How to Add an AI Chatbot to BigCommerce

A BigCommerce shopper asks where an order is, whether a product qualifies for a return, or which size to choose. If your team answers every question manually, the same support work keeps interrupting sales and operations. This guide shows you how to add an ai chatbot to bigcommerce with AssistLoop, BigCommerce Script Manager, and a live storefront test.

The setup has five parts: train the agent on approved store content, prepare the widget, add the storefront script, test the result, and configure human handoff before sending shoppers to it. AssistLoop lists setup at approximately five minutes once your content and widget settings are ready.

Last verified: August 2026.

What you’ll need

Prepare these items before opening either dashboard:

• A BigCommerce store with access to Script Manager and the storefront where the widget should appear.
• An AssistLoop account with an AI agent ready for training.
• Approved product and policy content. Useful sources include website links, PDF, DOCX, and TXT files, pasted text, and exact Q&A pairs.
• One test product page, one shipping or policy question, one order question, and one question the agent should hand to a person.
• A desktop browser and a mobile browser for the live storefront check.

Your first content set does not need to cover every product. It does need to cover the questions customers ask before they buy and after they place an order. Start with product details, shipping, returns, refunds, warranties, and contact instructions.

AssistLoop lists setup at approximately five minutes after the agent content and widget settings are ready. The BigCommerce installation still needs a live storefront test.

Step 1. Create and train your AssistLoop agent

Create an AI agent in AssistLoop before you place its widget on the store. In the AssistLoop dashboard, select Create AI Agent, name the agent, and open Training Sources. Add the content a shopper is allowed to rely on, then test the answers in AssistLoop before you install anything in BigCommerce.

Use the source that fits the information in Train on your data:

• Use Website Links for public product, shipping, and store information.
• Use Upload Files for PDF, DOCX, or TXT documents.
• Use Paste Text for short policies or internal instructions.
• Use Exact Q&A when the wording must stay precise, such as a refund condition, warranty term, or return deadline.

Source quality controls answer quality. If the return policy is missing from the knowledge base, the widget has no approved source to use when a shopper asks about a return. Adding more model settings will not fix missing store content.

Build a small test set before moving on:

1. Ask for a specific product fact.
2. Ask about shipping or delivery.
3. Ask about returns or refunds.
4. Ask an order-related question.
5. Ask a billing dispute or account-specific question that should reach a person.

Use a new conversation for each test. Record any answer that misses a product detail or policy condition. Correct the source, then run that question again.

Step 2. Prepare the widget and copy the embed details

Open Widget Settings in the AssistLoop dashboard before adding the script to BigCommerce. The widget customization settings let you set the brand color, logo, avatar, agent name, greeting, light or dark theme, and suggested replies.

Each setting should answer a storefront problem:

• Brand color and logo: help shoppers recognize the widget as part of your store rather than an unrelated overlay.
• Agent name and avatar: tell the shopper what they are speaking with before the conversation starts.
• Greeting: direct the first interaction toward a useful store question, such as product fit, shipping, or returns.
• Light or dark theme: keep the widget readable against your storefront design.
• Suggested replies: give shoppers clear starting points, especially on product and policy pages.

Open Installation in the AssistLoop dashboard, then copy the approved value shown under Embed Script. The same screen shows the agent’s Agent ID. Record both values in your deployment notes. Use the script and Agent ID shown for this agent. Do not type a code example from a blog post or modify the script by hand.

If you have not created the agent yet, create your AI agent, add the approved store content, and return to this step with the script and Agent ID ready.

Step 3. Add the widget through BigCommerce Script Manager

BigCommerce Script Manager is the installation path for adding the AssistLoop storefront script without installing a traditional marketplace app or editing the theme directly.

Open the BigCommerce control panel and go to Storefront, then Script Manager, and select Create a Script. Enter the following settings for a storefront widget:

• Script name: AssistLoop.
• Description: AssistLoop AI support widget.
• Location: Footer.
• Select pages where the script will be added: All pages, or the specific customer-facing pages where the widget should appear.
• Script category: Functional.
• Script type: Script.
• Script code: the approved AssistLoop embed script copied in Step 2.
• Consent: use the Functional consent category so the script follows the store’s consent configuration.

