AI Chatbot for E-commerce: Answer Product Questions Without Hiring Support

If you run an online store in India, you already know the questions. “Do you deliver to Pune?” “What's your return policy?” “Is this available in medium?” “Can I pay with UPI?” They arrive at all hours — across WhatsApp, Instagram DMs, and your site's contact form — and they all have answers that are already written down somewhere in your store.

An AI chatbot trained on your store's documents can handle all of them automatically, at 2am, without you being awake. This is not about replacing customer support. It's about handling the 80% of questions that don't need a human, so you can focus on the 20% that do.

The questions a chatbot actually handles well

Not every question is a good fit for a chatbot. But the questions that come in highest volume are almost always the ones a chatbot handles best — they have consistent written answers that don't change often.

For most Indian e-commerce stores, that list looks like this:

• Return and exchange policy — conditions, timeframe, how to initiate
• Shipping timelines and cost — which pin codes you cover, estimated delivery time, free shipping threshold
• Payment options — UPI, COD, EMI, which banks, which wallets
• Product specifications — dimensions, materials, compatibility, size guides
• Availability — whether a size or colour is in stock (from your FAQ, not live inventory)
• Order process — what happens after checkout, how customers get tracking

These six categories cover the majority of pre-purchase and post-purchase questions your store receives. All of them have written answers. All of them can be uploaded as documents and made searchable by a chatbot.

The four documents to upload first

The quality of your chatbot depends entirely on the quality of what you upload. Before you configure anything, write — or compile — these four documents:

1. Return and exchange policy. Include the conditions (unused, original packaging, within 7 days), the exceptions (sale items, custom orders), and the exact steps a customer should follow. Write it the way a customer would ask about it — “can I return this?” — not just the formal policy heading.

2. Shipping policy. Which states or pin codes you deliver to. Standard delivery time. Express options if any. Cost and your free shipping threshold. COD availability and any associated charge. Cover every question you've ever received about shipping in one document.

3. Payment FAQ. Every payment method you accept. Whether COD is available everywhere or only for select pin codes. EMI options and minimum order value. What to do if a payment fails. Refund timeline back to the original payment method.

4. Product FAQ. The top 10–15 questions you get about your products. Sizes and fit if you sell apparel. Materials and care instructions if relevant. Compatibility if you sell electronics or accessories. This document changes most often — update it whenever your product range changes.

Why generic AI chatbots fail e-commerce stores

Most AI chatbots on the market do not restrict themselves to your documents. When a visitor asks a product question and the chatbot cannot find an answer in its context, it falls back on its general training — and produces a plausible-sounding answer it invented.

For e-commerce, that is a serious problem. A chatbot that confidently states the wrong return window, a delivery timeline you don't offer, or a product spec you don't carry will generate real orders — and real complaints — based on false information. A wrong answer in e-commerce does not just confuse a visitor. It creates a return request.

Minyut uses a cosine similarity threshold of 0.60. If no document chunk in your knowledge base scores above that threshold for a visitor's query, no LLM call is made at all. The chatbot responds: “I don't have that information in my knowledge base.” The visitor then knows to message you directly — which is the right outcome when the answer genuinely is not in your documents.

A chatbot that says “I don't know” is more trustworthy than one that guesses. In e-commerce especially, the cost of a wrong answer is measurable: a return, a dispute, or a one-star review. You can see exactly how these accuracy failures happen — and how to read your query logs to find gaps — in this breakdown of the four chatbot failure modes.

Embedding the chatbot on your store

Once you've uploaded your documents and configured the chatbot in Minyut's dashboard, embedding it takes one script tag. The widget runs inside a Shadow DOM — your store's CSS cannot affect it, and the widget's CSS cannot affect your store. It will not break your checkout page or your product pages.

For Shopify: paste the script tag into your theme's theme.liquid file inside the <head> block, or drop it into a Custom HTML section if you only want it on specific pages. On WooCommerce or any HTML-based store, the same one-line embed goes in your site header. A full step-by-step walkthrough for each platform is in the chatbot installation guide.

The chatbot appears as a floating button — bottom-right by default — and expands into a chat panel when clicked. On mobile it goes full-screen. Visitors can ask questions without leaving the product page, and return to browsing when they're done.

What this chatbot will not do

One thing worth being clear about: a document-based chatbot cannot pull live data from your order management system. It cannot tell a specific customer where their specific order is today — that requires a live integration with your shipping provider, which is a different category of tool entirely.

What it handles is static knowledge: your policies, your product information, your processes. For the majority of pre-purchase questions — the ones that make up 70–80% of your support volume — that is exactly what is needed.

Set this expectation in your chatbot's welcome message. Something like: “Ask me about our products, shipping, returns, or payment options. For order status, WhatsApp us at [number].” Visitors appreciate knowing what the chatbot can and cannot help with before they start typing.

If you want to see how it works with your own store's documents, start a free Minyut account — one chatbot, three documents, 100 queries per month, no card needed. Upload your return policy and shipping FAQ and you can test it against real visitor questions in under 10 minutes.