The place to start is by evaluating what you read. They might be right, they might be wrong, but you don’t know unless they tell you where they got their information. So if the article does not cite a study by title and give the journal reference, ignore it.
If there is a citation, follow it to the actual study and evaluate it. Is it an epidemiological study? If so, it is likely to be unreliable. You can do a cursory check by looking at a few numbers. First, how many people were studied (N = ?)? If it was only a few, the results will be very noisy. Second, how statistically reliable was the observed effect (P < ?)? If p is not less than 0.001, the effect is not very reliable. If, however, they say p < 0.00001, that’s something to pay attention to. Lastly, how large is the clinical effect? If it’s not at least 2.0, then it’s clinically irrelevant, even if it’s statistically significant.
Epidemiological data at their best can only show an association (or lack of one). Austin Bradford-Hill came up with some criteria, which, if followed, can perhaps be used to establish causality, but they involve extremely large effects, very high statistical significance, large populations, and other major considerations, plus there has to be a plausible mechanism of how the effect is achieved. In any other situation, epidemiological data can never do more than suggest hypotheses for further study.
If the cited study is a randomised controled trial, then it’s worth paying more attention to. But still, the criteria above still apply: effect size, statistical significance, N studied. Then you look at how the study was designed. Does their definition of “low-carb” or “keto” match ours? Was the study double-blinded? If not, why not? Who paid for the study? (If it was paid for by a drug company, the statistical certainty of a favourable result approaches 1.0.) Do the researchers appear independent? Do they appear to have an agenda?
Once you have looked at all these things, you are in a much better position to evaluate “a lot of articles written” that say this or that.