Copy Competition: Claude Opus 4.1 vs ChatGPT-5 Thinking
We Have New AI Models to Write With
The AI companies were super busy last week announcing all sorts of new products. The biggest announcement was OpenAI’s long awaited launch of ChatGPT-5. If only all product rollouts could have the same energy, anticipation, and marketing hype behind them. Not to be completely left behind, Anthropic announced its update of Claude, specifically Claude Opus 4.1.
Anthropic has positioned Claude Opus 4.1 as what they call a “drop-in upgrade” to Opus 4. What it really was, was a way to step on OpenAI’s launch which would come a few days later. But all’s fair in marketing, right. According to Anthropic, the new Claude Opus 4.1 improves performance in coding, agentic reasoning, and advanced writing tasks. Early benchmarks show Opus 4.1 outperforming its predecessor in technical writing as well as longer-form structured outputs.
A few days later, OpenAI launched ChatGPT-5 to wide fanfare. OpenAI calls it “a substantial advancement” in AI, touting improved intelligence, speed, and a lower rate of hallucinations compared to earlier models (we see you OpenAI o3 and o4-mini). They’re positioning GPT-5 as capable of delivering PhD-level expertise in writing.
Is The Written Output Really PhD Level?
So both models claim much improved writing. But which would you really rather have on your creative team? Let’s take a look.
One note before we get there. With both companies, you need to select the model you want. Claude defaults to Claude Sonnet 4 while ChatGPT-5 defaults to an “Auto” model that will select the underlying model for you based on your query (which isn’t necessarily the best one). For this evaluation, I selected Claude Opus 4.1 and ChatGPT-5 Thinking.
Existing copy and a simple prompt
I asked both models to rewrite some simple product copy for a summer dress. Here’s the prompt I used: Can you rewrite this product copy. Consider that one of the brand attributes is "joyful." Admittedly a very basic prompt.
Here's the existing copy: “Simple to wash in the airport sink and hang out to dry in the trade winds, this warm-weather midi-length dress includes prints inspired by island ecosystems. The lightweight TENCEL™ lyocell fabric dries quickly and easily loses wrinkles from travel for a laidback cover-up that evokes life in the tropics.”
The existing copy isn’t bad.
I especially like the beginning of the first sentence. It evokes an adventurous trip: If you wanted, you could wash it at the airport on your way to a tropical location. Simple to wash, easy to dry. Nice. Then you learn exactly what type of dress it is. I like the “inspired by island ecosystems” less because I don’t think wearing an ecosystem sounds all that sexy (though to each their own). The “easily loses wrinkles” phrase is kind of awkward, but the remainder of that sentence is fine. “Lightweight” works economically, but isn’t adding any additional shine.
How was Claude’s product copy rewrite?
Here are the results I got back from Claude Opus 4.1:
Splash it clean in any sink, let it dance dry in the breeze—this breezy midi celebrates spontaneous escapes with prints pulled straight from paradise. Our whisper-light TENCEL™ lyocell laughs off wrinkles and dries in a flash, so you're ready for the next adventure (or afternoon nap under the palms). Island-inspired prints bring the vacation vibes wherever you wander.
Thoughts on Claude’s work:
• The Claude rewrite definitely doesn’t beat the “simple to wash in the airport sink” line. “Splash it clean” requires a little more thinking on the reader’s part, and it offers a lot less for your imagination to feed on.
• You can’t really compare “hang it out to dry in the trade winds” to “let it dance dry” because, well, what does dance dry even mean? While you’re wearing it?
• “Whisper-light” might be a little better than simply “lightweight” depending on the rest of the copy, but I think “breezy” would have worked well. No one asked me.
• I don’t like “laughs off wrinkle” because I’ve seen that turn of phrase from Claude before, and it sounds forced.
• The word “adventure” doesn’t match with the laid back island vibe. But “afternoon nap under the palms” does, and it’s better than “evokes life in the tropics.” Give me an image, not just an idea.
• Claude’s last sentence is nice. Better than the ecosystem line of the original.
