The Description an AI Reads Is Not the One on Your Page
Nearly every Shopify theme outputs a block of JSON-LD on the product page. Buried in it is a description field, and in most themes it is generated the same way: your product description, HTML stripped out, cut off at a character limit. Something close to this:
{{ product.description | strip_html | truncate: 500 | json }}
Some themes use 5000. Plenty use 500. Either way, that flattened string is the version a machine reads when it is working out whether to put you on a shortlist. It is not the version you wrote. The customer gets headings, bullet points, images, a spec table, a tabbed section. The machine gets the first N characters of plain text, in the order you happened to write them.
"That flattened string is the version a machine reads when it is working out whether to put you on a shortlist. It is not the version you wrote."
Which makes the opening of your description matter far more than it looks like it should.
The schema description field, by the numbers
500
Typical character truncate limit (some themes use 5,000)
1 of 8
Core Product schema properties that's prose, not a discrete fact
2
Competing description fields if an SEO app also injects Product schema
Most descriptions open with a packing list
The pattern we see over and over: the description starts with dimensions, weight, materials, care instructions and what is in the box. All perfectly good information. It is just answering the question a customer asks second.
Strip the HTML, truncate at 500 characters, and what a language model receives is a product name followed by four hundred-odd characters of measurements. Nothing about who the product is for, what problem it solves, or what separates it from the other forty products the model is holding alongside it. The positioning almost always exists somewhere on the page. It sits below the fold, or in a tab, or in a metafield, and it never reaches the field the machine is reading.
The result is a product that reads as generic to the one reader who cannot look at your photography, your reviews or your brand to make up the difference.
What the field is actually for
Description is one of the core Product schema properties, sitting alongside:
- name
- image
- brand
- sku
- gtin
- offers
- aggregateRating
Every one of those others is a discrete fact. A price is a price. A GTIN is a GTIN. Description is the only place in the entire object where you get to say what the thing is and who should buy it.
"Description is the only place in the entire object where you get to say what the thing is and who should buy it."
So treat it as a positioning statement rather than a summary. It is the same discipline as the product title, which we covered in Your product titles don't say what the product is: what it is, who it is for, what makes it different. Two or three sentences of that at the top of the description does more work than five hundred characters of specification ever will.
The spec detail does not disappear, it moves. Weight, size and material have their own schema properties. Anything without a standard field can go into additionalProperty, or into metafields rendered further down the page. Machines would rather have those as typed fields anyway, which is the same argument as the catalogue work generally: structure beats prose whenever the reader is software.
How to check your own
It takes about a minute:
-
Open a product page and view source. Search for
"@type":"Product". - Read the description value as it appears there, not as it appears on the page.
- Check where it cuts off. That tells you your theme's truncate limit.
- Check whether that first sentence would help anyone choose between you and a competitor.
- Prefer Google's Rich Results Test if you'd rather not read raw JSON — it shows you the same parsed object more comfortably.
One thing worth checking while you are there. If an SEO app is also injecting Product schema, you may have two blocks with two different description values on the same page. That happens more often than you would think after an app install, and it is worth knowing which one is winning before you spend time rewriting the wrong one.
The bottom line
Take your ten highest-traffic products and read the schema description for each. Not the page copy. The field. If ten out of ten open with a measurement, you have found a cheap change with an unusually wide reach: the same edit improves the page for customers, the data going to Merchant Center, and what an AI assistant has to work with when someone asks it for a recommendation in your category. Ten products, an hour, no developer.
At beCommerce, we check exactly this field as part of every catalogue and structured-data audit. If you want a hand rewriting your schema descriptions across your best-selling products, let's talk.
Is your store making any of these mistakes?
We audit Shopify stores for exactly these kinds of conversion-killing patterns — the ones that have become normalised until you see them through a customer's eyes.
Book your free audit →
