It's notoriously poor at math. Don't use it for math. Google went through hoops to help, see the 2023-06-07 update here: https://bard.google.com/updates. So if it doesn't understand it's a math topic, it uses poor defaults from its standard, limited LLM reasoning abilities.
In fact, don't use it for facts either because it's very bad at this too, basically giving me a long of self-assured falsehoods about city facts for my home city. It really turned me off. How can we use these things in any setting where their replies actually matter and not just for fun? The worst problem with LLM's isn't even that they're often wrong, but that they're so sure of themselves and try to be convincing, and don't know when to stop. That's downright harmful to humanity.