Sub-grade spoke
Return & Refund Policy — can an agent confirm your return terms before transacting?
Returns are the trust-tail of a transaction — and agents acting on a user's behalf treat them as a hard precondition, not a nice-to-have. Sites that publish MerchantReturnPolicy schema with structured return window, restocking fee, and method get cited as agent-eligible commerce. Sites that bury the policy in a PDF or a footer link get filtered. Two-thirds of shoppers check the policy before buying — agents check it on the user's behalf, every time, at machine speed.
By Chris Mühlnickel · 2026-05-16
What is Return & Refund Policy?
Return & Refund Policy is whether your site publishes MerchantReturnPolicy Schema.org markup — with return window, restocking fee, return method, and refund mode — and surfaces the same terms as plain-English copy reachable from the product page.
By the numbers
- 67% — of shoppers check the return policy before buying — and 8 in 10 read it pre-purchase. (ClickPost — Ecommerce Return Statistics)
- $849.9B — in merchandise expected to be returned in 2025 — 15.8% of all annual US retail sales. (NRF — 2025 Retail Returns Landscape)
- 19.3% — of online sales are returned in 2025 — significantly higher than the cross-channel rate. (NRF — 2025 Retail Returns Landscape)
Why it matters
Returns are where transactional trust is actually tested. The 67% of shoppers ClickPost reports as checking return policies pre-purchase is the human-side anchor; agents check the policy on the user's behalf every time, deterministically, at machine speed. Sites that publish MerchantReturnPolicy markup get treated as confidently reversible transactions — the agent can recommend them without leaving the user exposed to a one-way commitment. Sites without the schema force the agent to assume the worst case, which means filtering them out of recommendations involving expensive or untested SKUs. The $849.9B in 2025 US returns NRF measures is the scale of the trust market that returns markup gives agents access to.
The structured-vs-visible-text gap is wider on returns than on shipping. Most ecommerce sites have a return policy — usually buried in a PDF, a footer link, or a separate /returns page that requires another fetch. To an agent comparing three merchants for a user's order, "return terms exist somewhere on this site" is functionally indistinguishable from "no return terms" when the comparison happens at machine speed. The structured competitor wins by default because the structured competitor is the one whose terms the agent can actually confirm at the right step in the flow.
Returns are the failure mode that breaks agent commerce. The 19.3% online return rate NRF measures means roughly one in five agent-initiated transactions will trigger a return. If the agent can't confirm the return path pre-purchase, the user blames the agent for the bad transaction — and the underlying merchant gets routed off the agent's eligible set. Sites without MerchantReturnPolicy schema take the reputation hit twice: once at the recommendation step (downranked) and again when returns fail (rated badly via vendor reputation). Sites that ship the schema get cited as agent-safe.
MerchantReturnPolicy is the cheapest agent-commerce eligibility ticket. The schema is well-defined, supported by Google Merchant, supported by Shopify/WooCommerce/BigCommerce out of the box, and required by ACP/UCP-style agent checkout protocols. The work is hours; the payoff is being inside the agent-eligible merchant set as that share grows. Sites that publish it in 2026 are positioned for the channel as it scales; sites that wait are filtered out when the first wave of agent-checkout traffic arrives.
Where it's heading
Return terms become a binary gate for agent-completed transactions. Within 12-24 months expect ACP, UCP, and AP2-style flows to require machine-readable MerchantReturnPolicy as a hard precondition — not just a preference. Sites without the schema drop out of the eligible merchant pool. The 67% pre-purchase check rate today becomes 100% mechanical pre-check in agent flows.
Returns become a pricing signal, not just a trust signal. Real-time return-cost data flows into agent recommendations — agents weigh restocking fees, return shipping cost, and refund-mode against the listed price when comparing merchants. The site with a 30-day full-refund policy beats the site with a 14-day store-credit policy at the same nominal price; the structured comparison happens machine-side. Sites that publish complete returns schema win the implicit pricing layer; sites that hide terms lose the comparisons.
Returns instrumentation becomes regulatory. EU and US consumer-protection regimes are tightening on return-policy transparency. The compliance answer maps onto the agent-readability answer: publish structured terms accessible from the product page. Sites already running MerchantReturnPolicy markup are positioned for both layers; sites delaying disclosure to the policy page face dual pressure.
Common mistakes
- Return policy in a PDF or behind a separate page. Agents won't fetch an extra resource to discover terms — the comparison happens with what's structurally extractable on the product page.
- MerchantReturnPolicy missing returnFees or restocking-fee data. The schema technically validates; the agent treats the absence as 'fees apply but unknown' and downranks accordingly.
- Visible policy and schema disagree. Site updates the visible return window but forgets the schema. Agents read the structured number; humans read the visible one. Customer service handles the gap.
- No country or region in `applicableCountry`. A global site with a US-only return policy reads as having no policy for international customers.
- Implying full refund but executing store credit. Schema must reflect actual refund mode. The mismatch routes back through vendor-reputation channels and compounds across listings.
Frequently asked
What's the minimum return-policy markup my product pages need?
A MerchantReturnPolicy block referenced from the Product Offer, with returnPolicyCategory (e.g. MerchantReturnFiniteReturnWindow), merchantReturnDays, returnMethod, returnFees, and applicableCountry. The fields exist because Google Merchant requires them — and the same fields are exactly what agents read to confirm a transaction is reversible. A Spekto audit reports the per-page gaps.
Does returns markup belong on every product page, or sitewide?
Both, ideally — sitewide as a default policy, per-product when the SKU has different terms. The schema supports MerchantReturnPolicy at the Organization level and Offer-level overrides. Sites with uniform terms ship one block referenced from every product; sites with category-varying terms (final-sale items, custom-made goods, perishables) override per-SKU. The audit catches the case where the visible policy and the structured policy disagree.
What if my product is genuinely non-returnable — final sale, custom-made?
Publish that explicitly with returnPolicyCategory: MerchantReturnNotPermitted (or the equivalent specific to your case). Agents handle non-returnable items differently — they warn the user before transacting, but they don't refuse to recommend. Sites that publish a clear non-returnable signal earn more agent trust than sites that hide return terms and force the agent to assume the worst.
How do return policies interact with agent-checkout protocols (ACP, UCP)?
Return terms are pre-purchase confirmation data in ACP and UCP flows. The agent reads MerchantReturnPolicy before submitting payment; if the policy is missing or unparseable, the transaction degrades to manual handoff or routes to a competing merchant whose policy is machine-readable. The schema isn't an SEO nice-to-have once these flows scale — it's the access gate.
Should restocking fees be machine-readable or buried in fine print?
Machine-readable. The returnFees field on MerchantReturnPolicy carries the value; agents read it and surface it during the user's decision. Sites that bury restocking fees in fine print get flagged by agent comparison tools as 'hidden return cost' and downranked relative to transparent competitors. The structured field is the cleaner UX and the cleaner schema simultaneously.
What about refund mode — money back vs. store credit?
MerchantReturnPolicy supports refundType with values like FullRefund, StoreCreditRefund, and ExchangeRefund. Sites that ship store-credit-only refunds and publish that structurally are rated honestly by agents; sites that imply full refund but execute store credit get bad reviews routed back through vendor reputation. The schema is the alignment mechanism.
How often should return policies be re-validated?
On every deploy that touches commerce schema, ideally in CI via Google's Rich Results Test or the Schema Markup Validator API. The common failure mode is the schema drifting from the visible policy — site updates the return window from 30 to 14 days in marketing copy, forgets to update the schema. The dual-source-of-truth disagrees, agents pick the worse number, customer service handles the complaints.