AI Chatbot for Shopify: What It Can and Can’t Handle
What an AI chatbot for Shopify can handle (order status, product and returns questions), what it can't, the options compared, and how to set one up.
Qazi Ihsan
Xenova Solutions

Last updated: October 2026
An AI chatbot on a Shopify store can reliably handle the repetitive questions: order status, delivery times, returns policy, sizing and product details, as long as it answers from your own catalogue, policies and order data. It cannot safely make judgement calls, such as approving a refund outside your policy, calming an angry customer or settling a damaged-goods dispute. The stores that get value from one treat it as a first line that resolves the routine questions and hands everything else to a person, with the conversation attached.
This guide is for small and mid-sized Shopify stores in the UK and US weighing up whether a chatbot is worth it. It is built around the questions store owners actually ask in forums: will it answer accurately, can it do anything beyond reciting the FAQ, and will customers hate it?
What an AI chatbot can handle on a Shopify store
Many store inboxes are dominated by a small set of questions that have one correct answer already written down somewhere. That is where a chatbot does well:
- Product questions before purchase: materials, dimensions, sizing, compatibility, care instructions, and what is in the box. These come straight from your product pages.
- “Where is my order?”: status, tracking link and expected delivery, but only if the bot is connected to your order data (more on this below).
- Delivery and returns policy: costs, cut-off times, which countries you ship to and how to start a return, answered from your policy pages.
- Product recommendations: narrowing your catalogue down to two or three options based on what the shopper describes.
- Out-of-hours cover: answering at 11pm instead of leaving a question to sit until morning, when the shopper may already have left.
- Collecting details before a handover: order number, email, photos of a damaged item, so your team starts with everything they need.
Shopify’s free Inbox app now includes an AI assistant that its listing describes as connected to your catalogue, policies and orders. For many small stores that is the sensible place to start before paying for anything.
What it can’t (or shouldn’t) handle
The limits matter more than the features, because a confident wrong answer costs more than no answer.
- Exceptions to your policy. A bot should never promise a refund, discount or replacement that your policy does not already allow. In Moffatt v Air Canada (2024), a Canadian tribunal held the airline responsible for refund information its website chatbot gave a customer, and rejected the argument that the bot was responsible for its own words. Expect the same principle to apply to your store.
- Statutory rights, stated wrongly. If you sell to UK consumers, GOV.UK guidance says you must tell customers they can cancel an online order up to 14 days after it is delivered. Your bot must repeat your policy and those rights accurately, not paraphrase them loosely.
- Complaints and upset customers. Someone whose order is late for a wedding wants a person, not a script.
- Disputes and payments. Chargebacks, suspected fraud, payment failures and damaged-goods claims need human judgement and, often, evidence.
- Anything outside your data. If the answer is not in your catalogue, policies or orders, a well-built bot says it does not know and offers a person. A badly built one guesses.
There is also a technical limit people miss. Shopify apps can by default only read orders from the last 60 days; Shopify’s developer documentation says access to older orders needs a separate permission that Shopify must approve, and customer personal data needs its own approval too. If your customers often ask about orders placed months ago, check this before choosing a tool.
Answers or actions: the question that decides which tool you need
Many Shopify chatbot apps answer questions from a knowledge base and look up order tracking. Fewer can act inside your store: change a delivery address, cancel an unfulfilled order, create a return or apply a discount code. Before comparing tools, decide which you need.
| Option | Typical cost | Answers from your data | Takes actions in Shopify | Best for |
|---|---|---|---|---|
| Shopify Inbox AI assistant | Free app | Yes: catalogue, policies and orders | Limited; check the current listing | Small stores testing whether chat is worth it |
| Third-party chatbot app | Monthly subscription, often tiered by conversations, seats or features | Yes, from a knowledge base you upload | Varies by app; some can, many only answer | Stores with a helpdesk already in place |
| Custom-built assistant | From £2,500 for an MVP (our AI Chatbot MVP, 3–4 weeks) | Yes, from any system you connect | Yes, with approval rules you set | Stores with several systems, languages or unusual workflows |
Our rule of thumb: if most of your messages are “where is my order” and policy questions, start with an off-the-shelf option. A custom build earns its cost when the bot needs to work across Shopify plus a helpdesk, a warehouse or 3PL system and a returns platform, or when you want actions with checks, such as cancellations allowed only before fulfilment. For a fuller breakdown, see how much an AI chatbot costs.
