An eCommerce chatbot answers your customers over chat, on your website, WhatsApp or Instagram, without a person typing every reply. It works well for questions that come up again and again and have a clear answer: shipping, payment methods, exchanges or order status. It goes wrong when it makes things up, blocks customers from reaching a person, or decides exceptions that belong to your team.

This article covers which questions are worth automating and how human support, a rule-based chatbot and an AI agent differ. It also covers which limits to set and what to check before choosing a customer service chatbot for your online store. The focus is support: recommending products and selling over chat is a different job.

What is an eCommerce chatbot?

Whether you call it an eCommerce chatbot, a retail chatbot or an online shop chatbot, the term usually covers one of two tools:

Rule-based chatbot. It follows a flow you design in advance: a menu, buttons or keywords that lead to fixed answers. It’s predictable and easy to review, but it gets stuck when a customer types something the flow didn’t anticipate.
AI agent. It understands messages in the customer’s own words and looks for the answer in the information you gave it: policies, FAQs and your catalog. When it’s connected to your store, it can also use order data. It can ask a clarifying question or hand the conversation to your team when it can’t confirm something.

Many stores combine both: buttons for the most common topics and an AI agent for everything else. Either way, a chatbot is only as good as the information it draws on and the rules you give it.

What a chatbot can handle in your store

A support chatbot pays off where questions repeat and the answer already exists somewhere:

FAQs. Shipping times and costs, payment methods, store hours, size guides or product care. For example: “Do you deliver on weekends?”
Order status. The classic “where is my order?” With your store connected, the chatbot replies with the actual status and the tracking link. If the data isn’t there, it says so instead of guessing.
Returns and exchanges. It explains your policy, asks for missing details (order number, reason, a photo) and leaves the case ready for your team.
Cancellations. It tells customers whether an order can still be canceled under your policy and hands off anything that needs approval.
After-hours questions. It answers what it can at any time and leaves the rest organized for your team the next morning.

Today 20:15

Hi! If the size doesn’t fit, can I exchange it? 20:15

Hi! Yes: you have 30 days from delivery to exchange the size, and the first exchange is free. Here are the steps. 20:15

See how to exchange

Illustrative Instant Answers example on Instagram: Luma, a fictional store, answers an exchange question with its approved policy. Time frames and conditions are demonstration content.

What not to automate: a product that arrived damaged, a disputed payment, a policy exception or an angry customer. Those need judgment, and the chatbot should pass the conversation to a person.

Human support, rule-based chatbot or AI agent

You don’t have to pick just one. Each handles a different kind of question best:

Human support, rule-based chatbot or AI agent
	Human support	Rule-based chatbot	AI agent
Best at	Sensitive cases, exceptions and complaints	Short, predictable flows	Questions in the customer’s own words
Availability	Your team’s hours	24/7	24/7
What it needs	People, shifts and an organized inbox	Designing every branch of the flow	Approved information, store data and handoff rules
Main risk	Delays during peaks	Gets stuck on anything off the menu	Wrong answers if the information is outdated
Example	“My order arrived broken”	“1. Shipping 2. Returns 3. Talk to someone”	“Will it arrive by Friday?”

Comparison of three ways to handle chat questions: what each one handles best, when it’s available, what it needs, its main risk and an example.

If I order today, will it arrive by Friday?

11:02

Sorry, I didn’t understand that. Choose an option:

Shipping Returns Talk to someone

11:02

If I order today, will it arrive by Friday?

11:02

Hi! Orders placed before 2 p.m. ship the same day, and standard delivery takes 2 to 3 business days. If you order today, it should arrive before Friday.

See shipping policy

11:02

Illustrative webchat example from Luma, a fictional store: a rule-based chatbot doesn’t understand a delivery question and repeats its menu; an AI agent answers with the shipping policy. Names and delivery times are demonstration content.

In a 2024 Gartner survey, 64% of 5,728 customers said they would prefer companies didn’t use AI for customer service. Their top concern was that it would get harder to reach a person. The answer isn’t to avoid AI but to use it for repetitive work and always leave a clear path to your team. When a conversation moves to a person, its context should move with it so customers don’t have to repeat themselves.

Limits of a customer service chatbot

Before you switch it on, decide what it must not do:

Don’t make things up. Your store can be held accountable for what its chatbot says. In 2024, a tribunal in British Columbia, Canada, ordered Air Canada to compensate a passenger after its chatbot misstated the airline’s bereavement fare policy. The decision says the company is responsible for all the information on its website, and the chatbot is part of the website. Use only approved, current information, and prefer “I can’t confirm that” to a guess.
Don’t decide exceptions. Your team approves refunds, late exchanges and compensation. The chatbot can collect the details and explain the next steps.
Don’t pretend to be a person. Say it’s a virtual assistant. In the European Union, the AI Act’s transparency rules have applied since August 2026. They require that people know they’re talking to a machine.
Protect order data. Before sharing an order status or an address, confirm the person asking owns the order, and ask only for the data you need.
On WhatsApp, serve your customers. Meta’s terms for WhatsApp Business restrict companies that offer a general-purpose AI assistant as their main product. A chatbot that serves your store’s customers is part of your service, not the product you sell.

