An AI chatbot for ecommerce, because your store already answers this — just not at 2am
CustomerBot reads your product pages, shipping policy, returns terms and size charts, then answers shoppers in their own language — with the page it took the answer from attached.
Free plan · No credit card · One script tag to install
Store support
Online · replies instantly
What is an AI chatbot for ecommerce?
It is a chat widget on your storefront that talks to shoppers in ordinary language — answering shipping, returns, sizing and product questions, recommending between items, and handing over to a person when it should. The useful kind is trained on your own store pages, so it quotes your delivery window rather than a plausible-sounding one.
The line that matters is between content and live data. Most of what shoppers ask before buying is already written on your site. Order status and stock are not, and never will be — those need a connection to your systems.
How live lookups are wired upEight jobs it does in a store
Tagged by where the answer comes from, so nothing here is a surprise later.
- Live data
Order tracking and delivery status
The biggest single bucket for most stores.
- Live data
Stock and back-in-stock questions
Changes by the minute, so a crawl will not do it.
- From a page
Shipping costs, zones and cut-offs
Straight off your shipping policy page.
- From a page
Returns, refunds and exchange windows
Peaks after every sale and every Christmas.
- From a page
Sizing, fit and materials
The question that decides the sale.
- From a page
Product comparison and recommendation
From catalogue copy, for a shopper who described a need.
- From a page
Compatibility and “will this work with…”
Technical categories live and die on this one.
- From a page
Capturing the email when the answer is no
Out of stock is a lead, not a lost visitor.
Five questions, most of your ticket volume
Ecommerce support is unusually repetitive, and that is good news: repetitive questions are exactly the ones a grounded assistant handles well.
- 01
“Where is my order?”
Usually the single biggest bucket
Connect your order-lookup endpoint as an API tool and the assistant asks for the order number, calls your API mid-conversation and reads the real tracking status back. No copy-pasting between tabs.
- 02
“Can I return this?”
Peaks after every sale
Your returns policy page becomes the source of truth. When the window changes from 14 days to 30, you edit the page and re-crawl — the assistant is corrected everywhere at once.
- 03
“Will this fit?”
The one that decides the sale
Upload the size chart PDF. The assistant reads it, including the tables, and answers with the measurement rather than a link to a page the shopper then has to interpret.
- 04
“Do you ship to my country?”
Highest bounce risk
It detects the language the shopper writes in and replies in it, quoting your own shipping zones. A question asked in Spanish gets a Spanish answer from an English policy page.
- 05
“Is this in stock in blue?”
Time-sensitive
Point an API tool at your inventory endpoint. The model decides when to call it, passes the variant, and answers from live data instead of a cached page.
It learns your catalogue, not the category
A general model knows what a return policy is. It does not know yours. CustomerBot crawls the pages you point it at, splits them into passages, embeds them into a vector index, and retrieves the relevant ones before the model writes a word.
The practical consequence: when you edit a page and re-crawl, the assistant is corrected immediately. There is no retraining step, no fine-tune, and no window where the bot is quoting last season's policy.
How website training worksProduct pages
Crawl the whole catalogue from one URL
Shipping & returns
Policy pages, verbatim
Size charts
PDF upload, tables included
Care & materials
Paste plain text
Pre-sale Q&A
Exact wording you want repeated
Help centre
Bulk-train from a sitemap
Give it your order API and “where is my order” stops being a ticket
Describe the endpoint in a sentence. CustomerBot writes the tool schema, you fill in the URL, method and any auth header, and the assistant starts calling it whenever a shopper's question needs live data.
It can chain up to four tool calls in a single reply, so “where is my order and can I change the address” is one answer rather than two tickets.
See how API tools workWhat you type
“Look up an order by its number and return the status and tracking link.”
What gets generated
{
"name": "get_order_status",
"method": "GET",
"endpoint": "https://api.yourstore.com
/orders/{order_number}",
"parameters": {
"order_number": {
"type": "string",
"required": true
}
}
}Six things to judge any ecommerce chatbot on
The demos all look the same. These are the questions that separate them, in the order they tend to bite.
Where do its answers come from?
Scripted flows only say what somebody typed in. A general model will invent a returns window. Retrieval over your own store pages is the only one that handles a question nobody scripted and still gets it right.
Can it reach live order data?
Ask specifically how. A native platform app, or a tool you point at your own endpoint, are both fine answers — “it learns your website” is not, because order status was never on your website.
What does it do when it does not know?
The honest behaviour is to say so and offer a person. In a store, a confidently invented delivery date is worse than no answer at all, because someone acts on it.
Does it sell, or only deflect?
Deflection saves you support hours. Recommending between two products, or taking an email when something is out of stock, is where a chat widget makes money rather than saving it.
How does it handle your languages?
If you ship internationally, ask whether you maintain one knowledge base or one per language. The difference compounds every time you edit a policy.
What does peak week cost?
Check the price of the tier above yours before Black Friday, not during it. Per-conversation pricing is comfortable until your best month arrives.
One tag, any storefront
Paste this before </body> in your theme. Shopify, WooCommerce, BigCommerce, Webflow, or your own React storefront — same line.
<script src="https://customerbot.co/script.js" data-bot-id="YOUR_BOT_ID"></script>Questions store owners ask
What is an AI chatbot for ecommerce?
An AI chatbot for ecommerce is a chat widget on your storefront that talks to shoppers in natural language — answering questions about shipping, returns, sizing and products, recommending between items, and passing the conversation to a person when it should. The useful kind is trained on your own store pages, so it quotes your delivery window and your refund policy rather than a generic guess. Anything that changes minute to minute, like order status or stock, comes from a live connection to your systems instead.
Is there a free AI chatbot for ecommerce?
Ours has a free plan — one assistant, 50 conversations a month, no card — which is enough to crawl a real storefront and ask it your own returns policy before you commit. Free tiers across this category vary a lot in what they restrict, so it is worth checking whether the cap is on volume or on answer quality.
Can it recommend products?
It recommends from what it has read. Crawl your collections and it can steer a shopper who describes a use case towards the right item and link to it. It is not a merchandising engine ranking by margin or stock age — it is a well-read assistant that knows your catalogue copy and can reason about it.
Does the chatbot know my live order status?
Only if you give it access. CustomerBot lets you describe an endpoint in plain English — "GET my order status by order number" — and it generates the tool definition for you. The assistant then calls that endpoint during the conversation and answers from the response. Without a tool it answers from your published policies only, which is the safer default.
Will it invent a discount code to close a sale?
No. Answers are grounded in the content you trained it on and it cites the source it used. If nothing in your store content covers the question, it falls back rather than improvising — and the unanswered question shows up in your analytics so you know what page to write next.
Can it capture emails from shoppers who leave?
Yes. Lead capture is built in: the assistant asks for a name and email inside the conversation, saves it against the transcript, and you can filter by status and export the whole list as CSV.
How do I put it on my store?
One script tag before the closing body tag. The same tag works on Shopify, WooCommerce, BigCommerce, Webflow or a custom storefront. There is nothing to install and no app permissions to grant.
What happens during a traffic spike like Black Friday?
The widget is served from a CDN and the assistant answers concurrently — there is no queue. Your plan sets a monthly conversation allowance (5,000 on Hobbyist, 10,000 on Pro); if you run past it the bot keeps replying and shows a usage warning rather than going dark.
Installing on your platform
Keep reading
Point it at your store and see what it already knows
Paste your storefront URL, wait for the crawl, then ask it the question your shoppers ask most. It takes about two minutes.
Start free