There is no one criterion. It's the surrounding context that matters. Sometimes it's my familiarity with the field (works with most technology and science reporting, plus with articles that engage in the most obvious forms of manipulative charts and statistics). Sometimes it's other conflicting reporting (e.g. the case of a march I reported elsewhere, where independent evidence from around the web disproved a staged photo in the article, and the conclusion it was meant to support). Sometimes it's people with first-hand knowledge or the subject of the article themselves calling reporting into question. Sometimes it's a combination of all three.
Ultimately, it's a matter of trust. In any given situation, how much can I trust my knowledge, or my ability to interpret surrounding context, or my ability to judge the trustworthiness of a third party, vs. how much I can trust the article itself? I do my best to increase my knowledge and hone my judgement, and while I am and will be making mistakes, the current state of news reporting isn't presenting a particularly high bar here.