Voters in the US go to the polls on 5 November to elect their next president.
The election was initially a rematch of 2020 but it was upended in July when President Joe Biden ended his campaign and endorsed Vice-President Kamala Harris.
The big question now is - will the result mean a second Donald Trump term or America’s first woman president?
As election day approaches, we'll be keeping track of the polls and seeing what effect the campaign has on the race for the White House.
What do the polls say about who won the debate?
Just over 67 million people tuned in to watch Harris and Trump go head to head in the debate in Pennsylvania on 10 September. But what do the polls tell us about who won?
A Reuters/Ipsos poll of 1,400 registered voters who had heard at least something about the debate found that 53% said Harris won and 24% said Trump won. It also suggested Harris had a lead of five points over her rival nationally, 47% to 42% - up from 45% to 41% in August.
A YouGov poll of 1,400 adults in the US had similar conclusions - of those who had watched the debate, 55% said Harris won and 25% said Trump. Even so, it found no change in voting intentions, with Harris having the same lead of 46% to 45% as before the debate.
There was also no bump for Harris in a Morning Consult poll of 3,300 likely voters that put her lead at 50% to 45% - although Trump was down one point from 46% in their poll before the debate.
The data we have at the moment suggests that although a majority of those watching the debate felt Harris came out on top, her performance might not necessarily translate to more votes because so many Americans have already made their minds up on who they are supporting.
Who is leading national polls?
In the months leading up to Biden’s decision to drop out of the race, polls consistently showed him trailing former president Trump. Although hypothetical at the time, several polls suggested Harris wouldn’t fare much better.
But the race tightened after she hit the campaign trail and she developed a small lead over her rival in an average of national polls that she has maintained since. The latest national polling averages for the two candidates are shown below, rounded to the nearest whole number.
In the poll tracker chart below, the trend lines show how those averages have changed since Harris entered the race and the dots show the spread of the individual poll results.
While these national polls are a useful guide as to how popular a candidate is across the country as a whole, they're not necessarily an accurate way to predict the result of the election.
That's because the US uses an electoral college system, in which each state is given a number of votes roughly in line with the size of its population. A total of 538 electoral college votes are up for grabs, so a candidate needs to hit 270 to win.
There are 50 states in the US but because most of them nearly always vote for the same party, in reality there are just a handful where both candidates stand a chance of winning. These are the places where the election will be won and lost and are known as battleground states.
How are these averages created?
The figures we have used in the graphics above are averages created by polling analysis website 538, which is part of American news network ABC News. To create them, 538 collect the data from individual polls carried out both nationally and in battleground states by lots of polling companies.
As part of their quality control, 538 only include polls from companies that meet certain criteria, like being transparent about how many people they polled, when the poll was carried out and how the poll was conducted (telephone calls, text message, online, etc).
Can we trust the polls?
At the moment, the polls suggest that Kamala Harris and Donald Trump are within a couple of percentage points of each other both nationally and in battleground states - and when the race is that close, it’s very hard to predict winners.
Polls underestimated support for Trump in both 2016 and 2020. Polling companies will be trying to fix that problem in a number of ways, including how to make their results reflect the make-up of the voting population.
Those adjustments are difficult to get right and pollsters still have to make educated guesses about other factors like who will actually turn up to vote on 5 November.