the CODIS database uses
STRs ("short tandem repeats") at 13 (pre-2017) or 20 (post-2017) sites in the genome. the GEDmatch database uses an autosomal profile that consists of anything up to a million
SNPs (single nucleotide polymorphisms). A GEDmatch autosomal profile will return matches by degree of relatedness expressed in centimorgans (centimorgans are a unit of genetic distance); there are typical ranges of shared centimorgans for genealogical relationships (parent/child share around 3400cM; first cousins share around 700-1200cM; third cousins on average share 53cM). By comparing an autosomal profile on GEDmatch (or a similar service like FTDNA which also allows autosomal uploads and use by law enforcement) you'll get a list of people who share DNA with that profile ordered by relatedness. The GEDmatch system has enough uploaded DNA profiles, at present, that it's estimated that a majority of Americans of European descent likely can be identified through the shared genetic material of relatives as distant as fifth cousins; the method involved requires finding a family tree for, ideally, multiple matching profiles and looking for common ancestry and tracing forward in time from those common ancestors to descendants who may be a potential UID or suspect (this was the method used to identify the Golden State Killer).