Will customers hate talking to a bot?
This is a common objection, and it is fair. Shoppers dislike bots that loop, give generic links, or block them from reaching a person. They do not mind a quick, correct answer at midnight. The difference is in the design:
- Say it is an AI assistant in the first message. Being upfront avoids the feeling of being tricked.
- Always offer a person. Shopify Inbox, for example, passes the chat to your staff during your availability hours when a customer asks for a human, and gives them your store’s email address outside those hours.
- Hand over with context. The transcript, order number and any photos should go with the handover so the customer never repeats themselves.
- Keep the scope narrow. A bot that does five things well beats one that attempts twenty and fumbles half.
- Remember the conversation. A bot that forgets what was said two messages ago is a frequent complaint in merchant forums. Test for it.
How to set one up properly: 6 steps
- Read your last two to three months of messages. Tag the top ten question types. This tells you what the bot must handle, and whether you need answers or actions.
- Fix your source content first. Fill gaps in product descriptions, sizing charts, delivery and returns pages. The bot can only be as accurate as these.
- Connect order data if “where is my order” is a big share of your inbox. Without it, the bot can only point people to a tracking page.
- Write the rules. What it must never do (promise refunds, discuss payments), when it must hand over, and what it says when it does not know.
- Test with real past questions before launch. Run at least 50 genuine questions from your inbox through it and check every answer against your policies.
- Review transcripts weekly for the first month. Add missing answers to your pages or knowledge base, and tighten rules where it went wrong.
Our experience
We built an AI product description generator for an e-commerce client, which the client rated 5/5 on Fiverr. The lesson carries straight over to store chatbots: the AI was only as good as the product data behind it. Where a listing was missing materials or measurements, the output was vaguer. That is why we tell store owners to fix their catalogue and policy pages before switching any chatbot on; it is the cheapest accuracy improvement available. You can see more of our projects on our work page.
Off-the-shelf app or custom build?
Most small Shopify stores should start with Shopify Inbox or an established app, measure which questions it resolves, and only then decide whether a custom build is justified. Consider custom when:
- You need answers that combine data from more than one system, such as Shopify orders plus your 3PL’s tracking or your returns platform.
- You want the bot to take actions with your own approval rules.
- You sell in several languages or have complex products (made-to-order, bundles, compatibility rules) that a generic knowledge base handles poorly.
- You want to own the conversation data and the logic rather than rent them.
If your product knowledge lives in long documents, manuals or spec sheets, the build usually uses retrieval rather than training a model; our guide to RAG vs fine-tuning explains the difference.
FAQ
Can a Shopify chatbot check order status?
Yes, if it is connected to your Shopify order data. Shopify’s free Inbox app lists orders among its AI assistant’s data sources, and many third-party apps offer it too. By default, apps can read orders from the last 60 days; older orders need extra permission approved by Shopify.
Is a free chatbot good enough for a small Shopify store?
Often, yes. If most of your messages are order status, delivery and returns questions, start with a free or low-cost option and track how many conversations it resolves without a person. Move to a paid or custom tool only when you hit a clear limit, such as needing actions or several connected systems.
How do I stop the chatbot giving wrong answers?
Limit it to your own catalogue, policies and order data, tell it to say “I don’t know” and offer a person when the answer is not there, test it on at least 50 real past questions before launch, and review transcripts weekly. Never let it make promises your written policy does not already make.
Want help deciding?
If you are unsure whether your store needs an off-the-shelf app or something custom, book a free 30-minute scoping call. We will look at your question mix and tell you honestly which route fits. Our package options, including the AI Chatbot MVP, are on our packages page.