The "reputable surveys" done on "articles published in scientific journals" were exactly the same as those worthless example survey questions he listed.
End of psychoak's quote
There are a lot of reasonable discussions about the methodologies binning of papers into "support" and "don't support" categories. Your article that begins with "Global warming alarmists and their allies in the liberal media..." is clearly not one of them.
This basic debate as you can see if you actually read the parent Popular Technology article is about what it means to agree with Climate Change. The 97% study took a broad view. They basically took the view that if you thought that humans were causing substantial climate change, you were in the consensus. But, some people took a narrower view. They took the view that you had to mostly agree with very specific models (usually IPCC models).
This is a reasonable agreement about methodology that got blown up by the Forbes back into some grand conspiracy involving the liberal media and evil scientists.
A key paragraph:
“Cook et al. (2013) is based on a straw man argument because it does not correctly define the IPCC AGW theory, which is NOT that human emissions have contributed 50%+ of the global warming since 1900 but that almost 90-100% of the observed global warming was induced by human emission,” Scafetta responded. “What my papers say is that the IPCC [United Nations Intergovernmental Panel on Climate Change] view is erroneous because about 40-70% of the global warming observed from 1900 to 2000 was induced by the sun.”
(Scarfetta really doesn't like the IPCC as seen in the next paragraph in the article)
Basically, one party is saying "No, I don't agree because I don't match up with the IPCC models", while the other side is saying "well, the big picture is that you do agree because your models explicitly note that humans have a substantial effect on climate change, its just not as large as what is predicted by the IPCC models".
In any case, in any binning exercise where you have to binarize things into two categories, there will be problems along the edges. Ask anybody with any knowledge of machine learning or something. So the binning might not be perfect along the edges of the categories, but thats not surprise, and its a statistical exercise. Misclassifying a few papers out of a batch of 12,000 is not a big deal (although its inevitable that nutters will not understand this and make a big deal out of it). As I said at the beginning, there can and is reasonable discussion about the binning of the individual papers. But given the vast bulk of the papers, moving a few of them around isn't going to change the overall number much.