Someone writes to a pet shop's assistant: "I'm looking for a bed for my dog, who weighs 25 kilos, and I want to be able to put the cover in the washing machine." The shop has exactly the right bed. The trouble is that its product page says "Nordic Bed L, 90 × 70 cm, grey" and nothing else. Not what size of dog it's for, not whether the cover comes off, not whether it can be washed.

With that product page, an AI assistant has two options, and both are bad: recommend it without knowing whether it fits, or leave it out. A good member of staff would know, because they've had the bed in their hands. The AI only knows what's on the page.

The data AI needs to recommend well

If a detail isn't on the product page, as far as your shop's assistant is concerned it doesn't exist. These are the details that matter most when someone asks for advice:

  • A title that says what the product is. "Nordic Bed L" gives the model name; "Large dog bed with washable cover" says what it is. A good title has both.
  • A description that explains who it's for and what it does, not just its dimensions. "For dogs up to 30 kg" or "made for sleeping indoors, not for the car" is exactly what someone who can't decide is looking for.
  • Specific attributes, always in the same place. Dimensions, material, colour, weight, compatibility. If the material is in the title on some pages, in the description on others and missing on the rest, the AI struggles to compare them.
  • Variants with clear names, and their own price and stock. A bed in three sizes needs each size to have its dimensions and availability, not just an unexplained "S/M/L".
  • Up-to-date price and stock. A perfect recommendation for a product that's sold out is still a bad recommendation.
  • The answers to the usual questions. If everyone asks whether the cover is machine-washable, that answer belongs on the product page.

It isn't just for your assistant

Google Merchant Center asks you to include a product's most relevant attributes in its description, such as size, material, intended age range or special features, and notes that Google may use that text to find your item. The specification OpenAI publishes for shops sending their products to ChatGPT asks for much the same, in separate fields: identifiers, description, price, stock and variants with their attributes, such as colour or size.

It doesn't matter whether your catalogue is being read by your site's search, a semantic search engine, an AI shopping assistant or ChatGPT: they all rely heavily on what you've written. A well-described catalogue works for all of them at once, and for the customer reading it directly.

How to review your catalogue in an afternoon

You don't need to rewrite everything. Start where it matters most:

  1. Pick your twenty best sellers, plus the twenty that get the most visits without selling. That's where an incomplete page costs you most.
  2. Read their pages as someone who doesn't know the brand. Could they tell who each product is for and when it would suit them? If the answer only exists in your head, it's missing from the page.
  3. Check the variants. Every size, colour or model should have a name that makes sense, its own price and its own stock.
  4. Look at the questions you get by email, WhatsApp or social media. Every repeated question is a detail missing from a product page.
  5. Make the attributes consistent. Decide where the material goes and where the dimensions go, and always put them there.

What you don't need to do

This isn't about writing more for the sake of it. A 500-word description packed with adjectives helps no more than three sentences with the facts that matter. Stuffing pages with repeated keywords doesn't help AI, which searches by meaning, and it helps your customers even less. What makes the difference is that the detail someone needs to decide is written down, correct and where they'd expect to find it.

SmartShop AI, a sales assistant for Shopify and WooCommerce stores, works with the data already in your shop: there's no separate catalogue to prepare, but the clearer your product pages are, the better it can recommend. How it works walks you through connecting your shop, step by step.