How to set up ChatLab for your e-commerce store

Properly configuring ChatLab is key to ensuring your chatbot delivers precise product information and real-time order updates. With the right setup, ChatLab can answer detailed product queries, fetch order statuses, and assist customers around the clock.

Before you start

Create and train your chatbot before sharing it with customers. For installation, you also need permission to edit and publish the website. See How to create your first chatbot.

What Is Important in a Chatbot for E-Commerce

Start with the customer experience you want to support. Product discovery, cart actions and order lookup are separate capabilities with different prerequisites.

Help a customer find a product

See AI Search for catalog indexing and data freshness, and Offer Cards for visual results.

Help a customer add a selected product

This requires a supported integration and storefront setup, not just a product card. Read Adding products to the cart and test the actual store cart before enabling the feature for customers.

Help a customer track an existing order

Read Checking order status for lookup requirements, privacy considerations and unsuccessful lookups. Order details do not by themselves enable refunds or changes to an order.

Configuration checklist

• Accurate product information - ensure customers receive correct details about specifications and offers
• Clear presentation of offers - highlight discounts, bundles, and promotions effectively
• Up-to-date data - provide real-time pricing, stock levels, and order status (with integration enabled)
• Order tracking & user data - allow customers to check order details and shipping status when integrated

How to Provide Product Information to ChatLab

Before teaching ChatLab about your store, decide how it will ingest product data:

• Website Scan: ChatLab crawls and indexes your product pages (descriptions, specifications, and images) into a static knowledge base. Ideal for rich context and detailed answers when your catalogue is relatively stable. Read more: Adding New Website Sources

• API Integration: Connect via Shopify, WooCommerce, PrestaShop, or other platform APIs to fetch live product data, pricing, stock levels, and handle order lookups in real time. Some integrations also support searching products by SKU or reference code. Best for frequently changing inventories and order management. Keyword search and order tools query the store API. Supported platforms also offer AI product search, which indexes a separate catalog for semantic matching. Its freshness depends on catalog refreshes and the platform's live lookups.

• Combine Both: Use a scan for deep contextual knowledge and keep integration active for up-to-date accuracy.

Available E-Commerce Integrations

Navigate to your bot and select the Connect tab to view all available integrations.

ChatLab supports the following e-commerce platforms:

• Shopify - product search (with in-stock filtering), order tracking
• WooCommerce - product search by name or SKU, category browsing, order tracking
• PrestaShop - product search by name or reference/SKU, variant support, order tracking
• Shoper - product search, order tracking
• Wix - order tracking
• CS-Cart - product search, order tracking
• Abicart - article search, order tracking
• BaseLinker - order lookup for a configured order source
• Product feed (XML) - AI catalog search without order tracking

ChatLab also supports hotel and booking integrations:

• Beds24 - room availability, property details, pricing
• Mews - support-assisted room categories, availability search, reservation lookup
• Hotres - apartment listings, availability checks
• IdoBooking - apartment booking, reservation management
• KWHotel - room information, availability and optional reservation lookup

Check the Shop Tab and Catalog

After connecting a supported store, open Shop to review its connection, product catalog, and search configuration. AI product search can be enabled and indexing started automatically when catalog auto-indexing is enabled and the bot has the required access and capacity. Existing bots may need manual training. Wait for training to finish and check catalog errors before testing. The Hotel tab provides the corresponding overview for supported accommodation integrations.

Scanning vs Integration: Which Is Better?

Website Scan

Pros:

• Full context AI research with rich answers
• Deep understanding of product descriptions

Cons:

• Requires manual or auto retraining to capture new or updated products
• May hit training-character quota when indexing large catalogs
• Potential misinterpretation of detailed specs
• May fail to extract information from tables or parameter lists

API Integration

Pros:

• Current prices and stock from supported API actions, subject to platform caching and availability
• Real-time order look-ups for tracking status and history
• SKU and reference code search available for WooCommerce and PrestaShop
• Shopify supports in-stock filtering to show only available products

Cons:

• Keyword search needs suitable search phrases; AI product search adds semantic matching when enabled and trained
• Dependent on API availability and rate limits
• Requires proper credentials and configuration
• Test the selected AI model with your actual catalog and customer questions

The Winning Formula: Combine Scan & Integration

Combine selected website sources with the appropriate integration. Schedule refreshes for indexed content, test price and stock answers, and treat checkout as the final confirmation of availability and price.

