AI gets things wrong more often than you think

When was your last conversation with Chat, Claude or Gemini? If you’re like 66% of Canadians, you’ve probably experimented with generative AI tools. Thirty per cent of Canadians use them daily or weekly for leisure, work or study.

AI is helping people plan recipes for birthday parties, fix broken faucets and find great restaurants downtown. But many people are also turning to it for more complicated matters like health advice, medical questions and travel planning.

While AI can be helpful and fast, you can’t always take what it says at face value. There are more and more stories of inaccurate health, medical and travel advice. If you and your kids are increasingly relying on AI, it’s important to question, challenge and fact-check the answers you’re getting, so the information is helpful rather than hazardous.

AI use in numbers

How are Canadians using AI? According to research from polling firm Leger, among Canadians who use AI, chatbots and assistants like ChatGPT are the most common tools, used by 73%. Sixty-four per cent of Canadians trust AI tools for simple household tasks.

But AI use is extending beyond cooking, DIY projects and gardening. In its Health and Media Tracking Survey, The Canadian Medical Association (CMA) surveyed 5,000 Canadians and found that 89% of Canadians go online for health information. Approximately half is using AI to diagnose or treat health issues. Twenty-seven per cent trust AI to provide accurate health information, and 38% have used ChatGPT for treatment advice.

Youth and teens are also turning to AI for emotional support or advice. The University of Ottawa surveyed 40,000 Ontario students in grades four to 12. Researchers found that one in five used AI as an emotional sounding board. Those that used AI were two times as likely to have clinical emotional problems as those who did not.

AI has also become a common travel planning tool. According to the Blue Cross Travel Study, which surveyed 2,000 Canadians, 49% of travellers use AI when planning trips. They compare prices across destinations and dates, build itineraries and discover destination ideas and inspiration.

Bad advice, real health consequences

In a December 2025 story, a 60-year-old man in the U.S. ended up in the hospital after some bad health advice from ChatGPT. He wanted to eliminate salt from his diet and asked Chat for an alternative. Chat recommended sodium bromide, a chemical sometimes used in pesticides.

He bought the bromide online and used it for three months. After regular use, he started to have paranoia and hallucinations. The doctors discovered that he was suffering from bromide toxicity.

The CMA found that survey respondents who followed advice from AI were five times more likely to say they experienced an adverse reaction or negative effect on their health as a result.

Stranded by AI

Imagine planning your dream family vacation or romantic anniversary trip only to find out that the itinerary you built with AI includes sites that don’t actually exist? It’s happening more than you think.

The BBC reported on a story about a couple who traveled to Peru and included an epic trek to the Sacred Canyon of Humantay in their itinerary. A chance encounter with a local guide steered them away from definite disappointment and potential danger.

The Canyon in question doesn’t exist. And the couple paid $160 USD each to get to the starting point of the trek, without a guide or verified destination.

Dana Yao and her partner had a similar experience. They planned to hike to the summit of Mount Misen in Japan for sunset. AI told them that the last ropeway down was at 5:30pm. But once they got up to the station to make their descent, they learned AI was wrong, and it was already closed.

Before you trust, verify

When turning to AI chatbots for everyday or more complicated questions, it’s important to know how they generate answers.

Carnegie Melon machine learning professor Rayid Ghani points out that AI only knows words. It’s amazing that Chat or Claude can seem to know about everything and anything within minutes. While it seems like real, logical advice, the large language models powering AI chatbots analyze huge collections of text and put together words and phrases that statistically feel like the right responses. Sometimes the responses are bang on. But sometimes they are what’s known as “hallucinations.” Whether it’s real or fake information, AI presents it the same way.

How can you get the benefits of AI without the risk? If you’re asking for important information with any kind of consequence, verify and fact-check. Don’t take everything AI gives you at face value. And keep having conversations. Share with your kids how you’re using AI and what you’re doing to make sure the information is accurate. Understand how your kids are using AI, what kinds of conversations they’re having and how they’re taking in the information. Even though it’s everywhere, AI is still relatively new in the mainstream. We’re all figuring it out, so staying curious and asking questions are the best ways to make the most of what it has to offer.