AI Product Descriptions: How to Write Them Step by Step
Short answer Start with facts, not a prompt. Put each product's specs, materials, sizes and customer questions in one sheet, write a short brief on tone and rules, and test on five products. Then fact-check every draft against the sheet, add what only you know, label AI text in your Google feed, and publish.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callThe workflow in seven steps
| Step | What you do | What you end with |
|---|---|---|
| 1. Gather facts | One row per product: specs, materials, sizes, care, who it's for, top customer questions | A product sheet |
| 2. Write a brief | Tone, length, structure, words to avoid, claims never to make | One page you reuse |
| 3. Pick where to generate | Your store's built-in tool, a chatbot, or a bulk tool | A place to work |
| 4. Test on five products | Run the brief, fix it, run again | A brief that works |
| 5. Fact-check | Compare every draft line to the sheet | Drafts with no invented details |
| 6. Add what only you know | Fit notes, use cases, customer feedback | Copy no other store has |
| 7. Format and publish | Store page, Google Shopping feed, marketplaces | Live descriptions, each in the right format |
Shopify's own advice for its AI tools sums up steps 2 to 5:
"You should give clear instructions, add context about your store, refine results through conversation, and review changes before you apply them." (Shopify Help Center: AI-powered tools)
Step 1: Gather the product facts
The AI knows nothing about your product except what you give it. Google explains why that matters:
"Keep in mind that generative models don't retrieve facts, but predict a likely sequence of words based on their training data." (Google Search Central)
Give it only a name, and it fills the gaps with what similar products usually have. That's how a cotton shirt becomes "moisture-wicking."
Build a sheet with one row per product. Useful columns:
- Name and variants: colors, sizes, flavors, scents.
- Hard specs: dimensions, weight, materials, capacity, compatibility.
- Care and use: washing, assembly, power, shelf life.
- Who it's for: the customer and the job the product does for them.
- Proof you hold: certifications, test results, warranty terms. Only list what you can show on paper.
- Customer questions: the questions that keep coming up in emails, reviews and returns.
Google's Shopping spec lists more to capture: shape, pattern, texture, intended age range and design (Google Merchant Center Help).
Step 2: Write a one-page brief
The brief is the instruction you paste in with every product, or save in your tool. Keep it short:
- Customer: "Busy parents buying for kids aged 3 to 6."
- Tone: two or three words, plus one description you like as a model.
- Length and structure: for example, a two-sentence opening, then four bullet points, then care notes.
- Order: the most important facts first. Google advises putting the key details in the first 160 to 500 characters, because shoppers must click to see the rest (Google Merchant Center Help).
- Never write: health, safety or environmental claims not in the sheet; comparisons with other brands; prices or shipping promises.
- Words to avoid: the stock phrases you're tired of seeing.
Rule 5 is the one that protects you. In the US, the FTC says:
"Advertising must be truthful and non-deceptive;" "Advertisers must have evidence to back up their claims" (FTC: Advertising FAQ's)
In Canada, the Competition Act covers the same ground:
"All representations, in any form whatsoever, that are false or misleading in a material respect are subject to the Act." (Competition Bureau Canada)
The law looks at the claim, not who wrote it. If the AI writes "waterproof" and the product isn't, the seller made the claim.
Step 3: Pick where to generate
Three common places, each with a different trade-off:
| Where | Good for | Watch out for |
|---|---|---|
| Your store's built-in tool | A few products a month; text lands in the product page | Availability varies by plan |
| A general chatbot | Small catalogs; full control of the prompt | Copy and paste each result; re-paste the brief each session |
| A bulk tool that reads a spreadsheet | Hundreds of products at once | Errors repeat across every row if the brief is wrong |
On Shopify, Sidekick works inside the pages you edit, product descriptions included: click its icon and it writes the text in place (Shopify Help Center). Whatever tool you use, the sheet and the brief are what make the drafts good.
Step 4: Test on five products first
Never run a whole catalog on an untested brief. Pick five products that differ: your best seller, a technical item, a cheap item, one with variants, and one with few facts.
