Almost every Shopify store has an FAQ section somewhere product page accordions, a dedicated FAQ page, maybe a help centre app bolted on top of the theme. Most of them also have FAQPage schema wrapped around that content, because some SEO checklist said to add it. And most of that markup is technically valid and practically useless to an AI agent trying to answer a question.
This is worth separating clearly, because "has FAQ schema" and "is machine-usable" get treated as the same achievement when they're not.
Valid markup isn't the same as extractable content
Schema validators check structure: is there a Question, does it have an acceptedAnswer, is the markup well-formed. They don't check whether the answer inside that acceptedAnswer field is actually a complete, standalone response to the question, or whether it's a half-sentence that only makes sense next to a product image and three paragraphs of surrounding copy.
An agent parsing FAQ schema is looking for a self-contained answer it can lift and use directly. If the acceptedAnswer text reads "Yes, see below for details" or "It depends on your specific situation, contact us for more info," the schema is valid, and the content is worthless. The validator has no way to catch that, because completeness isn't a structural property; it's a content property, and nobody's checking for it.
The pattern that breaks most Shopify FAQ implementations
The failure mode is consistent enough that it's worth naming directly: FAQ content gets written for a human who has visual and contextual scaffolding around it a page layout, a product image, an accordion that's already open to a related question and then that same text gets dropped into schema markup with the scaffolding stripped away.
A question like "does this run true to size" answered with "runs slightly small, we recommend sizing up" works fine visually, sitting under a product title with images alongside it. Pulled out as an isolated Q&A pair with no product context attached, an agent has no idea which product that answer belongs to unless the schema is scoped correctly, and even then, "sizing up" from what baseline is now ambiguous.
The fix isn't more schema. It's writing the answer as if it will be read with zero surrounding context, because for an agent, it will be.
What machine-usable actually requires
A few adjustments separate markup that validates from content that gets used:
Every accepted answer should be a complete sentence or two that fully resolves the question on its own, with no "see above," "as mentioned," or dependency on nearby text
Product-specific FAQ content needs the product identity restated inside the answer itself when there's any chance the Q&A pair gets extracted independently of its page context
One FAQPage block per logical topic works better than one large block covering unrelated questions about shipping, sizing, and returns all at once mixed-topic blocks make it harder for an agent to isolate the relevant pair
Avoid conditional hedge answers ("it depends," "varies by region") wherever a concrete default answer is possible; hedges are unusable to an agent trying to state something confidently
None of this changes how the FAQ looks to a human visitor. It only changes whether the content underneath the markup is worth extracting in the first place.
Where this fits in the bigger picture
This is really the same problem that shows up with metafields and structured data generally: Shopify stores are full of markup that satisfies a checklist without satisfying the actual consumer of that markup, which is increasingly a model doing extraction rather than a search engine doing indexing. Schema tells an agent where to look. It doesn't guarantee there's anything usable once it looks there.
The fix is worth treating as its own audit pass, separate from schema validation entirely, going through existing FAQ content and checking whether each answer would survive being read completely out of context. Most existing FAQ sections don't, and that gap is usually fixable in an afternoon once it's visible.
Worth checking before assuming FAQ schema is a solved problem on any given store. Valid and usable aren't the same claim, even though they get treated like one.
Top comments (1)
ngl i always wondered why my rich snippets weren't showing up even though the content was there, the marked up vs machine usable part finally makes sense now