An Airline's Chatbot Made Up a Refund Policy. The Airline Had to Pay.
The Air Canada chatbot ruling was a small refund dispute with a large lesson: when a support bot gives a customer the wrong policy, the business may still own the answer.
Case
Moffatt v. Air Canada
British Columbia Civil Resolution Tribunal, 2024 BCCRT 149.
Failure
Wrong refund guidance
The chatbot said a bereavement fare could be requested after travel. Air Canada's policy said otherwise.
Order
CAD 812.02
Damages, interest, and CRT fees. The amount was small; the precedent travelled.
In 2024, a grieving customer asked Air Canada's website chatbot about bereavement fares. The bot told him he could apply for the discount retroactively within 90 days of booking. He booked, flew, and submitted the claim. Air Canada refused because the real policy said the opposite.
He took the airline to the British Columbia Civil Resolution Tribunal. The tribunal found Air Canada liable for negligent misrepresentation and ordered it to pay. The damages were modest. The lesson for Shopify stores is not.
The short version
If a customer relies on a support bot's answer, a business may not be able to shrug and say the bot was wrong. For a Shopify store, that means return windows, warranty promises, shipping timelines, discounts, and sizing advice need to come from approved store content, not AI improvisation.
What actually happened?
The decision is Moffatt v. Air Canada, 2024 BCCRT 149. Jake Moffatt used Air Canada's chatbot while booking bereavement travel. The chatbot described a process for requesting the reduced fare after travel had already happened. The linked policy page said bereavement consideration did not apply after completed travel.
Air Canada argued that the customer should have checked the policy page and that the airline should not be liable for what the chatbot said. The tribunal rejected that framing. It treated the chatbot as part of Air Canada's website and found the customer's reliance reasonable.
You can read the source decision on CanLII . The key operational point is simple: if one part of your website contradicts another, the customer does not carry the burden of guessing which one is trustworthy.
Why a small award matters to a store doing $40k a month
It is tempting for a small Shopify founder to treat this as a big-airline problem. It is not. The reasoning does not depend on the size of the business. A solo store may be more exposed because policies are often scattered, edited quickly, or partly trapped in the founder's head.
The risky chatbot answer in ecommerce is usually not dramatic. It is a confident but false sentence like "you have 30 days to return this", "express shipping will arrive by Friday", "clearance items are refundable", or "this size runs small". Each one can create refunds, chargebacks, angry email threads, and trust loss.
This is not legal advice. For Australian stores, misleading statements about refunds, warranties, delivery, or consumer guarantees can already be sensitive. The practical lesson is narrower: do not put a chatbot in front of customers unless you can control what policy facts it is allowed to use.
The lesson is not "avoid AI". It is "ground it".
Repetitive support questions are real. WISMO, returns, sizing, warranty, and product-fit questions drain a small team. AI support can help, but only when the bot is constrained by your approved content.
Generative free-for-all
The bot sees a policy question, knows what policies usually sound like, and writes a plausible answer. If the real answer is missing, stale, or ambiguous, it may invent one.
Grounded support bot
The bot answers only from approved store content: your real policy pages, product details, FAQ, order-status rules, and uploaded documents. Missing answer? It hands off.
That constraint is not a weakness. It is the safety feature. A grounded bot cannot invent a 90-day retroactive refund policy if that policy does not exist in the approved content it can retrieve from.
What a Shopify store should do this week
- List your top 10 support questions. Include returns, shipping times, sizing, warranty, order changes, discounts, and international delivery. If the answer is not written down, your bot cannot safely answer it. Start with the Shopify AI chatbot knowledge-base checklist.
- Audit your current bot. Ask it those 10 questions and compare every answer against the real policy. Flag anything vague, overconfident, or inconsistent.
- Make "I do not know" an allowed answer. A bot that always answers will eventually answer something it should have handed to a person.
- Require source grounding. Your support AI should be able to show where an answer came from. Use the AI customer support transparency checklist to review disclosure, citations, and handoff.
- Keep one source of truth. If your footer policy, FAQ page, product page, and chatbot memory disagree, the bot may surface the wrong one. Fix the content before increasing automation.
Where DocMind fits
DocMind is built around this principle: answers come from approved store content, unknowns route to a human, and replies can be traced back to source material. It is deliberately less flashy than a bot that improvises every answer. That restraint is the point.
Ground before you automate
Turn approved store content into safer AI answers.
Build a support bot around real policies, FAQs, and product content so it can answer common questions without inventing promises.
FAQ
Can a small business be responsible for what its AI chatbot says?
The Moffatt v. Air Canada decision found Air Canada responsible for misleading information provided by its website chatbot. It was a Canadian tribunal decision, not universal legal advice, but the practical lesson is clear: customers can rely on what your business publishes through a bot.
Does this mean Shopify stores should avoid AI chatbots?
No. It means a store should use an AI support bot that answers from approved policies, product pages, FAQs, and help documents instead of improvising. The risk is not automation itself; the risk is a chatbot that makes up policy answers with confidence.
What is a grounded chatbot?
A grounded chatbot retrieves answers from approved store content such as return policies, shipping pages, product information, sizing guides, and help-center articles. When the answer is missing, it escalates or says it cannot answer instead of inventing a policy.
How do I stop a Shopify chatbot from inventing return or shipping policies?
Write your policies clearly, keep one source of truth, train the chatbot only on approved content, audit answers against real policy pages, and require human handoff for refund exceptions, chargebacks, damaged goods, fraud, or anything that requires judgment.
Why does the Air Canada chatbot ruling matter if the award was small?
The dollar amount was modest, but the operational lesson is large: a wrong chatbot answer can become a promise the customer acts on. For ecommerce stores, the more common cost is refunds, chargebacks, support cleanup, and trust lost after a confident but false answer.
Source note
Case details were checked against Moffatt v. Air Canada, 2024 BCCRT 149 on CanLII. This article is an operational risk guide for store owners, not legal advice.
About the author
Daniel Tang is the founder of DocMind. He studied Statistics at the University of Toronto, with coursework in machine learning and large language models, and previously worked at ByteDance as an AI Product Manager.