So
PolitiFact "fact checked" this poll. They gave it a "half true". I give their fact check a "half true".
They gave two main arguments: 1.) they questioned the poll's accuracy since it was an online poll and 2.) they questioned the use of the word "fabrication". I agree with both points but the poll's results can only be 100% true.
Explanations:
They say that since it was an online poll, it required the respondents to take an action, which makes it less than random - and less than accurate - from the start. The pollsters then ran the results through their formulas to try to take that factor out and then to "nationalize" their sample. They're obviously on a little bit of shaky ground there.
"Fabrication," as in "made up from whole cloth" is
not a neutral word and it's a real stretch to say that any of the media have done that, except for the ones that have been caught actually doing it. If I were one of the people surveyed, my first response would be "of course not" but then I'd think of the ones who've been caught and I'd be forced to answer "yes". So... by using that word, they pretty much rendered their survey meaningless. But they didn't render the results half true.
Like I said before, I think the survey should have been broken down by outlet and the language should have been more neutral - something more like "Do you think the reporting of _________ is accurate, somewhat accurate, etc." A breakdown by party preference or who they voted for would have made the survey more interesting but wouldn't have been necessary. It could have even made it less meaningful, since many of the responses from the different groups would be so predictable. The overall feelings would be far more meaningful.
Don't look at me, I just work here.