How Did ChatGPT-5 Do on The Same Assignment?
How did ChatGPT-5 Thinking perform? Here’s the result:
Slip into sunshine. This warm-weather, midi-length dress bursts with prints inspired by island ecosystems and an easygoing spirit. The lightweight TENCEL™ lyocell fabric rinses clean in an airport sink, shakes out wrinkles fast, and air-dries in a whisper—perfect for spontaneous beach walks and sunset dinners.
Thoughts on ChatGPT’s work:
• “Slip into sunshine” could be nice, but the line doesn’t get paid off. Does that just mean it’s a dress for warm weather? If I’m going to slip into sunshine, I want to look good and warm people around me. I don’t want the prints to be inspired and easygoing—I want to be those things.
• ChatGPT used the word “ecosystems,” which doesn’t work for me. I also didn’t like the use of “bursts.” I mean, when you’re in a dress, do you want to be bursting? I think not.
• This entire last sentence is crap, in my opinion. The rinses-shakes-air-dries construction is too workmanlike and box-checking for a vacation-vibing dress. Does the dress really shake out its own wrinkles or air dry in a “whisper.” The “perfect for…and” construction is too common, and the beach walks and sunset dinners feel cliché.
Bottom line: Go copywriter! Claude comes in second, and ChatGPT third.
Another Writing Assignment for Claude and ChatGPT
Let’s try another product copy rewrite, this one for a base layer hoodie. For this one I created a full brand-vibe document that directed the models. So rather than just one brand attribute to work with, the models had a document with multiple attributes, a brand archetype, brand copy examples, and more. While these short headline-teaser copy combinations don’t fully show off the benefit of using a brand-vibe document for generating product copy, longer copy made the benefit extremely clear.
Here’s Claude’s take on a short headline-teaser for a base layer hoodie:
Engineered for Peak Performance. Built for Every Adventure.
The ultimate body-mapped base layer that adapts to your every move, keeping you warm, dry, and ready for whatever the mountain throws your way.
Here’s ChatGPT’s version for the same piece of gear:
Freedom to Roam. Comfort to Stay.
Gear up with the Base Layer Quick Hoodie Top—made for pushing further, staying warmer, and loving every frosty moment.
Which Model Won This Round—Claude or ChatGPT?
Claude understood the assignment and executed it well. ChatGPT did neither. Yes, ChatGPT referenced the hoodie, but the product name appears immediately above this short description and is unnecessary.
• Claude’s quick headline is far superior. It tells me I get peak performance for every adventure. ChatGPT says I’m free to roam or stay home. Okay, sure, but couldn’t I just wear my pajamas for that?
• In the teaser description, Claude hits the most important feature—that it’s a body-mapped base layer—and that it’ll adapt to keep me warm, dry, and ready. ChatGPT’s version doesn’t really speak to me. You could interpret it as saying the hoodie will push further and stay warmer (presumably while I’m at home in my pajamas).
• Claude definitely leans into the idea of adventure, which was clearly requested in the vibe document, both in the heading and teaser description (whatever the mountain throws my way). Meanwhile, ChatGPT says I’ll love every frosty moment. Am I making a snowman? Drinking a beer? I’d rather be on the mountain.
Results of the Writing Competition Between Claude and ChatGPT
Claude wins, no contest
In both examples, but especially the second, Claude clearly wins in the product copy competition. But Claude didn’t just generate better copy. It fully completed the second assignment, which included five products and much longer product descriptions, with two options for each product. Meanwhile, ChatGPT reused the exact same headlines as well as copy descriptions over and over. For example, ChatGPT’s long description for every product started with “When the forecast says ‘stay home,’ the [product] says ‘watch this.’” That is a copywriter slacking on the job!
What’s more, ChatGPT wrote an average of 49 words per “long description,” while Claude averaged 170 and included tech specs and product features that were ignored by ChatGPT. Prime example of the PhD-level writer phoning it in.
More drop-in upgrades for both models will be coming soon. So think of this as one match in a long season.
