The whole point of using the data like this is to change people's opinion without them knowing why, so I doubt anyone can answer a survey like this accurately.

Perhaps if there were two similar candidates, this would be true. However, that wasn't the case here. These candidates and their supporters were polar opposites. If they were swayed at all, it wouldn't have been through subtlety. The stories would be "I was going to vote for Hillary, but then I saw [X] on Facebook and was so horrified that I decided to [not vote at all or vote for Trump]".

I'm pretty sure that polls like that are ineffective for discerning the impact of any type of marketing. The best evidence I can think of for whether somebody found something influential is whether they liked or re-shared a post, and there's plenty of evidence for that. Those are, after all, the sorts of metrics typically used for measuring the success of a social media campaign: https://www.socialmediaexaminer.com/10-metrics-to-track-for-...