Can I train a chatbot on my store without connecting an integration?

Yes. Website scraping alone produces a capable chatbot. A native integration adds live product search and order status, but a well-configured scrape covers product descriptions, policies, and FAQs very well.

What is the difference between URL filtering and element filtering?

They are two separate levels. URL filtering decides which pages enter training (Include only URLs that contain / Exclude URLs that contain). Element filtering decides which parts of each scanned page become training content (Include element IDs / Exclude element IDs using #id and .class selectors).

How do I stop the chatbot from learning menus and footers?

Use Exclude element IDs on product pages to remove headers, footers, navigation, related products and sidebars - so only the product's own content is trained. This is element-level filtering, applied to pages that already passed URL filtering.

What is the difference between an integration and scraping?

Scraping captures a static snapshot of your pages. A supported integration can add product search and order lookups. Available actions and refresh behavior depend on the integration; some product search uses a synchronized index. Scraped pages remain a snapshot.

Training a bot on an e-commerce store without an integration

Before you scan product pages: for products, an XML product feed (Google Merchant format) or an uploaded XLSX or CSV product file with AI Search is the better route. It works without a native integration, does not use training characters, and avoids prices and stock going stale between scans. See Integration with a product feed and AI Search. Use the scraping method below for informational pages, or for products when a feed or file is not an option.

You don't need a native integration to train a capable chatbot for your online store. With website scraping alone - configured properly - ChatLab can answer detailed questions about your products, shipping, returns and policies.

This guide shows the recommended setup: two separate trainings (informational pages and your product database) and the two levels of filtering that keep each one clean.

Integration vs. scraping - what each gives you

Scraping (no integration) - a static snapshot of your pages. Great for product descriptions, specifications, FAQs, shipping and return policies, and general store information. It does not know live stock or order status.
Integration - can add product search and order lookups, depending on the provider and enabled actions. Catalog search may use a synchronized index, so verify each integration's refresh behavior before promising real-time stock.

The best results come from keeping the two apart: scraping for informational pages, and an integration, a product feed or a product file for products. If none of those is possible, a well-configured scrape is a workable baseline.

If your store later connects a supported e-commerce integration (WooCommerce, Shopify, PrestaShop and others), its synchronized catalog appears separately under Training > E-commerce catalog, on the Products and Categories tabs. That catalog is synced from the integration - it is not the same as the scraped pages described here, which remain a static snapshot.

Where to find the filtering settings

First create your chatbot and check training capacity. Select it and navigate to Training > Links, click Add New Training: Website, then expand Advanced settings. Inside you'll find the accordions used throughout this guide: URL Filtering (Include only URLs that contain / Exclude URLs that contain), Query Parameters, Content Filtering and Scraping Behavior.

The two levels of filtering (this is the key concept)

ChatLab gives you two independent levels of control in a website training's advanced settings. They are easy to confuse, so keep them separate in your mind:

URL filtering - which pages enter training. Decides which addresses the crawler keeps. Set with Include only URLs that contain and Exclude URLs that contain (URL substrings, semicolon-separated, e.g. /product;/produkt).
Element filtering - which parts of each kept page become content. Applied after URL filtering, only to the pages that passed it. Set with Include element IDs and Exclude element IDs (CSS selectors: #id for IDs, .classname for classes, [data-hook="description"] for attributes, semicolon-separated, e.g. .product-description;#tab-description).

Think of it as a funnel: URL filtering picks the pages, then element filtering trims each page down to the valuable content.

Keep less, but more valuable content

RAG chatbots work best when each trained chunk is focused. On a product page, the product's own description is valuable - the surrounding menus, footers, "related products" and upsell blocks are noise that dilutes retrieval and burns training characters. Train on the product content only.

Set up two separate trainings

Splitting your store into two website sources keeps each one clean and easy to tune. Each training uses its own URL filtering (level 1) and element filtering (level 2).

1. Informational pages

Goal: keep the pages that explain your business (About, Contact, Shipping, Delivery, Returns, Warranty, Payment, Privacy, FAQ).

Level 1 - URL filtering: in Exclude URLs that contain, list your product and category patterns, e.g. /product;/products;/shop;/category;/collections. This keeps product pages out of this training.
Level 2 - element filtering: usually light here. Optionally use Exclude element IDs to drop the global header/footer, e.g. header;footer;#navigation.
Prefer training from your sitemap for predictable coverage.

