An easy way can be to download all UTR regions with Biomart and the use a simple RegEx with grep/perl/python to identify and count your motif.

Here's an example using the R interface to Biomart (just looking at 100 human genes and only the 5' UTR, to keep things simple):

library(biomaRt)
ensembl <- useMart("ensembl", dataset="hsapiens_gene_ensembl")
IDS <- getBM(attributes="ensembl_gene_id", mart=ensembl)[c(1:100),]
seqs5 <- getSequence(id=IDS, type="ensembl_gene_id", seqType="5utr", mart=ensembl)
hits <- seqs5[[2]][grep("TAGGTTAAG", seqs5[[1]])]

If you already have the genomic sequence sitting around and an annotation file for it, then you can just import the GTR into R, filter it so you only have UTR sequence, and then getSeq() on those followed by a similar grep(). This would likely be faster.