For each draft, ask: did it invent facts, does it sound like your shop, did it follow the structure, and does it answer the top customer question?
Fix the brief, not the drafts, and run the five again until they pass.
Step 5: Fact-check every draft
Compare each draft line by line with the product sheet. Google is direct about this:
"It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." (Google Search Central)
What AI gets wrong most often in product copy:
- Numbers: sizes, weights, capacities, counts in a pack.
- Materials: "solid wood" when it's veneer, "leather" when it's faux.
- Claims: "hypoallergenic," "non-toxic," "eco-friendly," "BPA-free," "clinically proven."
- Compatibility: "fits all models," "works with any phone."
- Promises: warranty length, delivery speed, "lifetime."
Delete any line you can't match to the sheet or to a document you hold.
Step 6: Add what only you know
A clean AI draft from the same supplier facts reads much like a competitor's AI draft from the same facts. What sets yours apart is what you learned from selling the product:
- Fit notes: "runs small; most customers size up."
- Real use: "holds a 15-inch laptop and a charger, not much more."
- The answer to the question customers ask before buying.
This matters for search too. Google warns that using AI "to generate many pages without adding value for users may violate Google's spam policy on scaled content abuse" (Google Search Central). Many products sharing the same supplier text is a related problem; see duplicate content.
Step 7: Format for each place you publish
One product often needs more than one version.
Your store's product page. Lead with the main benefit and key facts, then scannable bullets, then care and specs. For what else helps product pages get found, see ecommerce SEO.
Google Shopping (Merchant Center feed). Google has a separate field for AI text:
"For descriptions created using generative AI, use the structured description [structured_description] attribute." (Google Merchant Center Help)
You mark that field as AI-made (the value is called trained_algorithmic_media). Other feed rules from the same page:
| Rule | Detail |
|---|---|
| Length | 1 to 5,000 characters; longer text is cut |
| First lines | Key details in the first 160 to 500 characters |
| Describe only this product | No accessories, other products, business history or policies |
| No comparisons | Nothing like "even better than X" |
| No links | Only the product link field points to your page |
| No gimmicks | No all caps, symbols or promotional text |
If your store app sends the feed for you, check how it handles the AI label before you publish AI copy.
Marketplaces. Each one has its own style rules and character limits. Read them in your seller account before you format.
Example
A pottery studio with 60 mugs and bowls on Shopify
She builds a sheet in an afternoon: size, capacity, glaze, dishwasher and microwave notes. Her brief says "warm, plain, no 'artisanal'." On the test run, the AI calls two glazes "food-safe certified," which she has no paper for. She adds "no safety claims" to the brief, reruns, and spends about 5 minutes checking each draft.
A pet supply store with 800 products from a supplier spreadsheet
The supplier text appears on dozens of other sites. The owner adds two columns to the spreadsheet: "staff tip" and "top customer question." A bulk tool drafts all 800 overnight. Checking takes about 2 minutes each, roughly 27 hours spread over a month. AI-written descriptions go to Google Shopping in the structured description field, labeled as AI.
A Montreal kitchenware shop selling in English and French
The owner keeps one fact sheet and one brief per language. Each French draft is checked against the same sheet as the English one, so a size or material error can't slip into only one version.
FAQ
What should I give the AI to write a product description?
The product's real specs, materials, sizes, care, who it's for, and the questions customers ask. A name alone produces guesses.
Do I have to say a description was written by AI?
Google Search doesn't require a label on your own page; it suggests explaining how automation was used where readers would want to know. In a Google Merchant Center feed, AI-written descriptions belong in the structured description field, marked as AI-generated.
Can I paste the manufacturer's description into the AI to rewrite?
You can use it as a fact source, but a light rewrite adds nothing new. Add your own notes and customer answers so the page offers something other stores don't.
Different for you?
- Your product pages get no search traffic even with good copy: start with ecommerce SEO.
- Your AI drafts all sound the same: see AI content that keeps your voice.
- You worry Google will rank AI copy lower: read does Google penalize AI content.
- You'd rather have product pages written, checked and kept current for you: see the SEO agent.