Use the current field names and options in BigCommerce’s Script Manager documentation if your control panel presents a different form. The script location, category, type, consent setting, page visibility, and save action can change as BigCommerce updates the interface.

For storefront scope, choose the storefront or channel where the script should run. With Multi-Storefront enabled, check the storefront assignment before saving. A script assigned to one storefront does not prove that the same entry is active on the others.

Select Save or Save Script after reviewing the entry. If BigCommerce presents a publish action after saving, select Publish as well. Then check the entry’s status in Script Manager before opening the storefront.

The installation path depends on the storefront implementation. A Stencil storefront can use Script Manager for this storefront script. Catalyst storefronts need the script method supported by the Catalyst implementation, so confirm that Script Manager is available for the storefront before publishing. B2B Edition stores should check the storefront assigned to the B2B experience and test the customer-facing pages separately. Multi-Storefront stores must confirm scope for each storefront that should display the widget.

Do not assume that a successful save covers every storefront. Test each relevant storefront URL after publishing.

Step 4. Connect the Agent ID

The Agent ID identifies which AssistLoop agent the storefront widget should load. In AssistLoop, open Installation, copy the Agent ID displayed beside the approved Embed Script, and compare it with the value contained in the script you placed in Script Manager.

BigCommerce does not need a separate marketplace connection for this step. The Agent ID is part of the approved AssistLoop embed details. If you have multiple agents, make sure the script came from the agent trained on this store’s product and policy content. Using the wrong script can produce a working widget with the wrong knowledge base.

Return to Script Manager, open the AssistLoop entry, and confirm that the code field contains the complete approved script. Save the entry again if you changed it. Do not paste a second script beside the first one.

Step 5. Customize the widget

Return to Widget Settings in AssistLoop and check the storefront presentation before launch. Set the brand color and logo so shoppers recognize the widget. Give the agent a clear name and avatar so the conversation has an obvious identity.

Use the greeting to point shoppers toward product questions, shipping, or returns. Keep suggested replies tied to those first decisions. Pick the light or dark theme that remains readable against the BigCommerce storefront.

Open the widget customization settings if you need to review the available branding controls. Save the settings, then refresh the live storefront in a fresh browser session. The widget should reflect the saved greeting, styling, and suggested replies.

At this point, create your AI agent if you still need an AssistLoop account. Start with one storefront and a small approved content set before expanding the deployment.

Step 6. Publish and test the live storefront

Save or publish the Script Manager entry, then open the live storefront in a fresh browser session. Do not use only an unpublished theme preview. A successful test starts with the customer-facing URL where shoppers will see the widget.

Check the following on desktop and mobile:

• The widget appears on the intended page.
• The greeting and suggested replies match the current AssistLoop settings.
• The color, logo, avatar, name, and theme display correctly.
• A fresh conversation receives an answer from the trained agent.
• Product and policy answers match the approved store sources.
• The widget does not cover checkout controls, product selectors, cookie notices, or other mobile controls.

Run the test questions from Step 1 again. Review the conversation log in AssistLoop after each test. The log shows what the shopper experienced and gives you a place to find missing training sources.

Separate two checks that are easy to confuse. A visible widget confirms that the storefront script loaded. It does not confirm that the agent has the right content, uses the right Agent ID, or knows when to hand a conversation to a person.

Step 7. Configure human handoff before sending traffic

Some conversations should leave automation. A billing dispute, account-specific problem, or policy question that the agent cannot answer clearly should have a human path instead of another guessed response.

Human handoff is available on AssistLoop paid plans. It lets a visitor escalate during a conversation, and your team can pick up the thread in the shared inbox with its history. Test this from the live storefront by asking the agent a question that should be handled by a person, then confirm that the conversation reaches the inbox.

The Free plan includes 150 message credits per month and does not include human handoff. See AssistLoop pricing for the current plan details.

Last verified: August 2026.

Do not treat a failed handoff test as a BigCommerce installation problem until you confirm that the AssistLoop plan includes human handoff. A widget can load correctly while the selected plan does not provide the escalation path you are testing.

