There was a story a while back which showed Trump doing better in online polls than in phone polls, which might have been a sort of Bradley Effect but just as easily could show different sampling biases between the two methods.
More relevant I think is that the polls were pretty predictive of Trump's actual primary performance (Cruz was the candidate who consistently outperformed his polling averages, until the Acela primaries where he flamed out). When the geek journos whiffed on Trump, it was because they dismissed the polls, not that they believed them.
Also, at least IME Trump supporters are pretty happy to tell you about it. Loudly. In spite of your efforts to aggressively direct the conversation to a different topic.