2. Product database

Goal: one focused training over product pages, carrying only each product's own content.

Level 1 - URL filtering: in Include only URLs that contain, list only your product patterns, e.g. /product;/products. Everything else is skipped.
Level 2 - element filtering: this is where you keep only the product content. Use Include element IDs to capture the product information block, and/or Exclude element IDs to drop store chrome (header, footer, navigation, related products, upsells, sidebars). See the per-platform selectors below.

Test each training before expanding it

Start with Single address and a representative informational or product page, using the appropriate element filters.
Click Start scanning the website and wait for Training completed. A small source may finish in a few minutes; larger scans can take longer and continue in the background.
In Training > Links, click View and check that the product name, description, relevant attributes, price context and useful links remain. Excluding images can remove references needed for visual cards.
Apply the verified configuration to the corresponding sitemap or full scan. Use disjoint URL patterns so the informational and product trainings do not overlap.
Test product, shipping and return questions in Overview. Schedule retraining if available, and tell visitors to verify changeable prices and availability on the store page.

Per-platform element IDs (level 2 - starting points)

These are values for the Include element IDs and Exclude element IDs fields (level 2). They are CSS selectors, semicolon-separated. Selectors vary by theme - verify against your own store and adjust. The pattern is always the same: include the product information block, exclude global chrome and cross-sell blocks.

These examples are theme-dependent candidates, not guaranteed platform selectors. Ask your website maintainer to inspect the actual product page if you are unsure. Shopify theme structure and WooCommerce template customization explain why layouts vary. A selector that matches nothing, or only a description block, can omit names, prices, variants or images; always use the preview check above.

WooCommerce

Include element IDs: .summary.entry-summary;.woocommerce-product-details__short-description;#tab-description
Exclude element IDs: header;footer;.related.products;.up-sells;.cross-sells;#secondary

Shopify

Include element IDs: .product__info;.product-single__description;.rte
Exclude element IDs: header;footer;.announcement-bar;.product-recommendations

PrestaShop

Include element IDs: .product-information;.product-description;#description
Exclude element IDs: #header;#footer;#_desktop_top_menu;.block-categories;.featured-products;#search_filters

Magento / Adobe Commerce

Include element IDs: .product-info-main;.product.attribute.description;#description
Exclude element IDs: .page-header;.page-footer;.nav-sections;.block.related;.block.upsell;.sidebar

Shoper

Include element IDs: .productfull;.productfull__description;#product-description
Exclude element IDs: .header;.footer;.navbar;.mainmenu;.product-suggested;.basket

CS-Cart

Include element IDs: .ty-product-block;.ty-product-feature;.ty-wysiwyg-content
Exclude element IDs: .ty-header;.ty-footer;.ty-menu;.ty-product-block__advanced-list;.ty-mainbox-title

IdoSell / IAI

Include element IDs: #projector_longDescription;.product_description;#detail_atributes
Exclude element IDs: #header;#footer;#menu;.suggestedProducts;.breadcrumb

OpenCart

Include element IDs: #tab-description;#product .description;.product-description
Exclude element IDs: #header;#footer;#menu;#column-left;#column-right;.related

BigCommerce

Include element IDs: .productView-description;.productView-details;#tab-description
Exclude element IDs: .header;.footer;.navPages;.productCarousel;.sidebar

Shopware

Include element IDs: .product-detail-name;.product-detail-description;.product-detail-description-text
Exclude element IDs: .header-main;.footer-main;.nav-main;.product-detail-cross-selling;.cms-element-product-slider

Wix Stores

Include element IDs: [data-hook="description"];[data-hook="product-description"];[data-hook="info-section"]
Exclude element IDs: header;footer;[data-hook="header"];[data-hook="related-products"]

Other platforms (osCommerce, Abicart, Selly, RedCart, AtomStore, and more)

The same principle applies: open a product page, find the container holding the product name, description and attributes, put that in Include element IDs, and put the global header, footer, navigation menus, sidebars and "you may also like" blocks in Exclude element IDs.

Tips

Choose an AI model suited to e-commerce reasoning in the bot's settings.
Re-train periodically so descriptions stay current (scraping is a snapshot).
Keep informational and product trainings separate so you can tune the URL and element filters of each without affecting the other.