When Readers Do Not Fight Falsehood: An Exploration of Factors Influencing the Perceived Realism of False News on International Disputes

Abstract

This study examines the effects of misleading news—one type of false information presented by news media in the U.S. and China—in the context of international disputes. Through a web-based survey experiment, we tested how Chinese readers’ perception of false news is affected by the source of the news, the presence of visual elements, and general trust in mainstream Chinese media and that in mainstream U.S. media, as well as news literacy. Our results suggested false news reported by domestic media was perceived to better represent the reality of the covered issue than news presented by foreign media. This relationship was moderated by readers’ general trust in U.S. media and news literacy, which indicated media literacy training as a possible solution to counteract the effect of the news source. These findings not only advance current scholarship on misinformation by incorporating perspectives from non-Western media systems but also provide both foreign and domestic readers with timely and relevant methods to combat false information.

1. Introduction

The rise of misinformation after the 2016 U.S. election has revived scholars’ interest in the propagation and distribution of so-called “fake news”. Until 2012, the concept of fake news in the context of journalism included genres such as satire and parody, but recent scholarship has expanded the definition to fabrication, manipulation, and propaganda (Tandoc et al. 2018). According to this school of thought, fake news refers to “the intentional presentation of false or misleading claims in order to manipulate the audience’s cognitive processes” (Biloš 2019, p. 1139). Other scholars have created further typologies, where fake news is classified as intentionally misleading and is differentiated from false news, which is theorized as unintentional misinformation (Lazer et al. 2018). Related to this, others have differentiated among misinformation, disinformation, and malinformation, three terms that are often used in relation to fake news. The first refers to sharing false information without the intent of causing harm. The second refers to sharing false information knowingly to cause harm, and the third refers to instances when genuine but private information is shared publicly to cause harm (Wardle and Derakhshan 2017). Studies on the broader topic of fake news fall into five categories: typology, impact, dissemination, detection techniques, and ways to combat fake news (Wang et al. 2020).

Most existing research, however, focuses on Western nations’ domestic news production and consumption (Jahng et al. 2023; Tandoc et al. 2021; Wang et al. 2020). Lesser attention is paid to the comparative aspect such as comparing fake news coverage between different nations (but see for exception, Humprecht 2018). Biloš’s (2019) meta-analysis of academic work on fake news found that among 393 scholarly articles published in 57 countries and regions, more than 68% were published in English-speaking countries such as the U.S., England, Australia, and Canada. A relatively smaller number have examined this phenomenon in non-English speaking countries such as Russia and South Korea. Even when scholars adopt a cross-national perspective to investigate the dissemination of false information, it usually focuses on a comparison of Western democracies (Humprecht 2020).

Yet, as many scholars point out, false news as a global phenomenon can be shaped differently by distinct news cultures (Humprecht 2020). Fighting false news is also shifting to a “whole-of-the-world” approach that seeks to address both foreign and domestic vulnerabilities. For example, recent U.S. elections have been infused with rising concerns about the inaccurate, misleading, and fully fabricated information disseminated by other non-U.S. countries on social media that could erode people’s faith in the integrity of democratic elections. Similar concerns also emerge in the Russian–Ukraine war, the Israel–Palestine war, and other international affairs. As such, a comparison between democratic and authoritarian nations can further our understanding of how this problem manifests differently in non-Western media systems. Unlike domestic media, which often enjoys a large readership, international media outlets attract fewer readers from other nations (Thurman 2007). False news circulating on social media, however, can be powerful in setting the agenda for key issues on international relations, economy, and religion. The effect of such misinformation is stronger in countries characterized by low media trust and polarized political environments (Gupta et al. 2023; Humprecht 2018; Ognyanova et al. 2020).

As a response to the above urgent issue, this study examines the public’s perceived realism about false news circulated during international disputes involving the U.S. and China; here, perceived realism is defined as the degree to which individuals feel the false story can represent the reality of a related phenomenon (Balmas 2014). We conducted an online survey experiment to examine how Chinese audience’s perceptions of false news about the 2019 U.S.–China trade war are affected by the news source, presence of visual elements, general trust in mainstream Chinese media versus mainstream U.S. media, and media literacy. With these efforts, this study aims to not only advance the current scholarship on misinformation from the non-Western media systems’ view of point, but also to yield relevant and up-to-date empirics for the broader public who are constantly seeking antidotes to counteract false information.

1.1. The Propagation of False News in International Affairs

The years after the 2016 U.S. election saw an increase in the circulation of fake news. A survey of U.K. residents found that more than one-third of respondents had “come across a news story, believed it to be true, then later realized that it was either exaggerated, inaccurate or blatantly false” (Flintham et al. 2018, p. 376). Fake news was found to spread faster when fueled by global panic. For example, misinformation around COVID-19 led to a more than 900% increase in the number of English-language fact-checks in a span of two months (Brennen et al. 2020).

