I’d imagine some types of voters (depending on background, party affiliation, etc) are more likely to vote than others. If polls say candidates A and B have both 50% support, is it possible that supporters of A are more motivated than supporters of B and more likely to vote, meaning A actually has a significant lead? Or are poll results corrected for this effect and if so, how?
When polls are published, they are often accompanied by a line that says “of likely voters”. This is determined in a number of ways: often by merely asking if the respondent is likely to vote or by asking if they have voted in previous elections.
This British site has a good explanation of methodology of polling. A few minor details are different from those in the U.S. but the general principles are the same.