I Gave ChatGPT, Claude, and Gemini the Same Weak Sales Page. One of Them Wrote Copy That Made Marketers Stop Scrolling.
The test was simple: one mediocre sales page for a fictional productivity app, one identical prompt across all three AIs, zero tweaking. What came back was so different that three experienced copywriters I shared it with each picked a different winner — and then argued about it for 20 minutes. This article breaks down exactly what each AI did, where each one fell flat, and which model you should reach for depending on what your copy actually needs. If you write, sell, or market anything online, this comparison is going to change how you work.
The Prompt Was the Same. The Results Were Not Even Close.
The original sales page was deliberately bad in the most common ways. Generic headline. Vague benefits. A CTA that said "Get Started Today" — the copywriting equivalent of a shrug. The kind of page that exists on about 80% of SaaS websites right now.
The prompt I used across all three tools was identical: "Rewrite this sales page to be more persuasive. The target audience is busy professionals aged 30–45 who are skeptical of productivity apps. Lead with the pain point, not the product. Make the CTA specific and urgent."
ChatGPT (GPT-4o) came back fast — under 15 seconds — with a headline that read: "You're Not Lazy. Your System Is Broken." It restructured the page with a classic pain-agitate-solution framework, hit the skepticism angle directly, and rewrote the CTA to "Start Your First Distraction-Free Day — Free for 14 Days." Clean, competent, immediately usable.
Claude (Claude 3.5 Sonnet) took a different approach. Instead of jumping to a rewrite, it opened with a short note explaining its choices — which tone it was going for, why it led with a specific emotion, and where it deliberately slowed the copy down. The headline it produced was: "Every Productivity App Promised to Fix You. This One Doesn't." That line alone stopped people cold.
Gemini (Gemini 1.5 Pro) produced something polished but safer. The structure was solid, the language professional, but it felt like it was optimizing for "good enough" rather than "memorable." It swapped out the weak CTA but kept language like "streamline your workflow" — exactly the kind of phrase that makes a 38-year-old marketing director mentally check out.
What Each AI Actually Understands About Persuasion (This Is Where It Gets Interesting)
Most AI comparison articles stop at surface-level output. They show you the words and call it a day. But the more useful question is: what model of persuasion is each AI working from?
ChatGPT understands structure. It knows PAS (Problem-Agitate-Solution), AIDA (Attention-Interest-Desire-Action), and other frameworks at a deep level. If you ask it to rewrite copy using a specific framework — "Rewrite this using the before-after-bridge structure" — it executes cleanly and fast. It's the AI equivalent of a trained junior copywriter who has read every swipe file ever posted online.
Claude understands psychology. The reason "This One Doesn't" worked as a headline is because it anticipated the reader's internal objection before they could form it. That's not framework-following — that's objection-first thinking, and Claude does it instinctively. When I asked Claude why it made that choice, it explained: "Skeptical audiences shut down when they feel sold to. The goal was to get agreement before making any claims." That's a senior copywriter's instinct.
Gemini understands context and research. Where Gemini actually pulls ahead is when you give it more to work with — a competitor's page, an industry report, a list of real customer reviews. Feed it context and it synthesizes fast. On a raw rewrite prompt with no extra data, it plays it safe. Give it a brief that includes "Here are 10 one-star reviews of similar apps" and the output sharpens dramatically.
The mental model that changes how you use these tools: think of ChatGPT as your execution engine, Claude as your strategic thinking partner, and Gemini as your research-driven synthesizer. Use each for what it's actually good at, not interchangeably.
How to Run This Test on Your Own Copy in the Next 30 Minutes
You don't need a fictional productivity app. Use your actual sales page, landing page, or email — the one you've been meaning to fix for three months.
Step 1: Open all three in separate tabs — ChatGPT, Claude, and Gemini. Copy your existing copy into a document.
Step 2: Use this prompt in all three, word for word: "Rewrite this sales page. My audience is [describe them specifically — age, job, biggest frustration]. Lead with their pain point. Don't mention the product name until the second paragraph. Make every sentence either build desire or eliminate doubt. Rewrite the CTA to be specific and time-bound."
Step 3: Don't read the outputs immediately. Copy all three into a single Google Doc with the AI names hidden. Then read them cold, as if you're the customer.
Step 4: For whichever output wins your gut test, go back to that AI and run a second pass: "Now punch up the headline and the first sentence of each paragraph. Make them impossible to skip." This refinement step is where the gap between AI and human copy closes fast.
Step 5: Pull the strongest elements from each version. Claude's headline with ChatGPT's CTA and Gemini's supporting proof points is a genuinely viable combination — and it takes under 10 minutes to assemble once you have all three drafts in front of you.
The whole process, start to finish, runs about 25–30 minutes. That's faster than most people spend staring at a blinking cursor on a blank page.
The Part Most People Get Wrong
Most people treat AI copy like a first draft they still have to "make sound human." That framing is backwards, and it leads to output that is both worse than good AI copy and worse than good human copy.
The real mistake is giving AI a vague prompt and then editing the vagueness out manually. When you write "make this more persuasive," you're doing the thinking for the AI after the fact. The prompt is the strategy. If you can't describe your audience's specific fear, their biggest objection, and the one transformation your product delivers — in 2–3 sentences — no AI can compensate for that.
The second mistake is using only one tool and assuming the output is the ceiling. The marketers I showed this test to all made the same comment: "I didn't know Claude could do that." Most people default to ChatGPT because it's what they know. That's leaving real quality on the table, especially for anything emotionally driven — launch copy, testimonial framing, or any headline that needs to stop a scroll.
Finally, people compare AI outputs against each other instead of against the goal. The question isn't "which AI wrote better copy?" — it's "which version would make my specific reader click?" Run it past a real human in your target audience before you publish. AI writes the draft. Your audience validates it.
Key Takeaways
- ChatGPT for structure: It executes copywriting frameworks (PAS, AIDA, BAB) faster and cleaner than any other tool — ideal when you need a solid draft immediately.
- Claude for psychology: It leads with objections and reader emotions, making it the strongest choice for skeptical audiences or high-stakes conversion copy.
- Gemini for research-backed copy: It performs best when you give it real context — customer reviews, competitor pages, or industry data — to synthesize into persuasive language.
- The prompt is your strategy: Vague input produces vague output across all three tools — the more specifically you define your audience's pain and desired outcome, the better every AI performs.
- Combine, don't choose: The highest-converting output often pulls the headline from one AI, the CTA from another, and the supporting structure from a third — treat them as a team, not a competition.
What to Do Right Now
Open Claude and paste in any sales page or email you haven't been happy with. Use this exact prompt: "My audience is [specific description]. Rewrite this copy leading with their biggest fear, not my product's features. Make the headline something a skeptic would stop and read." Do that in the next 10 minutes, then run the same prompt in ChatGPT — and watch how differently two world-class AIs interpret the same brief.