what specifically is the fake part?
The over all concept?
or the genome statistics?
The concept: yes and no. There are indeed possible genetic insights into diseases, but right now there aren't very many likely, actionable insights. Maybe that will change in the future.
Ancestor statistics: quite dubious. They're have been stories circulating (maybe just rumors?) that they'll toss in unexpected results. Not going entirely down the whole "race is a social construct", but there is some truth to that: I doubt we'll ever be able to determine Czechs from Slovaks, for example. Plus none of that stuff really matters anyway. Define yourself by whatever culture you want, nobody cares.
Ethics and privacy: horrible. This isn't science research governed by strict ethics and controls of information but rather a for-profit company that will profit off your genetic information. Once you give them your information your data is forever compromised. It's not protected like your other medical records, and they'll hand it over to law enforcement. What will they do in 20 years? Remember when Google, Facebook, etc changed policies and lied about what they do with your data? Once you send it over today, there is no clawing it back.
They've got great marketing and big VC money behind them. That's a good indication that you're better staying far away if you have seen how these things have played out in the past.
there are a number of sites where people can contribute their genetics for purposes of exploring family relations ships and more.
This can get pretty deep
by way of example
http://familypedia.wikia.com/wiki/Haplogroup_R1b
https://www.eupedia.com/europe/Haplogroup_R1b_Y-DNA.shtml
https://www.eupedia.com/europe/european_y-dna_haplogroups.shtml
https://upload.wikimedia.org/wikipedia/commons/c/cb/Aryan_migration_map_and_haplogroup_tree.jpg
https://en.wikipedia.org/wiki/Haplogroup_R1a
so I think the knowledge base is expanding faster than may people suspect.
The utter complexity of the diversity of human haplogroups and their geographic distribution also makes a mess of traditional views of races.
Their data's accuracy is limited by the scope of their database. For example when running chimpanzee dna against a human database you find out their closest relatives are nigerian and/or black. From a traditional evolutionary standpoint this makes sense but any science can be perverted as truth and this scenario certainly just seems like fuel for racists.
In a case like Elizabeth Warren her DNA wasn't even compared to a sample of actual Native Americans making any results more or less useless and largely speculative.
When DNA is most effective is in direct comparison. These sites like 23 and me generally use different criteria (number of markers they use) and compare them to different databases of information. As far as rough estimates go it's not terrible but to claim it as definitive is a stretch at best.
do you even know how a genetic database works?
23andme is fake science that nobody should support.