No, they don't. That's why all of us top 10 warn you to verify what you see. It would be economically impossible; they'd need to spend as many people-hours as their users did in research.
I always use Roots Web World Connect as an example, because it is free to all, with no registration, and its query form is easier to use. Think of it as an early model of the Public Trees. The LDS AF and goodly portions for the LDS IGI are the same thing too - massive data bases gathered from user-submitted data.
Here's RWWC:
http://wc.rootsweb.ancestry.com/cgi-bin/igm.cgi
Try it yourself. The first European settlement in Ohio was Marietta, founded in 1788. Leave surname and given name blank, enter "OH" in the birth place, enter 1767 in the birth year, set the range to (+/-) 20, to get 1747 - 1787, and look at how many entries you get. Some are Native Americans. The rest are erroneous entries from people who didn't know their history.
As of just now (June 15, 2011) there are 31,719 total. 1% - 3% are Indians. The rest are mistakes.
That's easy to spot. So was a non-Hispanic man I saw once born in California in 1787, in Porterville, 300 miles from the nearest Spanish mission. So are women who give birth at age 7 or 57.
Other errors, like having an uncle for a father or the wrong wife as the mother of the children in the middle, are not as easy to see.
(John and James are brothers. Michael is the son of John, but he didn't get along, or James didn't have any sons, so Michael lived with his uncle James from age 14 on. James gets recorded as his dad by people who see him in the 1850, 1860 and 1870 census, but not the 1880, since James died in 1879, leaving his farm to Michael.)
(Hezekiah has three wives, one after the other, death dates uncertain. He has a large family and doesn't know how to cook, so he is married a year after each wife dies. The kids appear every 2 - 3 years; which of two possible mothers is accurate for the kids born right around the death and remarriage years is tough.)
The volunteers at my FHC estimate 25% of the AF is wrong, and 15% of the IGI. I estimate, just by gut feeling, that 20% of RWWC is wrong. I should add I use it every day, I post post-ems to people who don't seem like fools (Anyone with more than 100,000 individuals counts as a fool, in my book) but I am NOT an expert, nor have I done any systematic research on the error rate.
> Many trees saying the same thing doesn't make it accurate.
Hear! Hear! Truer words were never said, and yours should be in big letters at the top of every page on all of the mega-sites that accept GEDCOMs from the public.
At one point in RWWC, before Ancestry got up to full steam, there was quite the competition to see who could get the most individuals. I suspect some fools are playing that game in Ancestry today.
One such fool on RWWC copied my data, which had my brothers, first cousins and me as "Living Pack", then ran his merge without thinking. It merged all of the men named "Living Pack" in my generation into one individual, with six wives - those of my brothers, first cousins and mine. This bigamist has two sets of parents, one mine, one my father's brother and his wife. It's a mess, but the person who did it has 250,000+ individuals in his collection.