No need to be skeptical of it. But all companies use a reference for an "ancestral" population. The science is good, but the reality of DNA inheritance is messy. Consider that going back 4 generations you have 2x2x2x2=16 ancestors. But so does everyone else. ...this is obviously not sustainable, since there are 7,000,000,000 people on the planet and there cannot have 16x7,000,000,000 ancestors 4 generations ago because the earth's population was smaller in the past, not larger. More simply, what is happening is that people's ancestry lineages combine into common ancestors. This is relatedness. But DNA also changes (mutates) and recombines (large chunks of chromosomes switch places prior to reproduction), which messes up how we can reconstruct ancestry (mitochondria DNA and Y chromosome DNA doesn't do this recombining). If you're looking to find relatives only a few generations back, then just use archival data. If you're trying to find out if you and your sister are full-siblings, use DNA. And if you're trying to find some amount of distant ancestry, then these tests will provide a coarse guide to ethnicity (e.g., % DNA from Africa, East Asia, Middle East, N.America). This is because DNA companies don't have time machines. They have a database of DNA samples called a reference population. Reference populations are samples of DNA from tons of individuals throughout the world. You send them your DNA, they sequence it and compare it to the DNA sequences of their reference populations; if there is a strong match, then they will say something like 59% of your DNA matches with our reference population A...therefore you are 59% A, and 43% B, and so on.
But these reference populations are only "stand-ins" for ancestors, since they come from modern people--people who's own DNA has been changing across generations due to recombining with other people's DNA, mutating, and due to the people's ancestors migrating into new areas--just like your ancestors. So the reference population isn't a very secure estimate of ancestry. For example, say I submit a sample to 23andme and they tell me that I have 79% South Asian DNA ancestry. What this means is that my DNA matches at 79% to their reference population of people who's DNA they got from India. But it is very possible that the South Asian reference population has only recently moved into India in the last 500 years (but 23andme don't know that), from say Afghanistan and Iran, so really while the current reference population lives in India (and to whom I matched), the reference population's ancestors came from somewhere else.
More generally, people use ancestry subjectively, despite what the ancestry tests say. My dad's family is Dutch and Irish. My mom's family is from Hungary. I tend to identify (for arbitrary reasons) with being Hungarian more than being Irish. But it could have been otherwise--had my dad talked up his Irish pride a bit more during my childhood, I might have gotten a Celtic Cross tattoo as a teenager and then later deeply regretted it. Your pedigree could contain a bunch of average people, a thief and a princess. Most people with this pedigree would go around bragging about how they have royal blood. It's very subjective. Putting it more bluntly, look at the rise in "white nationalism" in right-wing fringe groups. There is no such thing as being a true white person in any genetic sense (since everyone is highly related, and Europe in particular has a history of different migrations) but these people, quite arbitrarily, subjectively believe in "whiteness."