AI Can Be Wrong: How to Use It Without Getting Burned

One of the things I love most about AI is how confidently it can help you solve a problem.

Ask it a question and, within seconds, you can get a polished explanation, a detailed plan, a professionally written email, or an answer to something you may have spent 20 minutes searching for online.

There is just one small problem.

Sometimes AI can be wrong.

A few weeks ago, I was working on a brokerage onboarding workbook when, out of the blue, ChatGPT randomly referred to me as “Jeff Sabo, Broker of Record, Century 21 Local Home Team.” 

A little strange, but I thought it might just be a placeholder name. Then it happened a second time.

Now I was a little concerned, so I said to ChatGPT:
“My name is Len Filiault, not Jeff Sabo.”

“Thank you for catching that, Len. That’s completely my mistake. I slipped into using the placeholder name from earlier brainstorming instead of your name. I’ll use Len Filiault going forward and make sure any personal messages, letters, or Coach’s Insights reflect your voice, not anyone else’s.”

But where did “Jeff Sabo” come from, and who is he? That was obviously my next question. 

“He’s a real estate broker and coach that I inadvertently pulled into our conversation from my own pattern matching. He has nothing to do with your project, and he should never have appeared in the manual.

That was my mistake.

From the beginning of this project, I’ve been helping you, Len Filiault, build the Century 21 Local Home Team Realty onboarding and career development system. At some point, while drafting example leadership messages, I mistakenly substituted another real estate professional’s name. It wasn’t based on anything in your document or anything you told me.

I’m glad you caught it.”

And here’s the really interesting part: I had no way of knowing whether that explanation was true either. ChatGPT may have hallucinated the name, then hallucinated an explanation for why it hallucinated the name. This was a relatively simple mistake that could easily be found and corrected before anything was used or printed. But other mistakes can be completely wrong and go unnoticed. 

AI can give you an answer that sounds incredibly reasonable, detailed and authoritative… and still be completely incorrect.

That is one of the most important things to understand if you’re going to use AI regularly in your business.

 

---

 

AI Doesn’t Actually “Know” Everything

Tools like ChatGPT are incredibly good at understanding language, recognizing patterns, organizing information and helping us work through ideas.

But they are not infallible databases of facts.

Sometimes AI may:

• misunderstand your question
• rely on outdated information
• confuse similar rules or regulations
• fill in missing information with an assumption
• misinterpret a document
• cite a rule that doesn’t apply to your situation
• confidently provide information that simply isn’t true

This is sometimes referred to as an AI hallucination.

The strange part is that the incorrect answer often doesn’t look incorrect.

It may sound just as confident as the right answer.

And that is where people can get into trouble.

 

---

 

Real Estate Is Especially Sensitive

Imagine asking AI:
“Which OREA form do I need in this situation?”

Or:
“Does this situation create multiple representation?”

Or:
“What does FINTRAC require me to do here?”

It might identify the questions you should be asking. It may even point you in the right direction. But those are not areas where I would blindly copy the answer and move on.

Real estate deals involve contracts, legislation, compliance requirements, taxes, deadlines and significant amounts of money. Those answers deserve verification.

 

---

 

Think of AI as a Very Smart Assistant

If you had a talented new assistant working for you, you might ask them to research something.

But if the answer affected a contract, your client’s finances or your legal obligations, you probably wouldn’t say: “Sounds good. Let’s go with that.” You’d verify it.

AI should be treated the same way. Use it to help you:

• understand an issue
• summarize complicated information
• identify possible solutions
• prepare questions
• compare options
• find areas that need further research

Then verify the important facts.

 

---

 

A Simple Rule: Ask. Verify. Apply.

When I’m using AI for something important, I like to think of the process in three steps.

1. ASK
– Use AI to help you understand the question.
– Give it context.
– Explain the situation.
– Ask it to identify the issues you should consider.

2. VERIFY
– Before relying on the answer, check the original source.
– Depending on the question, that might mean:

• RECO
• OREA
• FINTRAC
• CREA
• CRA
• Government of Ontario
• your brokerage policies
• your lawyer
• your accountant
• your Broker of Record

The more important the decision, the more important the verification. 

3. APPLY
Once you know the information is accurate, use AI to help you turn that knowledge into action. That might include creating a client explanation, checklist, email, meeting agenda or follow-up plan. That is where AI becomes incredibly useful.

 

---

 

Here are three tips to help you avoid becoming a real estate meme. 

#1) Ask AI to Challenge Its Own Answer

One of the easiest ways to get better results is to stop treating the first answer as the final answer.

Try asking:
“How confident are you in this answer?”
“What assumptions are you making?”
“What parts of this answer should I independently verify?”
“Could there be another interpretation?”

Those simple follow-up questions can dramatically improve the quality of the conversation.

#2) Ask for the Original Source

This is especially important when dealing with regulations, government programs or industry rules.

Instead of asking:
“What are the FINTRAC requirements?”

Try:
“Using current official FINTRAC sources, explain the requirement and provide the original source I should verify before relying on your answer.”

Or:
“Explain this Ontario real estate rule, but do not guess. If you cannot verify the current rule from an official source, tell me that.”

That last part is important. Give AI permission to say “I don’t know.”

Sometimes the most useful answer AI can give you is: “I can’t verify that reliably.”

#3) Be Careful With Dates

Another common problem is information that was once correct but has changed.

Real estate regulations change. Forms change. Government programs change. Interest rates change. Policies change. Technology changes…

Whenever your question involves something current, include the date or ask AI to verify the most recent information.

Instead of:
“What is the current rule?”

Try:
“As of today, what is the current Ontario rule? Please verify it using an official source and tell me the date of the source.”

 

---

 

This Month’s Prompt

The next time you are researching an important question, try adding this to your prompt:

“Act as a research assistant, not the final authority. Before answering, identify any assumptions you are making. If the information could have changed recently, verify it using current reputable sources. Prioritize original or official sources whenever possible. Clearly separate verified facts from your interpretation. Tell me what parts of your answer I should independently confirm before relying on it. If you cannot verify something, say so rather than guessing.”

That one addition can make your AI conversations considerably more useful.

 

---

 

AI can save us enormous amounts of time. But saving five minutes isn’t valuable if we spend five hours fixing a mistake.

The best AI users aren’t the people who blindly trust everything it tells them. They’re the people who know when to trust it, when to question it and when to verify it.

Use AI to help you think.

Use reliable sources to confirm the facts.

And use your own professional judgment before you act.

That’s how you get the benefits of AI without getting burned.