The Product Description Problem
Most small business product descriptions read like they were written for a spreadsheet, not a customer.
Weight. Dimensions. Material. Color options. Available sizes.
These are specs, not reasons to buy. And they're costing you sales every single day.
The customer reading your product page already found you. They're interested. The description's job is to close the gap between "this looks interesting" and "I need this." Most descriptions fail at that completely.
What a Great Product Description Actually Does
Before we talk about AI, it's worth understanding what you're trying to write.
A great product description does four things:
AI can help you do all four — if you set it up correctly.
The Setup That Makes AI Work
The mistake most people make with AI product copy is prompting it with specs and expecting magic. "Write a description for a 16oz soy candle, vanilla scent, hand-poured" produces generic garbage.
The setup that works is building a brand voice document first. This is a written brief that tells the AI:
You give Claude this document once. Then every product description starts from the right place.
The Prompt Structure
Once your brand voice document exists, your prompt for any product looks like this:
"Using the brand voice guide below, write a product description for [product name]. The customer is [specific person]. The main benefit is [outcome, not feature]. The price is [X]. Keep it under 100 words. No bullet points — flowing copy only."
Then paste your brand voice guide.
The output will be dramatically better than anything you'd get from a generic prompt — and it'll actually sound like you.
The Real-Time Workflow
Here's how this looks in practice for a business with 20+ products:
First session — build the brand voice document. 30-45 minutes with Claude, working through your tone, your customer, your best existing copy. Save this document somewhere permanent.
Second session — batch write descriptions. Open Claude, paste the voice document, then feed it each product one at a time. For a 20-product catalog, this takes about 90 minutes. For a 5-product line, 20 minutes.
Edit what needs editing — usually 10-20% of the output. The rest is usable as-is.
What used to take a full day takes a morning.
What to Do With Variations
One underused application: AI is excellent at writing multiple versions of the same description for different contexts.
Your website description might be 80 words of flowing brand copy. Your Etsy listing needs different keywords. Your Instagram caption needs a hook and a CTA. Your email product feature needs to be 2 sentences.
Same product, four different pieces of copy — each optimized for where it's going. With AI and a good brand voice document, you can produce all four in 10 minutes.
The SEO Angle
Product descriptions also need to work for search. Not in a keyword-stuffed way — Google is too smart for that now — but by naturally including the terms people actually use when searching for what you sell.
Before writing any description, spend 5 minutes on Google. Search for your product. Look at the autocomplete suggestions. Look at the "People also ask" section. These are the exact phrases your customers use.
Feed those to Claude alongside your brief: "Naturally incorporate these search phrases without making the copy feel forced: [list]."
The result is copy that converts humans and gets found by search engines.
The One Thing That Makes It Actually Sound Like You
AI-generated product copy has a tell — it's often slightly too formal, slightly too complete, slightly too safe.
Your edit pass should look for those moments and rough them up a little. Add a contraction where there isn't one. Break a sentence into two. Cut the last sentence if it's wrapping up too neatly.
Real brand voice has texture. The AI gives you the structure. You bring the texture.
The goal isn't AI copy. The goal is your copy, written faster.