We Fed the Same Terrible Product Description to ChatGPT, Claude, and Gemini — Only One Made Us Want to Buy
We took a real, genuinely lifeless product description — the kind you'd find on a mid-tier Amazon listing — and gave it to all three of the biggest AI writing tools on the planet. No extra instructions. Same prompt. Same dead copy. The results were so different it was almost uncomfortable. One version read like it was written by a bored intern. One was technically impressive but weirdly cold. And one made us stop and actually want to click "Add to Cart." If you write copy for a living, run a business, or use AI to produce content, what you're about to read will change which tool you reach for first.
The Raw Test: Same Prompt, Three Very Different Rewrites
The original copy we used was for a stainless steel water bottle. Here it is, word for word:
"This water bottle is made of stainless steel. It keeps drinks cold for 24 hours and hot for 12 hours. It has a leak-proof lid. It comes in multiple colors. It is durable and easy to clean."
That's the kind of copy that exists everywhere online. Technically accurate. Completely forgettable. We gave all three tools this exact prompt: "Rewrite this product description to be compelling and persuasive for an online store. Make someone actually want to buy it."
ChatGPT (GPT-4o) came back fast with something punchy. It opened with "Stay hydrated on your terms" and hit all the features, but framed them as benefits — 24 hours cold became cold enough to last your entire workday and then some. It had energy. It used second-person ("your morning coffee," "your weekend hike"). It felt like it was written for a real human.
Claude (Claude 3.5 Sonnet) took a slightly different angle. It wrote with more restraint — cleaner sentences, no exclamation points, a quieter confidence. The opening was "Some bottles just hold water. This one holds up." It was genuinely elegant. But it also felt like it was written for a slightly older, more considered buyer — someone who'd sigh at a flashy headline.
Gemini (Gemini 1.5 Pro) gave us something that read like a solid first draft. It improved on the original, no question. But it leaned heavily on adjectives — premium, superior, exceptional — without grounding them in specifics. It told us the bottle was great instead of showing us why.
The Hidden Difference: How Each Tool Thinks About Persuasion
Most comparisons stop at "which one sounds better." That's the surface level. The more important question is why they write differently — and once you understand that, you can use each one exactly right.
ChatGPT is optimized for engagement and momentum. It's been trained on an enormous range of marketing copy, blog posts, and consumer content. It naturally reaches for second-person voice, benefit-driven language, and a rhythm that moves the reader forward. When you're writing for a general online audience — e-commerce, landing pages, social captions — this is its home turf.
Claude is optimized for clarity and trust. Anthropic built Claude with a focus on being genuinely helpful rather than flashy. This shows in the writing. Claude resists hype. It won't call something "game-changing" unless it actually is. For brands that sell on credibility — SaaS products, high-ticket services, B2B — this restraint is actually a superpower. A Claude-written description feels like a recommendation from someone smart, not a sales pitch from someone desperate.
Gemini defaults to descriptive rather than persuasive. Google's model is extraordinary at research, summarization, and factual writing. But "persuasion" requires a kind of emotional logic that Gemini doesn't yet nail consistently. It knows what to say about a product, but it doesn't always know why someone should care. It produces copy that would pass a quality check but rarely stops anyone mid-scroll.
Here's the mental model that changes everything: think of ChatGPT as your copywriter, Claude as your editor, and Gemini as your researcher. They're not competing for the same job.
How to Use All Three Together to Write Copy That Actually Converts
You don't have to pick just one. In fact, the best AI-assisted copy workflow uses all three in sequence — and it takes less than 20 minutes.
Step 1: Start with ChatGPT for your first draft. Give it your raw product info and use this prompt: "Write a compelling product description for [product]. Focus on benefits over features. Use second-person voice. Make the reader feel like this product solves a specific problem in their day." You want momentum and energy here. Don't edit yet — just get something that has a pulse.
Step 2: Paste that draft into Claude and ask it to sharpen it. Use this prompt: "Edit this copy for clarity and credibility. Remove any hype or vague claims. Make sure every sentence earns its place. Keep the energy but cut anything that sounds like an ad." Claude will strip the fluff and tighten the logic. What comes back will be more trustworthy — and trust is what converts cold traffic.
Step 3: Run the final version through Gemini with a fact-check lens. Prompt: "Review this product description. Flag any claims that seem unsubstantiated or vague. Suggest specific, concrete details that could replace generic adjectives like 'premium' or 'high quality'." Gemini is excellent at catching weak spots. It'll point you toward specificity — and specific copy outperforms vague copy every single time.
Step 4: Paste the Gemini feedback back into ChatGPT or Claude for one final pass that incorporates those specifics. This whole loop — draft, sharpen, verify, finalize — is repeatable for any product in any category.
This isn't just a workflow. It's a system that uses each tool for what it's actually good at, rather than forcing all three to do the same job.
The Part Most People Get Wrong
Most people treat AI rewriting as a single-step task. They paste in bad copy, hit generate, skim the output, and post it. That's wrong — and it's why so much AI-assisted copy still sounds flat.
The mistake isn't using AI. The mistake is using one AI, once, with no clear brief. A vague prompt produces vague output. If you tell ChatGPT to "make this better," it will — barely. But if you tell it "rewrite this for a 35-year-old professional who commutes by bike and cares about sustainability," you'll get copy that feels written for a real person.
The second mistake is treating the AI's first output as the final version. Every professional copywriter rewrites at least three times. AI doesn't change that — it just makes each pass faster. The first output is raw material, not a finished product.
The third mistake is ignoring tone-matching. Claude's restrained, intelligent voice is wrong for a fast-fashion brand targeting 22-year-olds. ChatGPT's punchy energy can feel too aggressive for a premium wellness brand. Before you pick your tool, decide what emotional register your brand lives in — then choose accordingly.
Key Takeaways
- ChatGPT for drafts: Best at generating momentum-driven, benefit-first copy that engages general consumer audiences fast.
- Claude for refinement: Strips hype and builds trust — the right tool for high-ticket, B2B, or credibility-sensitive brands.
- Gemini for specificity: Strong at flagging vague claims and surfacing concrete details that make copy more believable.
- The three-tool workflow: Using ChatGPT → Claude → Gemini in sequence produces copy that is energetic, credible, and specific — all at once.
- Prompts determine output quality: A brief that names the target customer, their problem, and the desired tone will outperform any generic "make this better" instruction every single time.
What to Do Right Now
Open ChatGPT in one tab, Claude in another, and grab the worst product or service description you've written in the last six months. Run it through the three-step workflow above — start with this ChatGPT prompt: "Rewrite this as compelling copy for [your audience]. Lead with the biggest benefit. Use second-person voice. Cut anything that doesn't move the reader closer to buying." Do the full loop. It takes 15 minutes and the difference in the output will make you wonder what you were doing before.