Return policy data for AI shopping
Return terms are one of the last things an AI shopping agent resolves before it commits to a recommendation — and one of the least often published in a form it can read. A generous return policy stated only in prose is, to an agent, an unresolved risk. The competitor with thirty days in a structured field wins the close call.
A machine-readable return policy is schema.org MerchantReturnPolicy markup attached to a product offer, stating the return window, applicable country, return method and who pays return shipping, so an AI agent can evaluate purchase risk without interpreting prose.
Why returns decide close calls
When two products satisfy the same constraints, an agent has to break the tie on something it can defend. Risk is the natural tiebreaker: which purchase is more recoverable if it is wrong?
That question is answerable only if the terms are fields. An agent cannot responsibly summarise a policy page written for humans, with exceptions, category carve-outs and a link to a form — so it treats the terms as unknown, and unknown loses to thirty days free returns.
The fields
MerchantReturnPolicy attaches to the offer. Four fields carry almost all the decision weight.
| Field | What it answers | Why it matters |
|---|---|---|
| merchantReturnDays | How long do I have? | The single most-weighted return signal in a comparison |
| returnFees | Who pays return shipping? | FreeReturn versus a customer-paid return flips many close calls |
| returnMethod | How do I send it back? | In-store versus by mail changes practicality for the buyer |
| applicableCountry | Does this apply to me? | A policy that does not name a country cannot be applied to a delivery-constrained query |
| returnPolicyCategory | Is there a window at all? | Distinguishes a finite window from no returns or unlimited returns |
"offers": {
"@type": "Offer",
"price": "129.00",
"priceCurrency": "USD",
"availability": "https://schema.org/InStock",
"hasMerchantReturnPolicy": {
"@type": "MerchantReturnPolicy",
"applicableCountry": "US",
"returnPolicyCategory": "https://schema.org/MerchantReturnFiniteReturnWindow",
"merchantReturnDays": 30,
"returnMethod": "https://schema.org/ReturnByMail",
"returnFees": "https://schema.org/FreeReturn"
}
}What if my policy is restrictive?
Publish it anyway. A fourteen-day, customer-paid return window stated clearly is a known quantity an agent can weigh; an unstated policy is a risk it discounts more heavily than the actual terms usually deserve.
It also protects you at the other end: an agent that cannot read your terms may describe them optimistically by analogy with the category, which is a support problem you inherit.
- State exceptions where they apply
- If final-sale categories exist, emit a different policy on those offers rather than one blanket claim.
- Keep the page and the markup aligned
- A prose page saying 60 days against markup saying 30 is worse than either alone.
- Do not overstate
- Marking every product FreeReturn when returns are conditional produces recommendations you then have to honour or dispute.
Frequently asked questions
- Why do AI shopping agents care about return policies?
- Return terms are how an agent assesses purchase risk when several products satisfy the same constraints. A machine-readable window and fee structure lets it defend a recommendation; an unreadable policy is treated as unknown risk and usually loses.
- How do I make my return policy machine-readable?
- Attach schema.org MerchantReturnPolicy to your product offer with merchantReturnDays, returnFees, returnMethod, applicableCountry and returnPolicyCategory. A link to a prose policy page is not a substitute.
- Should I publish a restrictive return policy in structured data?
- Yes. Clear restrictive terms are weighed on their merits; absent terms are discounted more heavily than most real policies deserve, and an agent may otherwise describe your terms by analogy with the category and create a support problem.
- Does a return policy page with schema markup count?
- It helps, but the field that changes a product comparison is the one attached to the offer. Agents evaluate risk per product, and products in final-sale categories often carry different terms than the site-wide page implies.
See how your policy signals score
The scan checks shipping, returns and warranty for machine-readable terms and reports what is missing.