I find it is often bad at historical data that you ask for. It will point occasionally give (and point sources) at contemporary wages when I ask for wage data from decades ago. It more often blurs together sources from too wide a span when things changed a lot in a short time (particularly giving post-WW1 info when asked about early 1910 or pre-WW1 when asking about late 1910s).
It occasionally get things completely spatially wrong. I asked about Italian village stonemasons in the 1890s and get pointed to a page about English churches.
Asking it to do a long-term narrative/prediction can have some funny results. I picked a scenario for an Italilan blacksmith in a hamlet of Bisenti who married an “old maid” in 1919 (she missed her marriage years because of the war) and got a $300 dowry out of it. A lot of money given the exchange rate. So he bought some land - makes sense (at least in a few years). It acknowledged the war for grain, at least. But when it got to WW2, he was escaping the inflation of the Lira by trading his services for goods, eating the grain and olive oil from his own crops, etc. and quite prosperous. Except he was in an area occupied by German forces, so not very likely, I think we can say.