To achieve both deep contextual responses and live accuracy, configure ChatLab's custom behavior under Settings > Role & Behavior. Click Enable custom role and behavior instructions and add the following:

When a user asks about products or categories, first invoke API tools to retrieve current data; fallback to the scanned knowledge base only for contextual enrichment. After fetching live data, augment responses with detailed descriptions and recommendations from the knowledge base when relevant.

Read more about role customization: Role & Behavior Settings

Choose and Test Your AI Model

Use a model that handles your catalog and customer questions reliably. Test ambiguous product descriptions, exact product codes, variants, and order lookups. Compare quality and message credit usage before changing models.

Change the model in Settings > Model & Advanced.

Set Up Custom Instructions

These examples are for the enabled keyword-search actions. If AI product search replaces keyword search, adapt the instructions to the actions actually enabled in Settings > Actions instead of forcing an unavailable tool. Use your store's product language, not automatically English. Bot Settings save automatically; wait for the saved status before testing.

For WooCommerce

WooCommerce integration has searchProducts, getProductCategories, and searchProductBySku tools available. Add these instructions to your Settings > Role & Behavior:

First, classify the user question into one of:
- PRODUCT-BY-SKU (user provides a SKU, reference number, or product code)
- PRODUCT-SPECIFIC (variant/price/availability/recommendation)
- ASSORTMENT-OVERVIEW (brands we carry, categories, "do you sell X at all", broad offer)

If PRODUCT-BY-SKU:
- Call searchProductBySku with the provided SKU or reference code.
- Return the product details to the user.

If PRODUCT-SPECIFIC:
- Call searchProducts(text) first using a concise query in the shop's product language.
- If results are empty/weak:
  1) Call getProductCategories(normalized noun) to choose best categoryId.
  2) Call searchProductsInCategory with the chosen category ID and search phrase to show a few items as examples.
- Answer using tool results as primary truth (name/price/availability). Use Knowledge Base only to add extra details if consistent.

If ASSORTMENT-OVERVIEW:
- Build the overview primarily from tools, then enrich using Knowledge Base:
  1) Call getProductCategories(normalized noun) to find relevant categoryIds.
  2) For up to 3-5 top categories, call searchProductsInCategory with the chosen category ID and search phrase.
  3) Aggregate from tool results; always label as non-exhaustive.
  4) Enrich the overview with Knowledge Base ONLY where it adds coverage the tools may miss.
- If tool results and KB conflict: prefer tool for concrete product facts; prefer KB for store-level statements.

For Shopify

Shopify integration has searchProducts tool with an option to show only in-stock products. Enable the "In-stock only" toggle in the Shopify product search action configuration to filter out unavailable items automatically. Add these instructions:

First, classify the user question into one of:
- PRODUCT-SPECIFIC (variant/price/availability/recommendation)
- ASSORTMENT-OVERVIEW (brands we carry, categories, "do you sell X at all", broad offer)

If PRODUCT-SPECIFIC:
- Call searchProducts first using a concise query in the shop's product language.
- If results are empty/weak, try up to 3 reformulations (simplify, synonym, category).
- Answer using tool results as primary truth; use Knowledge Base only for relevant extra context.

If ASSORTMENT-OVERVIEW:
- Use Knowledge Base as the primary source.
- Do not rely on searchProducts for full lists; optionally call it only to provide a few examples and label them "examples".
- Always be explicit when a list is non-exhaustive (e.g., "Examples include ...").

For PrestaShop

PrestaShop integration supports searchProducts and searchProductByReference tools. The reference search allows customers to find products by SKU, EAN-13 barcode, ISBN, or MPN code. Add these instructions:

First, classify the user question into one of:
- PRODUCT-BY-REFERENCE (user provides a SKU, EAN barcode, ISBN, MPN, or reference code)
- PRODUCT-SPECIFIC (variant/price/availability/recommendation)
- ASSORTMENT-OVERVIEW (brands we carry, categories, "do you sell X at all", broad offer)

If PRODUCT-BY-REFERENCE:
- Call searchProductByReference with the provided reference code.
- Return the product details to the user.

If PRODUCT-SPECIFIC:
- Call searchProducts first using a concise query in the shop's product language.
- If results are empty/weak, try up to 3 reformulations (simplify, synonym, category).
- Answer using tool results as primary truth; use Knowledge Base only for relevant extra context.

If ASSORTMENT-OVERVIEW:
- Use Knowledge Base as the primary source.
- Do not rely on searchProducts for full lists; optionally call it only to provide a few examples and label them "examples".
- Always be explicit when a list is non-exhaustive (e.g., "Examples include ...").

