As I suggested on slack, just have the bot post a small, neutral message stating that it found similar content at another site, and give a link. Humans can then review the material and decide whether to make a stronger accusation.

This will not only defuse negative reactions from people who don't want to be accused by a bot (either falsely or as a matter of a gray area) but it also reduces the spammishness of the bot by making the bot-posted comment smaller.