Used-Item Returns on Shopify: The Rules to Give Your AI Support Agent
An AI support agent can handle used-item return tickets safely if you give it three things: rules on condition, a limit on how much credit it can give, and a short list of cases it must pass to your team. Write those as yes/no rules before you go live, and the agent will give every customer the same answer.
This guide covers only the rules for items that are no longer new. For the full returns setup (return windows, refund limits, what never to automate), read our guide to setting up Shopify returns and refunds with AI. For final-sale and discounted items, see the Shopify sale-item return policy guide. Both are listed at the end of this article.
Why used items need their own rules
A standard return rule asks two questions: is the order inside the window, and is the product returnable? Used-item tickets add a third question that the order data can't answer: what state is the item in?
The customer's words are all your agent has to go on. "I only wore it once." "I opened it but it doesn't suit me." "I used it and it broke." Each of these needs a different answer. If you don't spell out the rules, the agent has to guess, and guessing is what leads to wrong refunds.
Rule 1: Define condition levels the agent can check
Give the agent a fixed list of condition levels, written as facts the customer can confirm. Avoid vague words like "lightly used." For each level, here is what the customer can confirm and the default outcome:
- Tried: tags on, not washed, not worn outside, packaging kept. Default outcome: normal return under your policy.
- Used, resellable: worn or used, no damage, could sell as open box or second grade. Default outcome: partial store credit, within your limit.
- Used up or unsellable: opened hygiene product, used cosmetics, worn underwear or swimwear, personalized item. Default outcome: polite refusal quoting the policy.
- Used, now faulty: "It stopped working," "it broke during use." Default outcome: hand off to your team or warranty process.
Turn each level into questions the agent asks before it decides:
- Are the tags still attached?
- Has the item been washed, worn outside or set up for regular use?
- Do you still have the original packaging?
- Is anything damaged, or did something stop working?
If any answer is unclear, the rule should say: ask for photos, then pass the ticket to your team. The agent should never decide a condition from a vague answer.
Rule 2: Map product types to what is allowed
Condition alone isn't enough. A worn jacket and opened face cream need different rules. Tag your products by type in Shopify, then give the agent one line per type.
Example rules for a store that sells apparel and skincare:
- Apparel, tried: accept the return.
- Apparel, used and resellable: offer partial store credit, never a cash refund.
- Underwear and swimwear, any use: refuse, quote "Non-returnable items."
- Skincare, sealed: accept the return.
- Skincare, opened: refuse, offer a product match for their skin type.
- Personalized items, any condition: refuse unless faulty.
Keep these rules exactly in line with your public return policy. If your policy page says "unused items only," the agent should never promise more. Whatever it offers on a used item is a goodwill gesture, and the reply should say so.
Rule 3: Set credit limits the agent can't go over
Partial credit is the outcome most likely to go wrong without a limit. Set it in three layers:
- A fixed share per product type. Choose it from your own margins and what you can resell the item for. The same case always gets the same amount.
- A cap per ticket in your currency. Above that amount, the ticket goes to your team, even when the rule would allow it.
- The form of the credit. Store credit or a single-use discount code, not money back to the card. Add an expiry if your policy allows it.
Also decide whether the item must come back. Some stores ask for the item back before giving credit, and others let the customer keep it. Look at your own return shipping cost and resale value to decide, then write it as a rule so the agent never has to choose.
Rule 4: Write the handoff triggers
These tickets must go to a person, even when the condition looks clear:
- Possible defect. "It broke," "it stopped working," "it arrived like this." This is a fault, not a used-item return. Send it to your warranty support process, covered in our warranty guide listed at the end of this article.
- Repeat used-item returns. The customer has had used-item credit before. Set the number that triggers a review.
- High-value items. Anything above your per-ticket cap.
- Customers outside your home market. Consumer rules differ by country. Check your local consumer rules before refusing, and let your team handle cross-border cases until you have.
- Disputes and anger. Threats of a chargeback, a bad review or legal action.
- Photos that don't match the answers. The customer says "tags on" and the photo shows none.
A short, clear list works better than a long one. The agent needs triggers it can spot in the ticket or the order, not judgement calls.
Rule 5: Fix what every reply must contain
You don't need a script for every case. You need four parts in every used-item reply:
- The decision, stated plainly.
- The reason, quoting one line of your policy.
- What happens next (label, code, or "our team will reply in this thread").
- Something useful when the answer is no: a size guide, care tips or a better product match.
A refusal with nothing useful in it is the reply most likely to turn into a complaint. Make the fourth part required for every refusal rule.
Test the rules before going live
Run old tickets through your rules before turning anything on. Take a handful of past used-item tickets from your inbox and check:
- Did the rules give the answer your team actually gave, or a better one?
- Did any ticket fit two levels at once? If so, make the condition questions tighter.
- Did any credit go over your cap?
- Did every possible defect go to a person?
Fix the rules until every test ticket gets one clear outcome.
Where Relay fits
Relay answers, refunds and resolves your Shopify tickets 24/7, following your rules. You write the condition levels, product rules and credit limits once. Relay understands the request, finds the Shopify order, applies your rules and acts. That can mean sending a discount code, processing a refund or cancelling an order, within the permissions you give it. When a ticket hits one of your handoff triggers, it brings in your team.
Most AI support tools reply. Relay resolves.
Setup comes down to the steps shown on the site:
- Teach Relay your business.
- Set your rules.
- Connect and go live.
The rules in this guide go straight into the second step.
FAQ
Can an AI agent decide whether an item is "used"? Only from facts the customer confirms, such as tags attached, washed or not, packaging kept. Give it a fixed list of questions and condition levels. When the answers are unclear or don't match the photos, the rule should pass the ticket to your team.
Should the agent give partial refunds or store credit for used items? Store credit or a discount code is easier to control, because you set the amount and expiry in advance. Whichever you choose, set a fixed share per product type and a cap per ticket so the agent never goes over.
What used-item cases should never be automated? Possible defects, repeat returners, high-value items, cross-border customers whose consumer rules you haven't checked, and any ticket with a dispute or chargeback threat. For the wider list, see our guide to Shopify returns and refunds with AI, listed at the end of this article.
Can I refuse every return on a used item? Only if your return policy says so and the customer could see it before buying. Consumer rules vary by country, so check your local consumer rules before refusing.

