When the chatbot’s policy wasn’t the policy.

The case

In November 2022, Jake Moffatt asked Air Canada’s website chatbot about a bereavement fare. The bot said he could book now and apply for the discount afterward, within 90 days of the ticket date. He bought two one-way tickets and applied for the refund. Air Canada declined it.

The published policy on the same website said the discount did not apply after travel was complete. The chatbot had contradicted the page it linked to. Moffatt saved a screenshot. In February 2024, the British Columbia Civil Resolution Tribunal ordered Air Canada to pay him $812.02 Canadian in damages, interest, and fees. The decision does not establish whether the airline later paid.

The ruling was about this claim, in this tribunal. It is not a blanket rule for every chatbot answer. The practical point is closer to home: a fluent answer can be wrong even when it appears inside the company’s own help window.

Try the source check

1. Choose one AI answer you might rely on. Write down its exact claim.
2. Open the linked policy, original document, or official page. Find the sentence that governs the decision.
3. Mark where the answer and source agree, differ, or leave something out. If money, health, or a deadline is involved, confirm with the responsible person before acting.

The goal is a habit, not a guarantee: keep the original source beside the polished answer.