See also
Quality control for OTU sequences
Sometimes primers amplify the wrong target, e.g. human DNA instead of 16S.
One way to identify this type of OTU is use ublast to search a large nucleotide database (e.g. NCBI nt), or submit the OTUs to the NCBI BLAST web site. Take the top hit, or top few hits, for each OTU, delete those containing expected strings for your gene such as "16S" or "SSU" and examine the rest more closely.
Another method is to use the sintax command to predict taxonomy and examine those with low boostrap confidence at domain or phylum level. This should give a smaller set which you can examine further, e.g. by submitting to NCBI BLAST.
A third option is to use the usearch_global command to align the OTUs to a large sequence database such as SILVA. Examine any which fail to align or have low identities (say, <80%).