Home › Blog

AI shopping South Africa preparation starts with accurate product information: the exact item, price, availability, delivery conditions and seller must agree across the product page, submitted data and checkout. Fix contradictory facts before buying more traffic or paying for an AI shopping integration. Eligibility is not a guarantee of recommendation or sale.

By David Gous | Official guidance checked: 4 November 2026

A customer finds the right dress, selects a size and reaches checkout. The price changes, the size disappears or a delivery charge arrives without warning. The problem is already serious before anyone asks whether an AI assistant can recommend the shop.

This guide introduces the Easy Rank Product Truth Check, an editorial worksheet that compares five customer decisions with the systems supplying the answers. It is not a platform score. The examples are fictional and do not describe audited clients or measured revenue gains.

Check identity, price, availability, delivery and confidence for one exact item. Record evidence, not a vague pass mark for the whole store.

Choose a meaningful test product: a popular item, a product with several sizes or a bundle with complicated delivery. Keep its identifier and selected variation constant throughout the check. Comparing the cheapest size with a different size at checkout produces a misleading finding.

Keep a note of when each source was checked. A cached page or delayed data-source refresh may explain a discrepancy, but the explanation does not remove the customer problem. Identify which system owns the fact and which systems copy it.

Use consistent product identifiers and identify variants honestly. Never invent a GTIN to make a warning disappear.

Start with your own stock records. The product name, SKU, supplier reference and relevant identifiers should refer to the same physical item. Google’s product data specification explains supported attributes and their requirements. Apply the rules for your product, not a copied template from an unrelated retailer.

A rooibos seller should distinguish a box of tea bags from a bulk loose-leaf bag. Net weight, pack count and processing type should be explicit. A photograph of several packs must not imply a multipack when the price buys one. Do not infer certification from the appearance of packaging or another supplier’s catalogue.

Preserve established identifiers unless there is a reason to change them. Document replacements and mapping. Repeatedly renaming the same product creates maintenance work and makes comparisons harder. The goal is a dependable record that a customer, staff member and connected system can recognise.

A product family can be available while the advertised size or colour is sold out. Test the variation, not just the parent page.

Google’s availability specification distinguishes stock, out-of-stock, preorder and backorder states. The submitted status must reflect the actual offer and agree with the website. Do not describe an item as immediately available merely because you accept an enquiry about sourcing it.

Suppose the shop has three medium blue shirts and no large blue shirts. A customer selecting large should not be encouraged to pay for stock that does not exist. If backorders are intentionally accepted, explain the arrangement and expected timing accurately. A general “available” badge is inadequate when the selected option changes the answer.

Check basket quantities too. Staff need a clear process for reconciling physical stock, online orders and sales made elsewhere. Software cannot keep accurate availability when the underlying record is not maintained. Connect this work to the ownership checks in our website handover guide.

A variant should have enough detail for the reader to understand exactly what changes and what stays the same.

Google’s size attribute guidance describes using size information for applicable products. A size label alone may not resolve customer uncertainty. Add meaningful measurements and explain whether they refer to a garment or a person. Do not promise fit from an illustration.

For a hypothetical blouse, the useful information includes the selected size, colour, fabric, garment measurements and care instructions. For tea, it may include bag count, net weight and whether the product is loose leaf or portioned. Keep units visible and avoid describing materially different packs as one interchangeable item.

If your store uses variants, review Google’s ProductGroup guidance with the implementer. The markup should describe the actual page structure and relationships. Do not paste a large example into a theme without checking what WooCommerce and the existing SEO integration already produce.

Customers should be able to understand where you deliver, what it costs and what conditions apply before committing to the purchase.

Google’s shipping specification covers supported cost and timing information. Use account-level or item-level settings appropriately. The relevant requirements vary by destination and programme, so verify the current guidance rather than borrowing another country’s settings.

A hypothetical Bellville collection point should state its real address, hours, confirmation process and any collection deadline. Do not promise same-day collection when stock is held elsewhere. A remote-area surcharge or an excluded delivery destination should not emerge as a surprise after the order is placed.

Test one supported urban destination and one destination with different conditions. If you do not deliver to a location, make that explicit. Use our service-area planning guide for honest coverage descriptions. A list of South African towns is not proof of fulfilment capacity.

State who sells the item, its condition and where the customer can get help. Keep the applicable policies accessible and consistent with the offer.

Google’s condition attribute distinguishes supported product conditions. A refurbished item, clearance garment and new retail pack should not be described interchangeably. Explain material defects or exclusions factually and obtain appropriate advice on legal policy requirements.

The customer should be able to find the return route, relevant time conditions, warranty information where applicable and a functioning contact method. Do not copy an overseas returns policy and assume it fits the business or South African requirements. This article is a data-quality guide, not legal advice.

Where customers enquire through WhatsApp, explain what information to send and who responds. The WhatsApp Business SEO checklist helps test that handover. A message button is useful only when it reaches the right business and the business can handle the enquiry.

Google’s merchant-listing structured-data guide explains relevant Product and Offer information and validation. Ask which component generates the markup. Multiple plugins producing conflicting offers should be investigated before more markup is added.

Inspect a representative page after selecting the relevant option. Check the supported validation tool, visible content and submitted data separately. A technically valid fictional review is still fictional. Do not manufacture ratings, identifiers, testimonials or availability to satisfy a checker.

Make the core information readable in ordinary page text. The AI agent readiness guide explains why access and understandable content matter alongside structured information. No special “AI shopping schema” can repair an inaccurate offer.

Start with one product and a recoverable record. Change only the relevant settings, then test the customer journey before expanding the work.

In the product editor, identify whether the item is simple or variable. For a variable product, inspect its attributes and variations, then the selected variation’s price, stock and image. Menu labels can differ by editor and extension. Follow the installed setup and WooCommerce’s variable-product documentation.

For our fictional blue-shirt example, select medium, confirm its SKU and price, inspect available quantity and add a permitted quantity to the basket. Attempt a quantity beyond the available amount without placing an order. Confirm the system responds appropriately to the store’s actual stock policy. Check delivery and the final item details.

Then inspect the corresponding submitted product record and any reported Merchant Center issue. Record the refresh time and compare again. Use a designated test process or staging where a change could affect orders; do not place a chargeable order merely to prove that a button works.

No. Supplying current information and meeting applicable requirements does not guarantee selection, recommendation order or a purchase. Verify the actual integration and country conditions.

Apply the same information-quality principles, but verify each platform’s integration separately. Do not assume that a documented Shopify connection automatically applies to WooCommerce.

Only represent the actual offer accurately. A lower price for another size, quantity or condition does not establish the price of the selected item.

No. Start with representative products and the most serious customer-facing discrepancies. Fix the source and repeat the check before expanding the scope.

Choose a product customers already care about. Complete the five-part check, repair the first material contradiction and record the retest. A larger catalogue project becomes easier to scope once the source of the problem is understood.

Discuss a product information and website review with Easy Rank. Send a public product URL and the customer problem, not passwords or customer records. The scope can cover the product page, selected data source and relevant purchase or enquiry route.

Official Google, WooCommerce and OpenAI guidance was reviewed on 4 November 2026. Platform features and eligibility can change. This article does not claim a live merchant-account audit, an AI recommendation experiment, a measured conversion gain or universal South African feature availability.

David Gous is the founder of Easy Rank, a division of Bizconnexity, working with South African businesses on websites, SEO and AI visibility. The Product Truth Check is an original editorial worksheet. Local examples are fictional; generated and library illustrations do not depict audited products or clients.