How to choose a chatbot for your online store

Eight questions to compare options:

Does it connect to your store? Without access to orders and products in Shopify, WooCommerce or VTEX, a chatbot can only answer general questions.
Does it work where customers message you? WhatsApp, Instagram, Messenger and website chat, with every conversation in one inbox.
Does it answer from information you control? Approved policies and documents that are easy to update, and an honest answer when it doesn’t know.
How does it hand off to your team? Check when it hands off (a request for a person, a complaint, an exception), to whom, with what context and what happens after hours.
What can it do, and what can’t it? Sharing a tracking link isn’t the same as canceling an order. Check which actions it allows and within which limits.
Can you test it before going live? Try ambiguous questions, out-of-scope topics and missing data before opening it to all your customers.
What does it measure? Conversations resolved without human help, handoffs, satisfaction (CSAT) and response times. It should also let you read real conversations.
What does it cost in total? Add up the plan, messages and channel fees. On WhatsApp, Meta charges per delivered message. Since October 1, 2026, each number gets 1,000 free service messages a month and pays from the 1,001st.

Instant Answers setup in Hellotext: uploads, incoming channels, escalation, tone, audience, properties to collect and agent prompt, with a panel to test the playbook.

Chatbots for Shopify, WooCommerce and WhatsApp

Shopify. For a Shopify AI chatbot, start with Shopify Inbox. It includes an optional AI agent that answers in your online store chat using your catalog, policies and published content. If that’s enough, it’s a good starting point. If customers also message you on WhatsApp or Instagram and you want every conversation in one inbox, connect Shopify to a multichannel platform.

WooCommerce. A WooCommerce chatbot usually comes through a plugin. The Hellotext plugin tracks orders, carts and store activity: the data a chatbot needs to answer questions about an order. You also add the webchat with the same plugin.

WhatsApp. When a customer messages you, a 24-hour window opens for free-form replies. Outside that window, you can only message them with templates approved by Meta. Whether it’s a Shopify WhatsApp chatbot or one for another platform, it needs access to orders and short handoff times. If your team takes too long to pick up a case, the window can close before anyone replies.

How Hellotext handles chat support

In Hellotext, chat support runs on playbooks: AI agents with a specific job that share the Inbox with your team. They reply on WhatsApp, Instagram and your website’s webchat; Messenger conversations also land in the Inbox.

Service & Experience playbooks in Hellotext: Instant Answers, Order-Update Delight, Review & UGC Builder, NPS Pulse, CSAT Pulse, Order Cancellation Assistant and Return & Exchange Helper.

Instant Answers answers FAQs from the information you approved and hands off what it can’t confirm.
Order-Update Delight answers “where is my order?” with your store’s data and shares the tracking link when there is one.
Return & Exchange Helper guides customers through your policy and passes anything that needs approval to your team.
Order Cancellation Assistant handles cancellation requests within your policy. With Shopify or VTEX connected, it handles permitted cancellations.
Custom Agent covers a job no other playbook does, with your instructions and documents.

Instant Answers is included from the Small plan, Custom Agent from Grow, and the three post-purchase playbooks in Pro. If you’d rather start with rules, the Starter plan includes the Journey Builder to build chatbots and automations. It has no monthly fee and charges 5% of sales attributed to Hellotext. When a playbook hands off a conversation, it moves to the Inbox with its history and is assigned according to the rules you set.

White Adidas Samba OG sneaker with black stripes — Illustrative Order-Update Delight example: it answers the status of order #1048 with a tracking button and, when the customer asks to change the address, hands the conversation to the team. Order details are demonstration content.

eCommerce chatbot FAQ

Will a chatbot replace my support team?
No. It handles repetitive questions and leaves your team the cases that need judgment: complaints, exceptions and customers who’d rather talk to a person.

What’s the difference between a chatbot and an eCommerce helpdesk?
A helpdesk organizes your team’s work: assignments, priorities and response times. A chatbot replies on its own. Many retail customer service software platforms combine both in one inbox.

Can I use an AI chatbot on WhatsApp?
Yes, to serve your own customers. Reply with free-form messages within 24 hours of the customer’s last message, and use approved templates outside that window.

How much does an eCommerce chatbot cost?
It depends on the plan, message volume and channels; on WhatsApp, Meta’s fees are added. Hellotext’s plans are on the pricing page.

How do I know if my chatbot is working?
Track how many conversations it resolves without human help, how many it hands off and why, and customer satisfaction after the conversation. Read a sample of conversations every week to find missing answers.

Start with the questions you answer most

Look through your team’s recent conversations and pick the ones with a clear answer. With those answers and your policies, turn on Instant Answers and decide when to bring in a person. If you sell on WhatsApp, see how Hellotext’s AI agents for WhatsApp work too.

Start for free or book a demo.

Back to resources

View all articles