The media is prone to tout new studies that claim sweeping results. “Coffee can cure cancer.” “Red wine is better for you than exercise!” However, most media reports on scientific topics become grossly distorted as they go from the study abstract to the listicle, similar to a game of telephone. Last Week Tonight host John Oliver explained this thoroughly well in his latest segment on scientific studies.

Oliver points out several things that are often overlooked when studies get published and covered in the media.

As their success largely depends on getting research published, scientists are working to get as many studies published as possible. Sometimes, they’ll tweak methods until they can produce something that’s seemingly a big link, even if it doesn’t really mean anything in the big picture.

“To get those results, there are all sorts of ways that consciously or not, you can tweak your study,” Oliver says. “You could alter how long it lasts, or you could make your random sample too small to be reliable, or engage in something that scientists call ‘p-hacking.’” Also known as “data-fishing,” p-hacking is essentially the process of sifting through data to find certain patterns that could be viewed as statistically significant. Unless you’re a scientist or a statistician, of course, it’s often difficult to guage how significant the results actually are in a greater context.

Replication studies, aka scientific fact-checking, aren’t done as often as they should be. The best way to confirm results from a study is to have another group of scientists, with different methods and sample sizes, repeat it and see if the link is still there. Oliver notes that these replication studies simply aren’t completed often enough.

Press releases will often change and exaggerate results, and journalists in turn often take those headlines and make them even more sensational or sexy for the web or TV. It’s “like a game of telephone, the substance gets distorted at every step,” Oliver said. “I could only imagine how someone who watched that [TV news] segment must describe it the next day.”

Finally, many studies are done in rodents, and the results often can’t be transferred over to humans without actual clinical studies. For example, if a study done in mice finds that lip balm can fight certain types of cancer cells, there’s no telling that the same results will appear in humans.

There are far more reasons why we can’t take all we read as actual scientific truth, but at the end of the day, it’s important to read articles and even studies themselves with healthy skepticism.