Why website chatbots fail, and how to avoid it
Last updated October 17, 2026 · Prices and features change, so check the linked sources before you buy.
Case 1: Air Canada’s bot invented a refund rule, and Air Canada paid
In November 2022, after his grandmother died, Jake Moffatt asked Air Canada’s website chatbot about bereavement fares. The bot told him he could apply for the reduced rate retroactively, within 90 days of buying the ticket. Air Canada’s own “Bereavement travel” page, which the bot linked to, said the opposite [1].
Air Canada argued it couldn’t be held liable for its chatbot. The BC Civil Resolution Tribunal called that “a remarkable submission”: a chatbot “is still just a part of Air Canada’s website… It makes no difference whether the information comes from a static page or a chatbot.” The tribunal also rejected the argument that the customer should have checked another page [1].
The February 14, 2024 decision ordered Air Canada to pay $812.02: $650.88 in damages, $36.14 in interest and $125 in tribunal fees [1][2]. The money was small. The reputational damage was not.
Case 2: NYC’s MyCity bot told businesses to break the law
In March 2024, The Markup found New York City’s Microsoft-powered MyCity chatbot giving business owners wrong legal answers. Among them: that landlords didn’t have to accept tenants on rental assistance, and that employers could take a cut of workers’ tips [3]. The city kept it online with added disclaimers [4].
Case 3: DPD’s bot swore at a customer
In January 2024, delivery firm DPD’s AI chat swore at a customer and wrote a poem criticising the company. DPD blamed an error after a system update and disabled that part of the bot [5].
The pattern regulators see
The U.S. Consumer Financial Protection Bureau’s 2023 review of bank chatbots flagged the same failures: wrong answers, “doom loops” that trap customers, and no timely path to a human [6].
8 rules that prevent this on a small-business site
- Answer only from your approved content. Pages and an FAQ you’ve reviewed, not the open internet.
- “I don’t know” must be a real option. Unknown question → collect name and contact → a human follows up.
- Never let the bot make policy. Refunds, discounts, exceptions, legal and medical answers go to a person.
- Keep one source of truth. When a policy changes, update the page and the bot together. Air Canada’s bot and its own page disagreed.
- Test before launch. Write 30–50 real customer questions, including awkward ones, and check every answer.
- Make the human exit obvious. A visible “talk to a person” option avoids the doom loop.
- Say it’s a bot. Visitors forgive a bot that admits it doesn’t know far more readily than one pretending to be human.
- Read the logs weekly. Unanswered questions are your next FAQ entries.
Health practices should add a ninth rule: keep patient health information out of a general website bot unless the vendor will sign the right agreements. Our free AI front-desk checklist for dental practices covers this.
This is the standard Botday builds to: answers come only from your content, unknowns become leads instead of guesses, and every bot is QA’d before it goes live. For tool options by industry, see Adoptly’s guides.
FAQ
What happened in the Air Canada chatbot case?
In Moffatt v. Air Canada, 2024 BCCRT 149, Air Canada’s website chatbot told a grieving customer he could claim a bereavement fare retroactively, which contradicted Air Canada’s own policy page. The BC Civil Resolution Tribunal found Air Canada responsible for its chatbot’s information and ordered it to pay $812.02 in damages, interest and fees.
Is a business legally responsible for what its chatbot says?
In the Air Canada case, the tribunal said a chatbot is part of the company’s website and the company is responsible for all information on it, whether it comes from a static page or a chatbot. Treat your bot’s answers as your own statements. This is general information, not legal advice.
How do I stop my chatbot from making things up?
Limit it to answering from content you have reviewed, give it an explicit “I don’t know, let me get a person” path, block it from making policy decisions such as refunds, test it with real questions before launch, and review the conversation logs weekly.
Should a website chatbot say it is a bot?
Yes. Being upfront sets expectations, and a visible route to a human avoids the “doom loop” problem regulators such as the CFPB have flagged.
Sources
- Moffatt v. Air Canada, 2024 BCCRT 149 (full decision, CanLII)
- CBC News: Air Canada found liable for chatbot’s bad advice on plane tickets
- The Markup: NYC’s AI Chatbot Tells Businesses to Break the Law (Mar 29, 2024)
- AP News: NYC’s AI chatbot was caught telling businesses to break the law. The city isn’t taking it down
- BBC: DPD error caused chatbot to swear at customer
- CFPB: Chatbots in consumer finance (Issue Spotlight, June 2023)