In short: A Google Merchant feed is a structured product file that lists a store's items with standard fields such as title, description, price, availability, image and link. Puhera syncs your product catalog from this feed, so the chatbot can answer questions about prices, stock and product details with current catalog data and recommend the right product.
What is a Google Merchant feed?
Google Merchant Center is the service stores use to show their products across Google's shopping surfaces. A product feed is the list of products you submit to it. Each item represents one product, and each field describes one attribute of that product.
A feed is usually an XML or tab-delimited text file, and most e-commerce platforms can generate one automatically. Many stores already maintain a feed for shopping ads or free listings. That makes it a ready-made, well-organized, machine-readable product source for a chatbot, too.
Why does a chatbot need a dedicated product source?
A chatbot's knowledge base is usually built from a website crawl and uploaded documents. Those sources are great for stable information such as return policies, shipping terms or payment methods. Product information, however, changes all the time: prices drop for a sale, items sell out and new products arrive.
A bot that relies only on crawled pages may be reading an outdated copy of a product page. A structured feed gives every product's price and availability in separate, unambiguous fields. That helps the bot answer questions like "Is this jacket available in a large?" or "Can you suggest running shoes under $80?" accurately.
Which feed fields matter for good answers?
Google's product data specification includes many attributes. These are the ones that most affect chatbot answer quality:
- id: the product's unique identifier. In many stores it equals the SKU, and it is the key for matching recommendations to sales.
- title: the product name. Titles that include brand, model, color and size make it easier for the bot to find the right item.
- description: the product description. If material, use cases and key features are here, the bot can answer "Is it waterproof?"
- price and sale_price: regular and discounted price. Keeping the sale price in its own field helps the bot describe a discount correctly.
- availability: stock status such as in stock, out of stock or preorder. This is what keeps the bot from recommending sold-out items.
- link and image_link: the product page and image URLs, needed to send the visitor straight to the product.
- brand, product_type and google_product_category: brand and category data, useful for questions like "What else do you have from this brand?"
- item_group_id, color, size: variant data that ties the different colors and sizes of one model together.
The richer and more consistent your feed, the clearer your bot's answers. Empty descriptions or titles like "Product 123" make the job hard for even the best AI model.
How do you keep prices and stock fresh?
A feed is only as valuable as it is current. If the price the chatbot sees differs from the price on your site, customer trust suffers. These practices reduce that risk:
- Generate the feed automatically: use the feed URL your e-commerce platform produces instead of maintaining a file by hand.
- Check before big sales: confirm the feed carries the new prices before a major promotion starts.
- Map availability correctly: make sure sold-out items are marked out of stock. Updating the status field is usually more consistent than deleting the item.
- Keep variants separate: when each size and color has its own id, the bot can say which variant is available.
- Test with real questions: after a sync, ask the bot about a few products in preview and compare the price and stock with your site.
What are the limits of a product feed?
A feed is a strong source, but it does not cover everything. Knowing its limits helps you plan the rest of your knowledge base:
- It contains product data only: return policies, delivery times and store rules are not in the feed. Use a website crawl, document uploads and an FAQ for those.
- It carries no order data: "Where is my order?" cannot be answered from a feed; that requires an integration with your order system.
- Stock usually arrives as a status: feeds typically say in stock or out of stock rather than an exact quantity, so the bot may not give unit counts.
- Freshness depends on regeneration: the bot sees data as current as the feed your platform produces.
How do a feed and a website crawl work together?
You get the best results when the feed works alongside your other sources, because each one answers a different kind of question:
- Product feed: "How much is this?", "Does it come in black?", "Can you suggest a similar, cheaper model?"
- Website crawl: "How long does shipping take?", "Where is your store?", "Which payment methods do you accept?"
- Documents and FAQ: size charts, user manuals, warranty terms and recurring special questions.
Picture a shoe store where a visitor asks, "Do you have this model in a size 9, and can I return it if it doesn't fit?" The bot takes availability from the feed and the return terms from the return policy page, covering both the product and the policy in a single answer.
Feed quality checklist
- Does every product have a unique, permanent id?
- Do titles include brand, model and a distinguishing attribute?
- Are descriptions filled in rather than empty or duplicated?
- Are sale prices in their own field?
- Does availability match what your site shows?
- Are variants linked by a shared group id?
How does Puhera use a Google Merchant feed?
Puhera's Google Merchant integration syncs your product catalog from your feed and adds it to the AI knowledge base. The catalog sits alongside your website crawl, PDF, Word and PowerPoint documents and FAQ as a source the bot answers from. Because every answer shows its source, you can see when product information came from the catalog.
At a high level, setup looks like this: add your feed, let the catalog sync, test a few product questions in preview and publish the bot. Stores running on T-Soft can also sync their catalog directly through the T-Soft integration.
A current catalog is also the foundation for product recommendations and revenue measurement. To see how recommended products turn into sales, read how to measure chatbot revenue and conversions. To prepare the rest of your sources, see our knowledge base guide.
Frequently asked questions
What should I do if a price in the feed differs from my site?
The bot reads product information from the synced catalog, so fix the price at the source of the feed, in your e-commerce platform. Once the catalog syncs again, ask the same question in preview to confirm the result.
What does the bot do when asked about a product that is not in the feed?
The bot answers only from the sources in its knowledge base. If a product is not in the feed or any other source, the right behavior is to say it does not have that information and, if needed, hand the visitor to a live agent.
Can a feed be used to check order status?
No. A feed contains product data only. Order lookup requires an integration that connects to your order system, such as T-Soft.