriter and founder of Words Sounds Visuals

Testing Claude vs ChatGPT for Copywriting on Three Real Briefs

I picked three jobs from my own queue instead of inventing a generic test, because a generic test produces a generic answer.

Brief one was a cold email opener for a coaching client’s re-engagement sequence, 40 words, matched to a specific brand voice document I’d already written for that client.

Brief two was a LinkedIn hook for a consultant, one line, built to stop a scroll.

Brief three was a landing page headline for a service business, anchored to a target keyword, under nine words.

I gave both models the identical brief text, the identical voice reference where one existed, and the identical word count.

Then I graded each draft the way I’d grade a junior copywriter’s first pass. Did it sound like the client, and did it need a full rewrite or a light edit?

Why Claude for Copywriting Won Two of Three Briefs

For the cold email opener, I pasted the client’s brand voice document into the brief. That document is dry, direct, and a little irreverent, and it names specific words the client won’t use.

Claude held that voice through the whole draft, tone consistent from the first line to the last.

For the landing page headline, I gave both models three approved examples from the client’s site instead of a written voice document.

Claude matched the phrasing style of those examples rather than defaulting to a template headline. The draft went to a shorter edit than the cold email one did.

Both wins had the same shape. Give Claude a reference (a voice document or a handful of approved lines) and it holds that reference across the length of the brief.

Where ChatGPT Still Wins

For the LinkedIn hook, I wasn’t holding a strict voice document, so I asked for 10 variants instead of one.

ChatGPT returned 10 distinct angles inside a single response, each built around a different hook mechanic:

• A direct question
• A contrarian claim
• A personal admission
• A number

On my run of this brief, Claude gave me four strong angles out of six.

If the job is fishing for one good line out of many rough ones, raw volume beats voice consistency. So I reached for ChatGPT on brief two, and I’ll reach for it again on the next brief shaped like it.

How Each Brief Came Back

Two out of three went to Claude, and the one that didn’t was the brief with no voice reference to hold onto in the first place.

Which Claude Model Is Best for Copywriting?

Sonnet 5, by default. That’s the model behind every draft in the test above, and it’s what I run for the vast majority of client copy: cold email openers, landing page headlines, LinkedIn hooks, full sequences.

I only reach for Opus 5 when the job is bigger than one draft: a new positioning angle from scratch, a full offer rebuild, a brief where I’m weighing five or more real trade-offs instead of one that’s already decided.

It’s slower and it costs more, so pulling it out for a 40-word cold email opener would be like hiring a strategist to proofread a tweet.

I don’t run client copy through Haiku. That tier is built for fast, cheap classification work. I keep it far away from anything a client will read.

So the honest answer to “which Claude model is best for copywriting” is Sonnet 5 by default, and Opus 5 only once the brief turns into a strategy problem instead of a drafting problem.

Claude vs ChatGPT for Email Subject Lines

Subject lines are the one job I run through both tools on purpose, and it’s the clearest version of the pattern from brief two above. A blank ChatGPT prompt beats a blank Claude prompt when the job is raw volume.

Packaging the formulas solves that, the same move as the skill section below.

I built a subject-line skill that returns 17 subject-and-preview-text pairs, one per proven formula, matched to whichever client’s voice document is loaded.

That’s the volume ChatGPT gives me on a blank prompt, except it’s tagged by formula and held to the brand instead of 10 variations on the same angle.

If you’re prompting from scratch with nothing packaged behind it, ChatGPT hands you more raw options faster. Once the formulas are packaged, Claude holds the brand voice across all 17 the same way it held the cold email opener’s voice document in the test above.

Why I Built a Copywriting Skill for Claude Instead of Prompting From Scratch

Pasting the same brand voice document into Claude by hand, every single time, gets old by the fourth client.

So I packaged it. A copywriting skill for Claude, in my case, is one file with the client’s voice rules, the words they ban, and four or five approved examples. Claude reads that file before it writes a single line.

I built mine out of the same voice-profile method I use on every client project. Claude reads it automatically instead of me pasting the same three paragraphs into every new chat.

That’s the difference between prompting Claude for copywriting once and building copywriting into how Claude works for you every time you open it.

I wrote about the full system behind it, the directives, the decision layer, and the gate that checks every draft, in how I use Claude Code for marketing.

What This Means for Your Copywriting Workflow

If you’re choosing one tool to standardize on, pick based on the shape of the job you run most often rather than a generic benchmark.

Voice-heavy work, client emails, landing pages, anything that has to sound like one specific business, points to Claude, especially once you’ve packaged a voice reference for it to read.

High-volume idea generation, 10 headline options, 20 subject lines to test against each other, points to ChatGPT.

I use both on purpose. The brief decides which one wins on a given day.

Whichever one drafts the brief, run the output through the same editing pass afterward. I’ve written separately about how to make AI writing sound more human, and that pass matters more than which model wrote the first draft.

Run your own version of this test before you settle on one tool for good. The brief that breaks the pattern for you isn’t the one that broke it for me.