For Abicart, CS-Cart, and Shoper

These integrations support product search, but action names vary. Use the enabled product-search action in place of searchProducts in this example. Wix and BaseLinker currently provide order lookup, not product search. Add these instructions:

First, classify the user question into one of:
- PRODUCT-SPECIFIC (variant/price/availability/recommendation)
- ASSORTMENT-OVERVIEW (brands we carry, categories, "do you sell X at all", broad offer)

If PRODUCT-SPECIFIC:
- Call searchProducts first using a concise query in the shop's product language.
- If results are empty/weak, try up to 3 reformulations (simplify, synonym, category).
- Answer using tool results as primary truth; use Knowledge Base only for relevant extra context.

If ASSORTMENT-OVERVIEW:
- Use Knowledge Base as the primary source.
- Do not rely on searchProducts for full lists; optionally call it only to provide a few examples and label them "examples".
- Always be explicit when a list is non-exhaustive (e.g., "Examples include ...").

How to Properly Scan Your E-Commerce Website

Plan scanning your website in portions rather than all at once for better control. Scanning has two filtering levels: URL filtering (which pages enter training) and element filtering (which parts of each page become content). The checklist below covers the essentials.

1. Define proper scanning scope - use URL exclusion filter:

   • Include only URLs for product and category pages (e.g., /products/ and /collections/)
   • Exclude low-value paths like /blog/, /tags/, and archive pages
   • For multilingual stores, restrict scan to a primary language

2. Exclude repeating elements - use HTML Element Exclusion filter for header, footer, sidebar

3. Include only relevant content - use CSS Selector Inclusion to focus on product descriptions and specifications while excluding reviews, ratings, similar products sections

4. Use Sitemap if available - helps identify all product and category pages more efficiently than crawling

For the full method - both filtering levels explained in depth, the recommended split into separate informational and product trainings, and ready-to-use element selectors for WooCommerce, Shopify, PrestaShop, Magento, Shoper, CS-Cart and more - see Training a bot on an e-commerce store without an integration. It is especially useful when you can't (or don't want to) enable an integration.

Read more about optimizing your website scan: How to Reduce Training Characters When Scanning a Website

Additional Features Worth Enabling

Offer Cards

Enable this setting under Settings > Model & Advanced to allow the chatbot to return product cards with names, images, and prices. This improves visual presentation, especially with real-time API queries.

Add this to your custom instructions to avoid duplicate links:

If knowledge base provides URLs of the source information, provide them as comma separated list in the response as a reference - only after you have answered user's question in the conversation and only if the links are not included in the product list

Read more: Offer Cards

Suggested Questions

Pre-populate quick-reply chips (shipping, returns, opening hours) under Settings > Chat Conversation. Enable dynamic suggested follow-ups for AI-generated suggestions.

Read more: Dynamic Suggested Follow-ups

Human Hand-off

Activate the Contact Human form and provide your support email for escalations.

Read more: Human Support Contact Form

Conversation & Client Summaries

When enabled, ChatLab automatically generates a summary for each conversation and maintains an up-to-date profile summary for returning customers.

Read more: Chatbot Summaries and Memory

Chat Frontend API

You can pass visitor context such as a user ID or email through the frontend API. Browser-supplied context is not proof of identity and does not automatically authorize order-history access. Never expose store API keys or private security tokens in widget configuration.

Read more: Chat API

Language Setup

ChatLab supports 90+ languages. Read more: Chatbot Language Setup

When using integration with Shopify, WooCommerce, or other e-commerce platforms, consider what language your system returns products in. Add this instruction to your Role & Behavior settings:

When searching for products use [shop language] product and category names in singular form.

Monitor & Iterate Daily

To ensure ChatLab remains accurate and effective:

• Review conversation logs regularly. Fine-tune answers or add content where the bot hesitates. Read more: Chatlogs
• Enable AI Conversation Insights to identify patterns and issues. Read more: AI Conversation Insights
• Leverage the automated Daily Activity Report. Read more: Daily Summary Email

Optimizing Chatbot Responses

Even well-configured chatbots benefit from regular fine-tuning. Analyze conversation logs to identify weak spots, misinterpreted queries, or repetitive fallback answers. Refine prompts, update training material, and experiment with system instructions.

Read more: How to Improve Chatbot Responses

Related articles

• Adding chatbot to your website
• Training a bot on an e-commerce store without an integration
• Integration with a product feed