E Commerce

Price Changes in AI Shopping Chat: What to Check

Check the exact variant, currency, offer conditions and source before quoting a price in AI shopping chat. Use this practical test worksheet.

6 min read
ecommercepricingAI chatproduct data

A shopper asks whether a product is still on sale. The answer gives one price, the product card shows another, and the merchant page opens with a third. Before adding a reassuring label, check whether those surfaces are describing the same offer.

The practical answer: bind a quoted price to an exact item, variant, merchant, currency and set of offer conditions. Check the source and when it was observed. If the current price or a promotion cannot be verified, say so and give the shopper a clear path to the relevant merchant page. A recent timestamp alone does not establish that the amount applies to this shopper.

Start with the offer, then check the number

Two cards with the same product name can represent different purchases. Size, color, software edition, license term, quantity and seller can change which price is relevant. A low starting price for one variant does not answer a question about a different variant.

Shopify's ProductVariant reference describes variants as specific versions of a product and exposes price separately from compare-at price and availability. Its contextual pricing reference also distinguishes pricing for a particular country. These distinctions are useful checks; they do not establish how a particular assistant is connected.

For each answer, identify the product and selected variant, the merchant, quantity or purchase term, and the country and currency relevant to the request. If the shopper has not supplied a necessary choice, ask for it before presenting an exact answer. Keep a product's starting price visibly separate from a quote for the selected item.

Separate a sale price from an eligible promotion

A crossed-out comparison price, a discount code, a member offer and cashback answer different questions. A shopper asking what they will pay needs the applicable conditions, not just the most attractive number available in a feed.

Check the source for dates, selected items, minimum purchase, membership or new-customer conditions, and whether the offer applies automatically or needs a code. Do not subtract an unverified discount from the listed price. Keep potential cashback separate from the amount due to the merchant; eligibility and later credit are separate questions.

When those conditions are absent, the useful answer is specific: explain which condition remains unknown and point to the offer's terms or merchant page. Avoid turning a plausible promotion into a confirmed saving.

Record what the source actually established

A feed update, a retrieved page and a successful answer are different observations. A feed may contain a valid item but an older price. A newly fetched page may describe a different variant. A field named price does not, by itself, establish the shopper's final checkout total.

Schema.org's Offer reference distinguishes price, currency, availability and a price-valid-until date. Structured fields help describe an offer; their presence does not prove that the source is current or that a connected assistant checked them correctly.

Keep the source identifier and observation time with the item being evaluated. Where a source supplies its own update time or offer expiry, retain that separately. Only call a price verified when the actual check supports that wording. Set any refresh interval around the connected source and buying context, then test it; there is no universal interval that proves every offer accurate.

Give uncertainty a useful next step

If sources disagree, preserve the difference instead of silently choosing the lower amount. If the selected variant cannot be confirmed, ask for the missing choice. If a current offer is unavailable, explain that limitation and link to the relevant merchant page for confirmation.

The handoff also needs checking. Opening a product page, adding a particular variant to a cart, and completing a purchase are separate actions. Test the supported destination and describe the action that actually happens. Do not label a link as add to cart when it opens a product page, or promise a final total before the required shipping, tax and offer conditions are known.

Run these six cases before relying on the answer

These are suggested test fixtures, not observed customer results or claims about automatic Brambles behavior. Use products and offers from the actual connected source.

  • Variant mismatch: ask about one size or edition while another has the lowest price. Check that the answer preserves the requested variant or asks for the missing choice.
  • Expired promotion: provide a listing with an offer that has ended. Check that the answer does not present the expired discount as currently available.
  • Conditional discount: ask about an offer requiring membership or a minimum purchase. Check that the condition remains visible and eligibility is not assumed.
  • Different currency or country: compare the same item in two supported markets. Check that each amount stays attached to its market and currency.
  • Source disagreement: use a feed and merchant page with different amounts. Record the difference, the relevant variant and the answer's explanation rather than treating either amount as automatically correct.
  • Unavailable current source: test a source that cannot be checked. Look for a clear limitation and useful destination, with no invented verification time or checkout guarantee.

Keep a small worksheet for every case

Record the question, exact item and variant, merchant and destination, country and currency, quantity or purchase term, and promotion conditions. Add the source and observation time, the assistant's wording, and the displayed card or link.

Then record the destination actually reached and what it showed for that same item. Grade the case as supported, contradicted or unresolved, with a short reason and a person responsible for the next check. Keep the original evidence when a later source changes. That makes it possible to distinguish an incorrect answer from an offer that changed after the observation.

Where Brambles fits

Brambles powers AI chat grounded in connected content and catalog information. Product information, merchant links and eligible product actions depend on the source and account setup. Retailer catalog routes include Shopify and CSV, JSON or API inputs; current catalog quality and freshness still need checking for the particular account.

Start with product discovery and the relevant Shopify integration route when planning that setup. Confirm the actual connection, selected product and destination before making a pricing or cart promise. Installing an assistant is one step; checking the shopper's complete path is another.

Measure answer quality and buying outcomes separately

Start with the number of comparable cases reviewed and how many were supported, contradicted or unresolved. Keep price mismatches separate from missing variants, unknown promotion eligibility and source failures. Record the source window and any cases you could not inspect.

Then examine merchant clicks, eligible cart actions and purchases where the connected events support them. Keep each denominator visible. A correct price answer does not prove a purchase, and a purchase does not prove the assistant caused it. The useful next move is to fix the specific source, answer or handoff failure and run the affected case again.

Explore Brambles.ai

Learn more about our AI-powered agentic commerce platform, agentic shopping, and shopping assistance solutions.

Explore More Insights

Discover more articles on AI, automation, and business innovation

View All Articles