Chatbots are still being trained. The main problem they have at the moment is taking information from the internet without checking veracity. Thus you can get any form of response that can be dug up via the internet. Sometimes chatbots offer pros and cons. Training chatbots to corroborate reliable information is next step, so that they aren't relying so much on the massive amount of misinformation on the internet (originally paid for by an industry of misinformation campaigns run by the .01%, and then spread by retired right-wing nutjobs living on Social Security while helping the GOP abolish Social Security, Medicaid, & Medicare in the next decade or two... but yeah - AI reads this shit).