“Order me some waterproof trainers in size 43, under 120 euros, to be delivered by Friday.” What sounded like science fiction just a short while ago is now part of everyday life: AI is doing the shopping. An AI shopping assistant searches dozens of shops in seconds, compares offers and adds the right product straight to the shopping basket.
This raises a new question for retailers: will my product even be selected? Whether AI agents in eCommerce recommend your product isn’t determined by your best product photo. It’s determined by your product data.
Agentic Commerce: How AI agents make purchasing decisions in eCommerce
Agentic Commerce refers to a form of shopping in which AI agents take charge of research, comparison and, in some cases, even the purchase itself. For retailers, this means that, alongside human customers, there is a second target group that operates in a completely different way.
People shop with their eyes. AI shopping agents, on the other hand, require specific, reliable information in order to understand, evaluate and compare products. The following factors are particularly relevant here:
Clarity: What exactly is the product in question?
Completeness: Are all relevant characteristics, such as size or colour, specified?
Availability: Is the product available in the desired variant?
Delivery promise: When will the goods arrive? And what are the delivery charges?
Terms and conditions: What are the price and returns policy?
For retailers, this means that it is no longer just a question of what information an online shop provides, but also whether AI agents can reliably find, understand and compare this information. If a piece of information appears only in the advertising text or on an image, the agent may well assume it does not exist. This brings the structured presentation of product information to the fore. Approaches such as Answer Engine Optimisation or Generative Engine Optimisation pursue precisely this goal: to present content in such a way that AI systems can reliably understand and process it. For online shops, structured product data is a key lever in this regard.
Step 1: Check the quality of your product data
Before you can optimise your product data, you need to know where you stand. Take a sample of 20 to 30 products from different categories – ideally your bestsellers – and check the following:
Does each product have a unique identifier? The GTIN or EAN and manufacturer number are the most important basis for agents to identify products beyond any doubt.
Are key attributes stored as separate fields or merely mentioned in the description text?
Are the values consistent? Does the same colour sometimes appear as ‘Blue’, sometimes as ‘blue’ and sometimes as ‘Navy’?
Are delivery times and availability specified precisely or only vaguely?
Do the details in the shop match those in the inventory management system and on marketplaces?
When carrying out your checks, look not only for individual missing details but also for recurring patterns. This is because gaps in product data quality rarely affect just individual items, but often entire categories. Identifying such systematic weaknesses early on allows you to tackle optimisation in a targeted manner, rather than having to laboriously correct individual products later on.
Step 2: Structured data in the online shop using JSON-LD and Schema.org
For AI agents to understand your product pages, they need the information in a standardised format. Two terms are key here:
Schema.org is a shared vocabulary agreed upon by the major search engines. The ‘Schema.org Product’ type defines how a product is described in terms of price, brand or availability.
JSON-LD is the format in which you embed this structured data into the source code of your product page, invisible to visitors.
Here’s what a simplified JSON-LD product looks like:
The following details are required on every product page: name, brand, GTIN, price (including currency) and availability. A welcome side effect: structured data in your online shop also forms the basis for rich snippets for products in standard Google Search. This means you simultaneously improve your visibility to AI agents and in standard search results.
Head of Development & AI at elio
Step 3: Maintain product attributes – ensure they are complete and consistent
Even the best markup is of little use if the underlying data is incomplete. Well-structured attributes form the basis for all subsequent steps. Three basic rules will help with this:
Facts in fields rather than running text: Properties such as ‘Waterproof: Yes’ or ‘Water column: 10,000 mm’ can be clearly interpreted by AI agents.
Consistent values: Use fixed terms for each attribute – for example, always ‘Black’ instead of ‘black’, ‘Black’ or ‘midnight black’.
Specify intended use and target audience: Details such as intended use or season help AI agents match products to suitable search criteria.
For products with variants, the following also applies: each variant requires its own structured product data with its own GTIN, price and availability. Otherwise, the agent will not know whether the size being searched for is actually available. For large product ranges, it is worth implementing a PIM system as a central data source at this stage at the latest.
Step 4: Specify delivery times, shipping and returns as structured data
‘Delivery by Friday’ was part of the original order. If the agent cannot reliably determine your delivery time, your product will be excluded from the selection, even if it were the best offer. Specifying the delivery time as structured data is therefore one of the most effective strategies, which is still missing from many online shops.
Schema.org provides dedicated fields for this in the product listing:
Information | Schema.org field | Example |
|---|---|---|
Delivery charges | shippingDetails → shippingRate | €4.90 |
Delivery country | shippingDetails → shippingDestination | DE |
Processing time | deliveryTime → handlingTime | 0–1 working days |
Delivery time | deliveryTime → transitTime | 1–3 working days |
Returns | hasMerchantReturnPolicy | 30 days, free of charge |
To ensure that this structured information is reliable for the agent, it must be up to date. An agent who sees ‘in stock’ but then finds the item is ‘sold out’ when placing an order will lose confidence in your shop. Stock levels and delivery times should therefore be retrieved automatically from the stock management system and not maintained manually.
Step 5: Check whether your product data is being recognised
You can check whether your markup is correct using free tools. The Schema Markup Validator shows which structured data is found on a page and where errors lie. Google ’s Rich Results Test also checks whether your data can be used by Google Search.
Pay particular attention to these common errors:
The price in the markup differs from the visible price, for example following a discount promotion.
Availability is set to ‘InStock’ regardless of actual stock levels.
Variants share a common markup without their own GTIN.
Mandatory fields are missing because a plugin only outputs basic data.
Only testing will reveal whether the processed product data actually works in practice. That is why you should not limit yourself to a one-off check. Schedule regular spot checks, particularly following updates to your shop system or major changes to your product range.
Checklist: How to make your shop AI-ready
Key points at a glance: Use this checklist to assess how well your shop is already prepared for AI agents.Weren’t you able to tick all the boxes? Well, now you know the key areas to focus on for your shop. Most shops still have room for improvement here, and that’s precisely where your opportunity lies to get ahead early on in Agentic Commerce.
Structured product data determines your visibility in AI search
AI agents are changing the way products are found, compared and purchased. They are not guided by the loudest advertising, but by the information they can reliably analyse. The more complete, unambiguous and up-to-date your product data is, the better an agent can understand your offering and factor it into a purchasing decision.
Structured product data is therefore no longer purely an IT issue. It is becoming a prerequisite for products to be considered at all in agent-driven commerce.
You should therefore check now how well your product data is already prepared for this shift. After all, anyone who only takes action once AI agents have taken over a significant proportion of product searches will start with an avoidable disadvantage.
Let’s review your shop together
Would you like to know how your products appear from an AI agent’s perspective? As an eCommerce agency specialising in AI and product data, we’ll analyse a selection of your product pages during a consultation. We’ll highlight the biggest gaps and outline how you can address them with a reasonable amount of effort.