It's wonderfully easy to contribute on the Web, and as an unfortunate side effect of this essential fact, hoaxes, cranks, and general nonsense abound. I'll euphemistically call this stuff “junk”. Because so much of what you see online is junk, smart people such as yourself develop a finely-tuned junk detector. This is fine—in any case it's important to have one if you plan to use the Web for anything serious.
But your junk detector is probabilistic, factoring in grammar, habits of speech and writing, vocabulary, the writer's opinions and personality, whether the page has pictures of kittens—anything but the actual argument itself, because the whole point of the junk detector is to avoid wasting the time of reading it. In other words it's like the worst possible use of ad hominem, a logical fallacy. You guess as much as you can about the author, then judge the value of the page based on that. I see no good way around this. Consequences:
Your main way of evaluating the quality of Web pages is subconscious.
The junk detector is not as accurate as actual critical thought.
False positives mean the reader misses out and the writer fails to connect (making good writing skills more important now than ever before).
False negatives mean you may be duped: the junk detector doesn't protect you from lies, logical fallacies, or really sophisticated forms of “junk”. By the time you decide to read the whole page, the junk detector is done working. Another, smarter junk detector had better kick in!
All of this applies in the non-Web world, too, but the Web is so full of junk, and it's so hard to avoid altogether, that the cheapest possible junk detector is highly rewarding and can instill a false sense of confidence.