How should product pages be written so AI assistants can use them?
Short answer
Put the facts a buyer compares into fields, show them on the page, and repeat them in Product structured data. Assistants quote specific, checkable facts: dimensions, materials, standards, prices per unit and stock status. Pages that describe a product in adjectives give them nothing to extract.
Start from fields, not prose
A product page that is generated from structured fields stays consistent across thousands of products and never contradicts its own schema. The unique text per product can then be short, because the comparison facts are already there.
What to show on the page
- A specification table with the attributes buyers compare
- Price and availability that match the structured data exactly
- Answers to the questions buyers ask before buying, as short question-led sections
- The manufacturer or brand as a named entity, not just a word in the title
Where it usually breaks
Schema prices that update on a different schedule from the page, out-of-stock items marked in stock, descriptions pasted with HTML that makes the JSON invalid, and the retailer named as brand instead of the manufacturer. Each of these makes search engines discard or distrust the data.
Send the problem. Get a scoped proposal.
Within 24 hours you get a fixed scope, a delivery window and a price, or a short note saying we are not the right agency for it.