Interest in the spread of fake news and misinformation online has increased dramatically since the 2016 US presidential election, and the relevance of misinformation to politics has only grown during the Covid-19 pandemic. However, we know little about levels of belief in fake news encountered shortly after publication, as well as what types of people are more likely to believe fake news. To address this gap in the literature, we fielded two studies in which we repeatedly asked representative samples of Americans to evaluate popular articles from non-credible and credible sources within 24-48 hours of their publication.

We find that, on average, false or misleading articles are rated as true 33.2% of the time; moreover, approximately 90% of individuals coded at least one false or misleading article as true when given a set of four false or misleading articles. While most demographic characteristics co-vary only slightly with the likelihood of correctly identifying fake news stories, we find a very strong relationship for ideological-congruence: both conservatives and liberals are much more likely to believe false/misleading information if it reflects their ideological perspectives than if it does not, and this effect is stronger than previously measured in other studies. We also find that older respondents are less likely to believe false/misleading information relative to younger respondents. This adds important context to the existing findings that older Americans are more likely to share fake news on Facebook [4], suggesting that sharing behavior may be divorced from actual belief.

Finally, using two different experiments, we tested whether a common prescription for lowering belief in fake news, encouraging news consumers to search for information about a fake news article, actually reduces belief in fake news. Paradoxically, we find that encouraging respondents to search for information to inform one's evaluation of an article's veracity increases the likelihood that an individual believes that fake news article. Evidence from real-time Google searches suggests that this pattern is driven by the existence of similar information elsewhere on the internet, even when the quality of that "supporting" information is also low. In our two experiments, we find that encouraging respondents to search for information only increases belief in false/misleading stories when Google search engines return information from low-quality sources that corroborate claims made in that fake news article.

When running a second study using the same survey instrument, but strictly limiting our articles to those about Covid-19, we find that slightly more respondents believed popular fake news about Covid-19 than general fake news articles. We also find that the salience of partisan divisions remains. Both the strong effect of ideological congruence and age identified in our first study hold when we focus our analysis strictly on fake news articles about Covid-19. Similarly, we again find that encouraging respondents to search for information about fake news articles Covid-19 also increases their belief in this information. This indicates that attempts by Google during the Covid-19 pandemic to remove or mitigate data voids may not have been as successful as we might have hoped.

These studies show that belief in online fake news is higher than previously identified [3,6,2,7,5,1] and suggests that efforts to encourage people to search for corroborating information may be exacerbating this problem.