Shopify Product Questions Before Purchase: AI Rules by Question Type

These tickets come from people who haven't paid yet. They need one fact before they check out. A reply that takes a day can lose you the sale. So can a vague one. This guide gives you a rule for each type of question.

Want product questions answered from your own rules, at any hour? Start the 14-day trial.

The three product detail questions that block a sale

Size, shipping and stock questions follow fixed rules. We compare them with after-sale tickets in our guide on pre-purchase vs post-purchase AI on Shopify, listed at the end of this article. The questions in this guide are different, because the answer depends on the details of one product.

• Compatibility: "Does this case fit the 2021 model?" The answer needs the list of compatible models, sizes and connectors. In Shopify, it usually lives in the description, a model list field or a compatibility chart.
• Ingredients: "Is this serum fragrance free?" "Does it contain nuts?" The answer needs the full ingredient list and your "free from" claims. In Shopify, it usually lives in an ingredients field, the product label or the description.
• Materials: "Is the lining 100% cotton?" "Is the frame solid wood?" The answer needs the composition of each part and the care instructions. In Shopify, it usually lives in a material field or product attribute, the description or a care guide.

Each of these questions has one right answer, and you already know it. You write it down once and decide what the agent is allowed to say about it.

Where the facts live, and what to teach your agent

In Shopify, product facts are usually spread across several places:

• Product description: the most common source, and the messiest. Facts get buried in marketing copy and go out of date when a supplier changes.
• Variants: color, size and sometimes material are stored as options. "Is the black one leather too?" depends on the variant, not the product.
• Metafields and category metafields: custom fields for structured facts such as fabric composition or compatible models. Shopify calls the standard ones product attributes in its product taxonomy.
• Your knowledge base: facts that apply to more than one product, such as a compatibility chart for a whole range, your allergen policy or your cross-contamination statement.

The safe approach is to copy the facts customers ask about most into the product information and knowledge base you teach your support agent. Don't assume the agent will find them in your product pages on its own. If two sources disagree, the agent shouldn't pick one. It should hand the ticket off.

How a good agent should handle a product question

Whatever tool you use, this is the behavior to expect and to write into your rules before you go live.

• Confirm the product and variant. If the customer writes "your blue jacket" and you sell three, the agent asks which one. For materials and compatibility, the variant often changes the answer.
• Match the question to a rule. Compatibility, ingredients and materials each get their own rule, and you set it (see below).
• Answer only from what you taught it. "The shell is 100% recycled polyester and the lining is cotton." No guessing about what the customer probably meant.
• Hand off when the fact isn't there. It tells the customer a team member will confirm. Your team then picks up a clear question, not a guess it has to undo.

The last step is what earns trust. When an agent says "I'll check with the team" instead of making up a spec, nobody gets a wrong answer. A made-up spec leads to a return, or worse.

Rules for each question type

Compatibility. Write compatibility as a list, never as a sentence. "Fits most recent models" can't be checked. "Fits iPad Pro 11 inch, 2018 to 2022" can.

• Automate: the model is on your list, or clearly not on it.
• Hand off: the model is new or missing from the list, the customer has modified the product, or the question combines two of your products ("Will this charger work with this dock?") and you never documented that pair.

Ingredients and allergens. This is the riskiest category, because the answer can affect someone's health.

• Automate: the ingredient is on the list, or plainly absent from a full published list ("Does it contain retinol?").
• Automate with your exact wording: "free from" claims you have approved, copied word for word.
• Always hand off: allergy guarantees, pregnancy or breastfeeding questions, interactions with medication, and anything a customer describes as a medical condition.

If you sell to parents, our guide on baby products support on Shopify, listed at the end of this article, applies the same care to safety tickets.

Materials and care.

• Automate: composition, care instructions, and whether a part is removable or washable, as long as that fact is written down.
• Hand off: durability promises ("Will it last outdoors all winter?"), certification requests you can't attach, and anything that looks like a warranty claim.

Write these rules once and decide exactly what the agent may answer. Start the 14-day trial.

Before launch: check the rules, not just the data

• Search past tickets for "compatible", "fit", "contain", "made of" and "allergy", and give each recurring question a rule from the section above.
• Get your allergen and "free from" wording approved by whoever signs off on labels.
• Pick ten real past tickets and check that each health question matches a hand-off rule you set.

To make these facts visible on the page itself, see our guide to Shopify product page design, listed at the end of this article.

Rule templates you can copy

Here are three rules written in plain language, the way you would type them when you set your rules in Relay. Change the product names and fields to match your store.

Compatibility rule: When a customer asks if a phone case fits their device, first confirm the exact device model. Answer "yes" only if that model is in the compatibility list in the knowledge base. Answer "no" only if the list or the product page says the case does not fit that model. If the model is not on the list, if the customer mentions a modified device, or if they ask about using the case with another product, tell them a team member will confirm and hand the ticket to the team.

Ingredients rule: When a customer asks if a skincare product contains an ingredient, first confirm the product and size. Answer only from the full ingredient list in the knowledge base. For "free from" questions, use only the approved "free from" statements, word for word. Never say a product is safe for an allergy, during pregnancy or breastfeeding, or alongside a medication. For those questions, and for any message that mentions a medical condition, do not answer. Tell the customer a team member will reply and hand the ticket to the team.

Materials rule: When a customer asks what a jacket is made of, first confirm the color or variant. Give the composition of each part (shell, lining, trim) exactly as written in the knowledge base, and add the care instructions if they ask. Do not promise durability or performance in specific conditions, and do not confirm a certification that is not documented in the knowledge base. Hand off warranty claims and any material question the knowledge base does not answer.

Each template does the same three things: it confirms the product first, it names the only source the answer may come from, and it lists exactly when a person takes over.

Answer the shopper who is one fact away from checkout. Start the 14-day trial.

FAQ

Can an AI agent answer product questions that aren't in my product description? Only if the fact is written somewhere you taught it, such as your knowledge base. If the fact isn't written down anywhere, the agent should hand the ticket to your team, not guess.

Should AI answer allergy questions for my Shopify store? It can repeat an ingredient list and the "free from" statements you approved, word for word. Allergy guarantees, cross-contamination questions and anything medical should always go to a person.

Which product questions should I automate first? Start with the ones that have one fixed answer: a model on your compatibility list, a published ingredient, a fabric composition. Leave anything involving health, guarantees or judgment to your team.

What happens when the product information is wrong? The agent will repeat the wrong fact, just like a new hire reading the same page would. Keep one source of truth for each fact, check your hand-off rules against real past tickets before launch and review hand-offs every week.