References in a table are mistakenly detected as cited papers

I recently wrote a paper with tables, that contained columns that look like this:

cite{paper1} & explaining a method from that paper 
hline
cite{paper2} & explaining a method from that paper 
hline

I recently uploaded that paper to arXiv, and saw that both SemanticScholar and Prophy detected ~90 Bibliography items in it, while it really only contained 39.

They both listed the descriptions from the table’s second column as reference titles, thus the number of detected references got that high.

Is there any bad practice that I should have avoided with those tables?