I have no doubt what constitutes evidence - and it is a very broad definition - anything at all.
where I probably differ from George is in what I consider USEFUL evidence.
As an example, someone coming on here and saying "chemtrails exist" is evidence that chemtrails exist - but I don't find it useful as evidence that chemtrails exist. When someone tells me that something exists I expect to see some objective evidence that supports the assertion - the sort of evidence that I could possibly gather myself in the right circumstances. Who is "someone"? What is the supporting data for his statement?
In fact I give such an unsupported statement so little weight as to make it useless.
If someone says "chemtrails exist because contrails disperse within (some time frame)" then I evaluate that against known facts - and there is a wealth of verifiable data that says that such a statement is false - contrails can last for hours. So it gets discarded too - but before I do so I have at least evaluated it against known factual information.
ditto with "elevated levels of aluminium" and the like - I evaluate them against known standard - like EPA limits - and I evaluate the process - and invariably I have found that limits have not been reached - often because the limit is for one thing (usually water) but the test was done on something else (usually dirt, soil, sludge)
The most useful evidence is verifiable - in theory anyone could do the same work and get the same results. Scientific papers that have been peer reviewed and their results replicated are the strongest evidence. Other papers that have been peer reviewed but perhaps not replicated yet, but which explain what happened, provide mechanisms for how it happened that are within the current knowledge base, etc., are pretty good too - but not quite so much.
The next most useful evidence is testimony (or papers, blogs, writings, etc) from people who say that something has happened, or exists, etc., and who have a known connection to the subject matter (qualifications, work there, etc) and are likely to know what they are talking about, even if the information is not directly verifiable. Papers not yet peer-reviewed by published by people with a history in the field and who provide sound reasoning would be the best evidence at this level - blogs by "experts" less so - depending on the level of expertise and the quality of the info offered, etc.
"contrail science", for example, is a blog with a great deal of information on it - but it is not peer reviewed (obviously). However the information on it is supported by a great deal of information from sources that have clearly done a great deal of work in the area, or by a significant amount of analysis (eg the "missile off california" case) and reasoning, so I find it generally useful evidence.
And so on down the chain.