Troubleshooting the BigCommerce installation

Use this order: load, placement, content, then handoff. It keeps a missing widget from being confused with an agent training problem.

The widget does not appear

First, open Script Manager and confirm that the entry is saved, enabled, and published. Check that its storefront visibility includes the page you are testing, then open the live storefront rather than an unpublished preview. If the store has more than one storefront, verify that the entry is attached to the storefront you opened.

If the entry is present but the widget still does not load, copy the approved script and Agent ID from AssistLoop again. Do not repair the code by adding a second copy of the script. Correct the Script Manager entry, save it, and test in a fresh browser session.

The widget appears on the wrong storefront or pages

Review the Script Manager page-visibility settings and the storefront or channel scope. With Multi-Storefront enabled, test each relevant storefront separately. A script assigned to one storefront does not prove that the same entry is active on the others.

Also check the page types selected in Script Manager. If the widget appears on a product page but not on another customer-facing page, the issue may be visibility scope rather than the AssistLoop agent.

Placement differs on Catalyst or B2B Edition

Confirm that the Script Manager location selected for the script is supported by the storefront type. Catalyst and B2B Edition can have different storefront behavior from a standard Stencil store, so check the current BigCommerce Script Manager documentation before changing the deployment location.

The widget loads but answers are wrong

The storefront installation is working. Return to AssistLoop training sources, add the missing product or policy content, and ask the same question in a new conversation. Check the conversation log after the retest so you can compare the answer with the approved source.

For a concrete AssistLoop check, compare the failed conversation with the agent’s training sources. If a return answer omits a condition that exists in an uploaded policy, add an exact Q&A pair for that condition, start a new conversation, and compare the result with the approved wording.

Handoff cannot be tested

Confirm that the AssistLoop plan includes human handoff before changing the BigCommerce script. The Free plan has no human handoff, so a failed escalation test on that plan is an account capability issue, not proof that Script Manager is broken.

What to do after the widget works

Review the conversation logs after the first round of testing. Collect questions the knowledge base did not answer well, then add the missing product or policy source and retest the same question in a new conversation.

If your store needs customer details during support, add lead capture so the agent can ask for those details inside the conversation. This capability is available on paid plans.

Use Agent Actions only for documented workflows that you have connected. AssistLoop can call a REST API, book meetings through supported connections, and capture leads. Do not present a BigCommerce order-status or catalog action as built in unless you have connected and tested the required workflow.

Start with one storefront and one focused content set. Once the live widget answers the prepared questions, stays out of mobile controls, and reaches your team when needed, expand the coverage to other storefronts.

When you’re ready to create the agent, create your AI agent in AssistLoop, train it on approved store content, and return to BigCommerce Script Manager with the generated embed details.

FAQ

Can you add an AI chatbot to a BigCommerce store without installing a marketplace app?

Yes. BigCommerce Script Manager can add an AssistLoop storefront widget without a traditional marketplace app or direct theme edits. Check the current BigCommerce Script Manager documentation for the labels and storefront settings in your control panel.

How do you choose the right pages for a BigCommerce chatbot script?

Choose the customer-facing pages where shoppers need support, then test the live URL for each selected page. Multi-Storefront stores should confirm the storefront scope separately instead of assuming one Script Manager entry covers every storefront.

What content should an AssistLoop agent use for an online store?

Start with approved product information, shipping details, returns, refunds, warranties, and contact instructions. AssistLoop accepts website links, documents, pasted text, and exact Q&A pairs, so policy wording can be supplied in the form that best preserves its intended meaning.

Can AssistLoop send BigCommerce shoppers to a human during a conversation?

Yes, human handoff is available on AssistLoop paid plans and sends the conversation to a shared inbox with its history. The Free plan does not include human handoff, so verify the selected plan before testing escalation.

How long does an AssistLoop website setup take?

AssistLoop lists setup at approximately five minutes after the agent content and widget settings are ready. BigCommerce still needs a live storefront check, including page visibility, mobile placement, answer quality, and any handoff path you plan to use.