"The focus of a robo-poll is being very quick and short so, as a consequence, they may or may not ask enough questions to help them judge the likelihood of someone voting or enough demographic questions to weight their sample appropriately," said David Hill, a Texas-based pollster who has worked for Republican candidates since 1984.
Scott Rasmussen, founder of Rasmussen Reports, said automated polls mean that each respondent hears the same questions the same way, without any nuances from a live operator.
He said his company uses an online polling tool to reach people without conventional telephones.
While pollsters usually adjust their data to reflect the demographics of the population they're sampling - such as age, race and gender - Rasmussen also weighs his data to reflect political party enrollment or preference. He calls it "the best single indicator of how" someone will vote.
Other pollsters don't do that, because they say party identification can be fluid and change from survey to survey.
"It goes up and down; it changes," said Evans Witt, chief executive officer of the nonpartisan Princeton Survey Research Associates. "Party ID is not stable. It varies."
Voters excited about a candidate might tell a pollster they identify with that nominee's party, and then decide they lean independent in the next survey if their enthusiasm has waned, Witt said.
Another complicating factor in reading the most recent polls is the impact of a Labor Department report Oct. 5 showing the jobless rate fell to 7.8 percent, the lowest since Obama became president in January 2009.
"We have to get to the point where we have enough polls that were conducted past the debate and at least long enough past the jobless numbers," Witt said. "I don't think we've had a huge move. We're seeing a good debate for Romney, a good jobless number for Obama and I don't think that fundamentally changed the race."
A number of interest groups have added to the morass of presidential election polling. The Service Employees International Union sponsors polls with the Democratic-leaning Daily Kos website. Wenzel Strategies of Columbus, Ohio, partners in polls with groups including the Family Research Council, which supports a ban on abortion and calls homosexuality "unnatural."
"For whatever reason, if a poll is sponsored, it can influence the results," said Frank Newport, editor-in-chief of the Gallup poll. "Be very suspicious of polls sponsored by a political party or advocacy group. When they report to you, they're spinning."
Paul J. Lavrakas, president of American Association for Public Opinion Research, based in Deerfield, Illinois, said respondents may decide to participate in a sponsored poll depending on whether they agree with the aims of the group behind it.
"You're going to get biased results," Lavrakas said.