Just a note that, in the article Lynda linked, I would take the AncestryDNA claims of distant cousin detection with a sizable grain of salt, maybe two grains. Their phasing algorithm--which is genotype phasing, meaning that it is based on estimates of population models, not some massive database containing actual trio-phased data--is proprietary and has never been revealed. Meaning we can only take Ancestry's word for it.
The 23andMe numbers, on the other hand, were derived from an academic, peer-reviewed study led by a then-lead scientist for 23andMe, Brenna Henn. You can scroll down that page a bit to the section titled "How many cousins do we have?" to find one of the summary charts and a link to the published study at PLOS.
I personally consider Ancestry's success numbers at 4th cousins and beyond to be highly inflated and suspect. We have no way to view or substantiate their data, and they don't have to give them to us. That the 23andMe and FTDNA numbers are reasonably in sync while Ancestry's are off the chart should tell us something.