AI shopping assistant that recommends your products
AI shopping assistant with personalized product recommendations: intelligent upsell and cross sell suggestions based on shopper queries and catalog data.
In short
An AI shopping assistant in StoreChart helps website visitors find the right product through interactive chat conversations. Built from a scan of your store's catalog, it suggests relevant items with accurate pricing and links, ensuring products you disable are never recommended.
Recommendations from your own products
The on-site agent uses your product list to recommend, too
- Recommendations happen inside the website chat the agent is already having
- Based on the product list built from your site scan, which you can edit
- Every recommendation includes a link to the product page, and the price when it appears on your site
- A product you switch off in the list isn't recommended, so you control what's offered
- No separate recommendation widget to install: it's part of the website chat
AI product recommendation features
Part of the same website chat that already answers questions
Turn on product recommendations in 3 steps
No separate component to enable: recommendations use your website chat setup
The scan builds your product list with name, price, specs and page link.
Edit details, delete duplicates and switch off products you don't want recommended.
Paste the line of code on your site, and from then on the agent can recommend a product whenever the conversation calls for one.
AI product recommendations: how the chat chooses
Example: a recommendation chat in a camping store
Illustrative only. A visitor writes, "Looking for a tent for two, not too heavy." The on-site agent doesn't throw a list at them:
- It asks one question, such as which season the tent is for.
- After the answer, it may ask one more, about budget.
- It recommends one or two products from the store's list and explains in a sentence why.
- It links to the product page, with the price as it appears on the site.
That's the difference from "similar products" blocks under a product page. The agent acts like a shopping assistant: it finds out what's needed first, and only then suggests something.
How the agent picks a product
For each message, the agent receives a short list of around a dozen products from the active products in the knowledge base.
- Matching: by each product's name, description, specs and keywords, against what was said in the chat.
- From the list only: the recommendation and its link must come from that list, so it doesn't send visitors to pages that don't exist.
- Broad questions: "What do you have?" usually gets a clarifying question, not a tour of the whole catalogue.
- No priority setting: you can't boost a product in the settings.
Product data that improves recommendations
A product with a generic name and no specs will rarely be picked, even if it's the best fit. After the scan, fill in the list with:
- Specs: sizes and materials.
- Keywords: in your customers' own words, such as "for two", "light" or "winter".
- Short description: who the product is for.
- Models, not variants: in a store with thousands of products that differ only by colour or size, describe models. The short list won't always contain the right variant.
Stock and prices in recommendations
The agent doesn't check stock against WooCommerce, Shopify or StoreChart. It recommends from what was scanned.
- Sold-out product: stays a candidate until you switch it off in the list.
- Price: the one read in the latest scan. A new sale appears only after another scan.
- Price ranges: a "from" price is stored that way in the list, and you can edit it by hand.
- StoreChart catalogue: if the scan found no products at all, the agent recommends from the catalogue, but without a link to the product page.
Rules that steer recommendations
Every rule you write is included in every answer. A few example rules:
- Ask about size or where it will be installed before recommending.
- Offer a cheaper option too when the visitor mentions a budget.
- Don't recommend more than two products in one answer.
- Don't recommend a certain category, such as made-to-order products.
Tracking recommendations
There's no separate recommendations report. The conversations list shows every chat with the page where it started, so you can see what the agent recommended.
- Wrong recommendation: usually caused by a thin description.
- Needless pointer to WhatsApp: the agent didn't find a product that does exist on your site.
- Sales: show in your orders and StoreChart's BI, without direct attribution to a chat.
Who it's for and what's not included
- Good fit: stores where choosing takes a question or two, such as furniture, sports gear, cosmetics, gifts and technical products.
- Unnecessary: a store with three products.
- Not included: adding to cart, creating a coupon and saving the visitor as a lead. The visitor buys from the product page as usual.
- Questions the agent can't answer: the visitor gets a link to the store's WhatsApp, where your team or the WhatsApp AI agent picks up.
- Where to manage it: the product list lives in the knowledge base, and the chat's look and line of code in the website chatbot settings.
Key integrations
Frequently asked questions about AI product recommendations
Is this a separate recommendation engine?
No. This is a capability of the AI agent in your website chat: it recommends a product during a conversation. It isn't a "customers also bought" widget on your pages, and it has no analytics or A/B testing of its own.
Can it recommend a sold-out product?
The agent doesn't check live stock. It recommends from your list of active products, so switch off a sold-out product in the list until it's back.
Is it a generic AI model that doesn't know my store?
No. Recommendations are based on your store's product list, descriptions and rules, the same information the agent uses to answer product questions.
Does this replace the website chat?
No, it's part of it. The same agent that answers questions in your website chat can also recommend a product when the conversation calls for one.
Can I promote a specific product?
There's no setting to boost a product. The agent chooses from products whose name, description, specs and keywords match what the visitor wrote, so a product with complete details has a better chance of being picked. You can also write a rule, for example to suggest the new model when it fits.
Does a suggested product include a link or image?
A link, yes. When the agent recommends a product that has a page on your site, it links to that page and includes the price if it's in the list, and the visitor sees the images on the page itself.
Can I turn off only the recommendations?
There's no separate switch for recommendations. You can switch off products in the list, or add a rule telling the agent not to recommend products, and it will keep answering questions.
Related capabilities
Help choosing a product, inside the chat
Recommendations from your own product list, inside your website chat