Migros brings MAYA AI into ChatGPT, what does a 60% add-to-cart rate mean for retailers?

The announced tasks fall under six headings: building a personalised basket, suggesting offers, adding products straight to the basket, sharing cooking recipes, analysing past purchases and reminding users of items that are running low.

In classic online shopping, the customer searches, opens a category, opens a product page, then adds to basket. In chat, those steps melt into one window. Every extra step is a door the customer can walk out of. Fewer doors means a shorter distance between intent and basket.

Right now MAYA AI works through written dialogue. Voice commands are said to be coming “soon”, with no firm date given.

A basket item can be dropped at checkout, or the basket abandoned. Conversion rate, a separate metric, shows how many visitors actually finish an order.

Because suggestion and intent meet on one screen. When the suggestion is accurate, the customer stops searching. When it misses, the customer walks out the door.

Orçun Onat, CEO of Migros One, describes AI as, in our translation, “a strategic element that fundamentally transforms the customer experience”. Migros presents itself as the first food retailer to integrate directly into ChatGPT, in partnership with OpenAI. Whether that means Türkiye or the world is unclear, so read it as the company’s own claim.

A business with a clean product list does not start from zero when a new channel opens. The same data can feed a website, a marketplace store and, later, an assistant, so the chance to grow average basket size arrives earlier.

The till record is usually just an abbreviated product name and a price. An assistant needs more to match a request like “gluten-free pasta”. The product is physically sitting on the shelf, but digitally it does not exist.

The producer is the first source of product information: ingredients, weight, use. A producer that does not pass this on in an orderly way slips to the back row of the chain’s digital shelves. Label claims also follow rules, which we look at in our piece on FİLE Market’s gluten-free Harras range.

The base is broad. E-commerce volume in Türkiye grew 52.2% in 2025 to TL 4.567 trillion and made up 19.5% of total trade (Ministry of Trade). According to 2026 data from TÜİK, the Turkish Statistical Institute, the share of people ordering goods or services online has passed 60% for the first time. That 60% is not MAYA AI’s add-to-cart rate; they measure different things.

The title should hold brand, product type, weight or volume and any variant. A till abbreviation like “Cheese 500” is vague for people and machines alike.

The assistant matches context, not just the product: which meal it suits, who it is for, what it goes with. Without that context, the product stays off the list even when the question is right.

Structured data marks up the price, stock and product name on a page in a standard format that machines can read. A product feed exports your whole product list to other platforms in a single file. At AINEO we work on the same principle: AINEO Motor runs calculation engines embedded in business articles on defined inputs. If the inputs are messy, the output cannot be trusted.

Every product needs a unique SKU code, current stock information and a consistent category. If stock changes several times a day, the system has to reflect that. An assistant built without this groundwork is like dressing the shop window when the till isn’t ringing.

The customer adds the suggested item, then finds out it is out of stock. Trust breaks on the first try, and next time the customer does not come back to that channel.

No, the same logic applies to store decisions. We cover a platform that lets mall tenants compare their sales with the mall average in our piece on ECE Türkiye’s ECEConnect.

Start with the items that move fastest at the till, because those are most likely what customers search for. Then move to the local or specialty products that set you apart, since they give an assistant the strongest reason to recommend you.

Write a natural question with your neighbourhood and a product you sell. Check whether your business appears, whether product details are right and whether address and hours are current.

The person who keeps the product list and the person who runs the website should sit at one table and update the data from one place. We also do this work, preparing product data for AI channels, as part of our AI consultancy.

The coverage says MAYA AI is reached by selecting Migros from the apps tab in ChatGPT. Whether the assistant is also available inside Migros’s own app is not clear from the news.

No. 60% is the share of MAYA AI product searchers who add the suggested item to their basket. No data on sales, orders or revenue growth has been released.

Migros’s setup is an integration built in partnership with OpenAI. For a small grocer, the realistic step today is to clean up product data and keep it consistent on readable channels such as a website and marketplaces.