Lots of heuristics. As many have mentioned, "a priori, don't believe anything on the internet" is a great starting point. You wouldn't trust a complete stranger as much as your best friend; almost everyone is a complete stranger on the internet, with the bonus of having a megaphone and opinions to spread. On the internet, no one knows you're a dog. But actually, most sources are humans, and most want to just share their perspective (as opposed to knowing the truth and deliberately lying), however flawed it might be.
Follow the money.
You used to be able to rely more on coincidence counting - meaning more sources saying the same thing means its more likely to be true. That has way less weight when one tweet gets shared, articles written about it, then people tweet the articles, and it comes full circle. You can manifest "facts" easily. And most news reports the same original source. But it's still useful if you can triangulate with multiple high-quality sources, or sources of conflicting persuasion. NBC reporting X is worth 1 unit. NBC, ABC, CNN reporting X is still just 1 unit. But CNN, Fox, Al Jazeera, and AP all reporting X is worth 2-3 units. There should be a logarithmic curve to your confidence. Any situation could flip on its head.
Think like a Bayesian. Every statement should have some prior probability of being correct, near 50/50. Increase the weight if the source is good, it jives with other things you are confident in; decrease it if the sources is bad or it doesn't sit right. De-rate "shocking" information - it tends to outspeed quality but nuanced info.
Knowing who to distrust is often more useful than knowing who to trust. It's often easier to infer as well. Create a mental "trustworthiness score". You start by identifying low quality reporting or straight up lies. This drops a sources' score. Other news from that source is also likely to be bad. Other sources that cluster near this source are also likely to be bad.
All media can be faked (manually or with AI), but at this time, deepfakes have some glaring limitations, and most fakes are in fact "cheapfakes" of the lowest possible effort. Why spend time photoshopping when you can just slap a misleading caption on an old photo?
The information firehose - tweets, shares, comments sections - can be useful for cutting edge info and building the big picture, but wielding it is an art. Most is gossip. Often the most popular/liked opinion is simply wrong, but also often, the truth is there deeper in the thread, but buried. Bullshitting is easy, but speaking facts is hard and time consuming. Look for "rich"/"deep" comments that show someone doing their homework.
I'm loathe to recommend any particular sources (you should develop your own), but I'm partial to the OSINT groups like Bellingcat. AP and Reuters are decent for "mainstream" media.