Visora
Sign up
← Back to blog

· Visora

GEOAI CitationsEcommerce

Why do assistants quote some return policies and ignore yours? The terms layer of AI shopping

A shopper asks an assistant "what happens if it does not fit" and the assistant answers in twelve words. Those twelve words are rarely invented. They come from whichever merchant published clear, quotable return terms. Your store either is that source or it is not, and the difference has almost nothing to do with design, ads, or how fast your site loads.

Return and warranty terms are the least glamorous pages you own. They are also, in our crawl data, among the most consistently cited blocks in shopping conversations, because they settle a decision that specs alone cannot. A product can be perfect and still lose to a store whose conditions the assistant could actually state.

## Why terms beat glossy pages in citations

Assistants do not reward effort. They reward extractability. A policy page that reads like a legal essay with conditions folded into paragraphs is nearly invisible to a retrieval system looking for a direct answer. A page that states the rule, the window, who pays shipping, and the exceptions in plain sentences is trivially quotable, so it gets quoted.

The math is unkind. When three stores sell the same item, the one with the clearest stated conditions becomes the default answer for the whole category. The other two are not merely uncited; they are described through someone else's rules, which is worse. A shopper who is told a competitor offers free returns has been given a reason to leave before the conversation even reaches your product.

## The four blocks an assistant needs to quote you

Most return and warranty pages fail because they bury one of these four facts. Answer all four explicitly and you become a citation source.

1. Window and trigger. State the number of days, what starts the clock (delivery, not order), and whether it varies by category. "30 days from delivery for most items" is quotable; "generous returns" is not. 2. Condition requirements. Say plainly whether the item must be unused, unopened, in original packaging, or with tags attached. Assistants compare these across stores, and vague conditions get read as unfavorable. 3. Cost and method. Who pays return shipping, whether a prepaid label exists, and how the refund is issued. Include the cross-border case explicitly, because international return cost is the single most asked-about detail in EU and APAC sessions. 4. Exceptions. Final-sale items, hygiene products, custom or made-to-order goods. Exceptions that live in a PDF or a footnote are effectively invisible, and an assistant will assume the strictest reading rather than risk being wrong.

## How to structure the page so it gets used

Put the policy in the page body as short, self-contained sentences, one rule per line or bullet. Add a FAQ section that mirrors the exact questions buyers ask, phrased as questions, because that is how they arrive. Then wrap the page in the markup that makes the Q&A machine-readable.

{ "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [{ "@type": "Question", "name": "How many days do I have to return an item?", "acceptedAnswer": { "@type": "Answer", "text": "30 days from delivery for most items; 14 days for electronics." } }] }

Two things make that snippet work: the question text matches real language, and the answer is a complete sentence with no marketing adjectives. If your return policy lives only as a styled page with no structured Q&A, you are relying on the assistant to parse prose it may not quote. Cross-check the schema against the visible text too, since structured data that contradicts the page is worse than none.

## The cross-border trap

Returns are where global stores lose citations fastest. A policy that says "returns accepted" without stating who pays duties, whether the original shipping is refunded, and where the item must be sent back gives an assistant nothing to work with. In multi-locale sessions we consistently see stores with localized return terms cited two to three times more often in DE and JA prompts than stores shipping identical conditions in English only. hreflang tells a search engine which version to serve; it does not give an assistant a local-language answer to quote.

## Common mistakes

  • Writing the policy once and never updating it when the operation changes
  • Splitting the rules between a policy page, a product-page note, and an email-only promise
  • Using phrases like "easy returns" or "hassle-free" as substitutes for actual numbers
  • Publishing structured data that contradicts the visible on-page text
  • Ignoring the warranty path entirely, so buyers learn about defects from a competitor's page

## FAQ

Does the return policy really affect AI citations?

Yes. Terms settle the last objection in a shopping decision, so assistants look for a quotable source. Stores that publish exact, structured conditions are used as that source; stores that do not are described through someone else's rules.

How long does it take to fix?

The content change is usually a day of writing. Citations tend to follow once the page is re-crawled and the assistant's sources refresh, which can take a few weeks. Treat it as a durable fix, not a same-week bump.

Should the terms live on the product page or a dedicated page?

Both, linked and consistent. A short summary on the PDP plus a canonical policy page with FAQ markup gives an assistant a quotable sentence at the point of decision and a citable source for deeper questions.

What if my terms are genuinely worse than competitors'?

Publish them clearly anyway. A shopper who learns the real cost from you still trusts you; a shopper who learns it after ordering does not, and an assistant that cannot state your terms will quietly route the question to a store whose terms it can.

If you want to see which of your pages an assistant can actually quote, run a free scan with Visora at [geovisora.com/audit](/audit) and read it alongside the methods on our [FAQ page](/faq). The order of fixes rarely matters as much as simply having exact, structured, bilingual terms in place before peak season.

Put this into practice

Audit your PDP or category page with Visora, then fix schema and FAQ gaps that block AI citations.

Run a free GEO audit →

https://geovisora.com/en/blog/return-policy-terms-ai-shopping-citations-2026