The Gallup religion poll is self-reported.[1] But some of the questions asked are objective, such as "How often do you attend church or synagogue -- every week, almost every week, about once a month, seldom or never?" Self-reported church attendance in the US is about twice what churches report actually show up. (That info used to be on adherents.org, but the site is gone.)
Questions along those lines are tough to get answered well. Who asks makes a big difference. Phone surveys on drinking get quite different results if the people calling are drinkers or non-drinkers.
I got asked to participate in a Gallup poll once. On numerous questions, my answer wasn't part of the "permitted" answers so they had to enter something only mildly related to my real opinion (and the interviewer kept badgering me to select one of the permitted answers, which was amusing).
These polls are unreliable from the outset, bad poll design defeats their accuracy even with well-meaning respondents. So nowadays, I just decline to participate.
Questions along those lines are tough to get answered well. Who asks makes a big difference. Phone surveys on drinking get quite different results if the people calling are drinkers or non-drinkers.
[1] https://news.gallup.com/poll/1690/religion.aspx