Commerce · website AI assistant
AI chatbot for E-commerce
Updated
The short version
Answer sizing, shipping, returns, and stock questions straight from your own product and policy pages — with a link to the page each answer came from.
Most store support tickets are not complaints. They're the same handful of pre-purchase questions — *does this fit*, *when will it arrive*, *can I return it* — asked by people who are standing at the checkout with their card out. Every hour those questions sit in an inbox is an hour the cart sits abandoned.
The catch is that a shopper will not accept a confident-sounding guess about a return window. A wrong refund policy is worse than no answer at all. So the assistant that helps a store has to do two things at once: answer instantly, and show its work.
That's the trade MySiteGPT makes. It trains on your product pages, shipping and returns policies, and FAQ, answers only from that content, and attaches a source card to every claim so the shopper can open the actual policy page. When your content doesn't cover the question, it says so and offers to take an email instead of inventing a number.
What online stores get asked
Real questions, in visitors' own words. Every one of these is answerable from content you probably already publish.
- “Do you ship to Canada, and how much is it?”
- “What's your return window, and who pays return postage?”
- “Is this true to size? I'm normally a medium.”
- “Is this back in stock, and can I get notified?”
- “What's the difference between these two models?”
- “Can I change the address on an order I already placed?”
How MySiteGPT helps online stores
Pre-purchase answers in seconds
Sizing, materials, compatibility, and delivery questions get answered at the moment of hesitation instead of the next business morning — from the product copy you already wrote.
Policy answers you can defend
Every returns or shipping answer carries a [n] citation the shopper can open. If a policy changes, you edit the page and re-sync; the assistant can't keep quoting last season's terms.
Never invents a refund window
When retrieval confidence is low, the bot returns your fallback message rather than guessing — and logs the miss so you can see exactly which policy page is thin.
Captures the shopper who's still deciding
Out-of-stock or bespoke questions turn into a name and email with an instant alert to your inbox, so a dead-end question becomes a follow-up instead of a bounce.
What to train it on, in order
The answers are only as good as the index behind them. Start at the top — these are ordered by how much question volume they absorb per page trained.
- 1
Shipping, returns, and warranty pages
The highest-volume questions in retail are policy questions. Train these first — they're short, factual, and they retrieve cleanly.
- 2
Product and collection pages
Crawl them so specs, materials, and compatibility details are answerable in the shopper's own words rather than through your filter UI.
- 3
Sizing guides and care instructions
Often buried a click too deep to be found. In the index they become the answer to the single most common fit question.
- 4
Your existing FAQ, as Q&A pairs
Q&A pairs give you exact control over the wording of sensitive answers — the ones legal or finance would want to sign off on.
Not sure your pages are ready to be retrieved from? Run one through the RAG readiness checker first, or audit the whole site with the sitemap analyzer.
What to measure once it's live
We don't publish other people's numbers as if they were yours. Track these instead — they are all visible in your own dashboard within a fortnight.
- Share of chats resolved without an email hand-off
- Unanswered questions per week — each one is a missing or thin content page
- Leads captured from out-of-stock and bespoke enquiries
- Pre-purchase questions arriving in the support inbox, before vs after
AI chatbot for E-commerce: frequently asked questions
Does it work on Shopify, WooCommerce, or a custom storefront?
Yes. The widget is a single script tag that goes in your theme's header or before the closing body tag, and it renders inside a Shadow DOM so it never inherits or breaks your storefront CSS. That's the same install on Shopify, WooCommerce, BigCommerce, Webflow, or hand-rolled HTML.
Can it look up a specific customer's order status?
Not on its own — it answers from the content you train it on, not from your order database. It can explain how to check an order, what your fulfilment times are, and where the tracking link is sent, then capture the customer's email so a human can pick up the specific case.
What happens when we change our returns policy?
Edit the page and re-sync the source. Content re-syncs automatically every 24 hours at a time you choose, or you can trigger it immediately after a change so the assistant stops quoting the old terms right away.
Will it answer international shoppers in their own language?
It replies in the visitor's language — 50+ are supported — while still answering strictly from your English content and citing the English source page.
Try it on your own content first.
The chat bubble on our homepage is the product, trained on our own site — ask it anything and watch every answer cite its source. Then train one on yours: 14 days, full features, no card.