A growing number of shoppers start with a question rather than a search: which pasta sauce has no added sugar, what is a good gift for someone who loves hot sauce, which sparkling water comes in cans. AI assistants answer by reading web pages, pulling out the facts and deciding what to recommend.
That makes a new kind of question worth asking about your website. Not just “does it look good?” but “can software read it?” Plenty of attractive product sites fail that second test without anyone noticing.
How AI tools see a website
Search engines and AI assistants send automated programs, called crawlers, to read pages. What they handle well is text in the page itself. What they handle less reliably:
- Words inside images, such as an ingredients panel photographed from the pack
- Information that only appears after scripts run, like product details loaded into tabs
- PDFs, which some tools read and some do not
- Content behind pop-ups or gates that need a click before anything is shown
If the facts a shopper cares about are only in those places, an assistant may simply not know them, and it will not recommend what it cannot describe.
Common ways product sites hide their best information
- The ingredients and nutrition panel exist only as a photo of the back of the pack.
- Sizes and varieties sit in a dropdown that only appears once a script runs.
- The where-to-buy list lives inside a map widget, with no text version.
- The brand story is a video with no written version.
- Every product shares one page, so there is no single address that describes one product.
The fixes
Write the facts as text
For each product, write out its name, size, what it is, ingredients or materials, allergens, how to use it and where to buy it. Keep the pack photos too, but treat them as illustration, not as the only place the information exists.
One page per product
Each product deserves its own page with a stable web address, a clear title and a description written for a person. It gives search engines and assistants one definite place to learn about that product, and one place to send people.
Add structured data
Structured data is code that labels the facts on a page: this is a product, this is its name, its brand, its image, its barcode number, its price and availability. Shopify, WordPress and most modern platforms can add it, and it removes the guesswork for anything reading the page.
Check what you are blocking
A file on your site called robots.txt tells crawlers what they may read. It is easy to block more than you meant to, sometimes by a setting left over from when the site was being built. Some AI companies use separate crawlers for training their models and for answering questions, and blocking some of them can keep you out of their answers. Whatever you decide, decide it deliberately.
Keep facts consistent everywhere
Your website, your marketplace listings and your retailers’ product pages should agree on names, sizes and ingredients. Conflicting details make any system less confident about recommending you.
How to test it
- Ask the assistants. Ask ChatGPT, Gemini or Perplexity about your product by name and by category, and see what they say.
- Look at the page without the extras. Read just the text of a product page. Is everything a shopper needs still there?
- Use Google’s free tools. Search Console shows how Google sees your pages, and the Rich Results Test checks your structured data.
An example
On the website we designed for Zia Nina, each range has its own page, and the facts on the jar are written out on the page: the four pasta sauces, the 25 oz (708 g) glass jar, and the colour band that tells each variety apart. The jars are there to be seen, but nothing important exists only in a picture.
If you are not sure your website is telling machines what it tells people, let’s talk.