How to Use AI to Write Product Descriptions That Actually Convert

I spent three weeks rewriting product descriptions for a friend's Etsy shop using ChatGPT, and sales went up 23%. Then I tried the exact same approach for another friend's Shopify store, and nothing happened. Same prompts. Same process. Completely different results. That's when I realized most advice about AI product descriptions misses the thing that actually matters.

What I Was Wrong About

I assumed the hard part would be getting ChatGPT to write well. It wasn't. The AI can string together compelling sentences about basically anything. The hard part was figuring out what to tell it in the first place.

My first attempts were embarrassing. I'd paste in "Write a product description for a handmade ceramic mug" and get back this smooth, professional copy that sounded exactly like every other ceramic mug listing on the internet. Technically correct. Zero personality. The kind of description your eyes slide right past.

The Etsy shop that worked? We spent two hours on the phone before I wrote a single prompt. I learned that her customers were mostly women buying gifts for other women. That they cared about the story behind the piece. That "handmade" wasn't the selling point — it was that each mug had slight variations that made it feel personal. The Shopify store? I just jumped straight into prompting. I had product specs but no customer insight.

The AI didn't fail. I failed at giving it the right information.

The Prompt Structure That Actually Worked

After testing probably 50 different approaches, here's what consistently produced descriptions that felt different:

Start with the customer, not the product. I'd write something like: "The person buying this is a 35-year-old woman looking for a birthday gift for her best friend. She wants something that feels thoughtful, not generic. She's browsing on her phone during lunch break. She'll spend 15 seconds max reading this."

Then I'd add the emotional outcome: "After buying, she wants to feel like she found something special that her friend won't already have."

Only then would I describe the actual product. Physical details, materials, dimensions. But framed around what matters to that specific buyer.

The difference was immediate. Instead of "This handcrafted ceramic mug features a unique glaze pattern," the AI started producing things like "The glaze on this one caught me mid-scroll. Yours will have its own pattern — close to this, but not identical. Which honestly makes it better for gift-giving."

That second version sounds like a person. The first sounds like a catalog.

The Kick: Why Your Best-Selling Products Need Different Prompts

Here's what nobody tells you about AI product descriptions, and I only figured this out after comparing conversion data across about 40 different listings.

Your best-selling products and your slow-moving products need completely opposite prompting strategies.

For products that already sell well, the AI's job is to reduce friction. Keep it short. Answer the obvious questions. Don't oversell. These buyers are already interested — they just need reassurance. I'd prompt: "Write a 50-word description that answers: what is it, who's it for, what makes it worth the price. No hype."

For products that aren't moving, the AI's job is to reframe. Find a new angle. Question the assumed use case. I'd prompt: "This product isn't selling. Come up with three completely different types of people who might want this, and write a description for each one."

One slow-selling candle became a bestseller after ChatGPT suggested positioning it as a "work from home reset ritual" instead of just a scented candle. Same product. Different story. The AI didn't invent that idea from nothing — it combined what I told it about the scent profile with a bunch of patterns it had seen in successful product marketing. But I never would have thought of it myself.

The biggest waste of time I see is people using the same generic prompt for every product. That's not how this works.

What Still Doesn't Work

I've tried using AI to write descriptions for technical products with lots of specifications, and it's consistently bad at this. It either buries the specs in fluffy language or lists them in a way that doesn't help anyone make a decision.

For anything where buyers are comparing specific features across products — electronics, tools, anything with measurements that matter — I've found AI descriptions actually perform worse than simple spec lists with minimal prose. The AI wants to sell, but these buyers want to compare.

Also, ChatGPT has this annoying habit of adding urgency language. "Don't miss out" and "Order now while supplies last" kept sneaking into outputs even when I explicitly said not to include them. I ended up adding "NO urgency tactics, no scarcity language, no calls to action" to every single prompt. Tedious, but necessary.

The other thing: AI descriptions all start to sound the same after a while. There's a cadence to them. A certain smoothness that tips people off. I started deliberately asking for "one slightly awkward sentence" or "one detail that's oddly specific" just to break the pattern. A description that mentioned "the mug holds exactly 12 oz, which is slightly less than a standard Starbucks tall" outperformed the polished version by a lot. Imperfection signals authenticity.

The friend with the Shopify store eventually saw results, but only after we did what we should have done first — talked to five of her existing customers about why they bought. Turns out they cared about something we hadn't mentioned in any description. I still wonder how many sales we lost in those first three weeks.

Heads up: Some links in this post may be affiliate links. I only recommend tools I've personally tested. Opinions are entirely my own.