Your returns policy was drafted by someone whose job is to limit liability. Every clause was negotiated for a human dispute: a customer, a call center, maybe a chargeback. That document is now being read by a buyer that does not get tired, does not skim, and does not forgive ambiguity. It reads your return window, your restocking fee, and your fit guarantee as inputs to a ranking function. Then it ranks you.
This is the part most boards have not internalized. McKinsey’s QuantumBlack team framed the shift with precision: agents optimize for delivered value, factors such as price, availability, service reliability, and reversibility. Reversibility is the analyst’s word for the return experience. And the consequence is blunt: merchants that expose clean inventory data, predictable fulfillment performance, and transparent substitution and return policies become default suppliers, often without ever winning a traditional moment of consideration.
Read that again from the CFO seat. You can become the default choice for an entire category without ever running a campaign, winning a search term, or designing a landing page. The lever is operational clarity, expressed as data.
Your terms became ranking factors while you were writing them for legal
The agent comparing two retailers with identical SKUs at identical prices has to break the tie somewhere. Return window is a clean numeric field. Restocking fee is a clean numeric field. Fit guarantee is a binary with conditions. These are exactly the variables a model can rank on, instantly, across every merchant in the consideration set.
The problem is where those terms live. Returns policies are written for human readers, buried in PDFs that machines skip entirely. When the agent cannot read a term, it does not pause to investigate. Publishing return terms as structured data attached to the offer lets an agent compare your conditions against a competitor’s without guessing, and guessing is where most brands quietly lose the shortlist.
That last phrase is the whole game. The agent resolves uncertainty by moving on. You never see the lost cart because there was no cart. There is no bounce metric for a shortlist you were silently dropped from.
The Cost of Doing Nothing is now a weekly compounding loss
Here is how I size the CODN for a retail client. Take your agent-attributable sessions, which are growing fast. Adobe Analytics found AI-sourced traffic to U.S. retail sites grew 393% year over year in the first quarter of 2026, and in March 2026 that traffic converted 42% better than organic. That converting traffic is the segment most exposed to your unstructured policy, because the conversion only happens if the agent clears you through its filter first.
Now run the counterfactual. Every week your return terms sit in prose, agents are being trained on hundreds of thousands of transactions that teach them which merchants are legible and dependable. The retailers building agent-ready infrastructure now will be the default choices agents learn to trust over hundreds of thousands of transactions. The retailers running a discovery project in Q4 will be competing for visibility they quietly lost in Q2. That is the CODN in one sentence. The loss compounds with every transaction. It is a preference that hardens against you with every transaction you are absent from.
Peak season concentrates the damage. Gift purchases carry higher return risk, so both shoppers and the agents acting for them scrutinize reversibility harder in Q4 than at any other point in the year. The window where your policy page matters most is also the window where it is least ready.
What to restructure before peak, and who owns it
Three moves. None requires a platform replatform.
First, lift return window, restocking fee, and fit guarantee out of the policy page and attach them to the offer as structured data. The full schema includes GTIN/MPN, hasMerchantReturnPolicy, shippingDetails, review nodes, the fields agents weigh when qualifying you. The hasMerchantReturnPolicy field is already a standard. Populate it.
Second, enforce a single source of truth. A feed that shows an item in stock while the storefront shows a two week lead time teaches an agent that your data cannot be relied on, and unreliable merchants get filtered before price is ever compared. If your feed and your policy page disagree on the return window, you are penalized for the inconsistency itself.
Third, if you already offer generous terms, make them legible so the agent can quote them. The generosity you cannot express is generosity you are paying for and getting no ranking credit for.
Ownership is the real decision. The policy page used to belong to legal and risk. It now sits at the intersection of merchandising, data, and digital. Assign one accountable owner with authority over all three, measured on agent-attributed conversion.
The retailer that treats its return policy as a product data asset this quarter becomes the default supplier next year. The one that treats it as legal boilerplate finds out, transaction by transaction, that the machine already made its choice.