Sequence labels are strings of text following a greater-than (>) symbol at the start of a line in a FASTA file or the @ symbol at the start of a read in a FASTQ file. The label is terminated by the end of the line.
By default, the complete label is stored and reported, unlike BLAST which truncates labels at the first white space (blank or tab). You can use the -trunclabels option to specfiy that labels should be truncated.
USEARCH often uses or embeds annotations in sequence
labels, e.g. cluster sizes.