Do you remember 2016? Polls were saying Clinton would beat Trump by a significant margin.
If you're approaching this logically, you'd notice the trend on data being unreliable when Trump is on the ballot.
It's mostly attributed to inaccuracies in putting appropriate weight on likely voters vs. unlikely voters. People considered unlikely to vote by pollsters went out and voted, and they voted for Trump.
Measuring racism is also something that polling is bad at. People simply don't like to admit to being racist. Is this related to the reason why polling on Trump is inaccurate? We don't know because there's no data on this. Some things polling just fails at. Can't do much when people won't provide you with data that may be relevant.
We do know that Trump's primary numbers were lower than polling indicated it would be. Does that mean his numbers in the general will be lower than the polls we're seeing right now? We don't know.
What effect did January 6 have on people's decisions? Some people may not want to talk about it. But the week before election they'll probably be seeing political ads showing video about Jan. 6 and ask people straight up "do you want this to happen again?" which might people who might say Jan. 6 wasn't a big deal to privately think otherwise just stay home on Election Day. Polling is based on past trends, so isn't going to be good a predicting anything after unprecedented events.
After this election pollsters have a baseline for how likely people will vote for a candidate lost the previous election, tried to overthrow the government, was convicted of felonies, had an assassination attempt vs. a candidate that suddenly became prominent after the sitting President and presumptive nominee dropped out the race 3.5 months before the election. But right now there's not a lot of data there on this particular scenario.
The data is simply too unreliable to make any prediction on anything. So... vote!
