SEM or SD depends on what you want to express (the difference in your replicates or difference between repeat experiments; fold-change; relative expression; quantitative expression...)
A simple google search yields
https://www.researchgate.net/post/How_do_I_publish_qPCR_data_in_a_bar_graph to help you in your decision.
I would pay particular attention to Jochen Wilhelm's answers as he is well-known in the qRT-PCR field.
SEM or SD does not matter, as long as you specify in the figure legend. The statistical analysis you're performing doesn't care what your figure depicts.
People like to use SEMs because they're usually smaller (especially for large datasets), and it means that graphs can be more compact and informative (rather than wasting the top half of the graph with error bars etc), but if a difference is statistically significant, then it's significant regardless of how you choose to display your error bars.
A better question would be "what statistical tests are appropriate for qPCR data", and for that I would second the advice to listen to Jochen.
I hope you be fine, about your question as I had a problem like you I asked and searched all about that. In addition, as I read the comments, I strongly agreed with them, because it just depends on your data and no differences between them to use for bars. For me I used STD for qRT-PCR data, it showed better for my results. Therefore, it is your choice.