Factors related to audiences, journalism, and technology can cause the propagation of false news around international affairs. Generally speaking, audiences tend to be less informed about international affairs than about domestic issues (Lanoszka 2019). They tend to rely on popular beliefs or conspiracy theories to evaluate the accuracy of messages from unfamiliar foreign sources. This information void occurs largely because humans tend to take less notice of faraway threats compared to events of importance that occur closer to home, which have a greater potential to affect their lives. As a result, audiences are likely to rely on their political ideology, heuristics, personal experiences, and prejudice in order to form opinions about international affairs. They also tend to treat positive news about their adversary nation with skepticism (Lanoszka 2019). When the audience is already primed to believe false news, the likelihood of its propagation increases. However, most of these concerns emerged from studies on factual news published by legitimate foreign media outlets. Although these mechanisms are theoretically relevant to how people may process false news, there remains a lack of empirics to test whether people would treat false news differently depending on whether it is from their familiar domestic media or from their less familiar foreign outlets. This present study aims to fill in this gap with a survey experiment to delve into this question.

False news could also be promoted inadvertently by journalistic practices and media organizations (Himma-Kadakas 2017), though some scholars contest this idea, proposing instead that news makes people resilient to misinformation (Altay et al. 2024). Due to the pressure of producing content for a fast-paced news cycle, journalists are forced to skip fact-checking processes or select click-bait news items that are cheap to produce (Reilly 2018). False news mimics the news values associated with traditional news (Khaldarova and Pantti 2016; Sui et al. 2023), and without appropriate checks, such content could be inadvertently published and distributed on mainstream channels. Since traditional media outlets are considered credible, news published by them—though, later on, found to be false—could spread fast and eventually mislead people.

Technological and political factors also propagate misinformation (Altay et al. 2023). One example is “political spamming”, where active Twitter users flood social media channels with tweets and hashtags connecting news organizations with false news, thereby undermining their credibility. In their analysis of the hashtag “fakenews”, Al-Rawi et al. (2018) found that liberal news organizations, especially CNN, were frequently mentioned in such posts on Twitter. CNN was also connected with other related hashtags such as #FakeNewsCNN, #CnnIsIsis, and #FraudNewsCNN. The connection of CNN to the fake news discourse was created by criticism from former president Trump and members of the Republican party. Bots and automated accounts spread the tweets from these political elites, which increased the audience for these messages. Al-Rawi et al. (2018) concluded that spamming bots could influence public opinion and determine the tone of the public agenda, thereby undermining the credibility of mainstream media.

1.2. Factors That Affect Readers’ Perceived Realism of False News

As the current society witnesses a rapid spread of unverified or fabricated content—especially on social media—it is imperative to look for effective ways to help audiences discern factual information. Indeed, through repeated exposure to false information, people’s trust in the media in general decreases. In the meantime, because false news tends to mimic factual news, the blending of the “perceived realism” of the content itself—that is, at least part of the information in the false story is believed to represent the reality—with suspicious sources can further deteriorate people’s capacity to identify or refute false news. Eventually, people may foster a sense of disillusionment, inefficacy, and cynicism, which could extend to their daily consumption of factual or real news. These concerns point to the importance of addressing the ways in which audiences perceive and interpret the realism of false news.

In the presence of competing narratives, audiences use various cues to evaluate the veracity of information (Szostek 2018). These cues include the relevance of news (Flintham et al. 2018; Tandoc et al. 2018), existing beliefs (Bronstein et al. 2019), and media consumption patterns (Nelson and Taneja 2018). For example, people are interested in finding out the veracity of news only when it is relevant to them. On the other hand, fake news consumers are likely to be heavy internet users, and people who are predisposed to believe in conspiracy theories are more likely to believe in fake news. The realism of international news could also be judged on the basis of personal experiences and cross-border relationships. For example, in his analysis of Ukrainian audiences’ news-seeking habits, Szostek (2018) found that people used information from family and friends to make sense of media narratives and decide which narratives to trust.

Other factors include prior exposure (Pennycook et al. 2018) and audience availability (Nelson and Taneja 2018). People who come across the same fake news item multiple times on social media are likely to perceive it as accurate, regardless of how partisan or implausible it sounds (Pennycook et al. 2018). The effects of this could be mediated by audience availability, which posits that heavy media users have the time to explore varied media offerings and are more likely to come across niche news sites offering extreme views, while light media users follow traditional and popular media sources. This creates a pattern where well-known, mainstream media organizations enjoy larger audiences than niche sites that are likely to carry misinformation (Nelson and Taneja 2018).

Beyond the above general considerations, this paper explores the effect of five factors that may contribute to readers’ perceived realism of the false news they encounter: source cues, structure of news content (topics and images), media trust, and media literacy.

Source cues. Audiences habitually rely on familiar sources to obtain credible news (Flintham et al. 2018). When they consume news from unfamiliar sources, they use content, social, and platform cues to judge the trustworthiness of the content (Ross Arguedas et al. 2024). They also use institutional news sources to verify news encountered on social media (Tandoc et al. 2018). Since people obtain information from multiple sources (Hmielowski 2012), they are likely to be exposed to fake news as well as factual news from traditional sites. When people have higher levels of exposure to fake than factual news, they are unable to distinguish between the two. However, people who consume factual and fake news in equal degrees do not find it difficult to differentiate between the two (Balmas 2014).

Structure of news content. News items that are serious in tone or deal with familiar topics are perceived as credible, while those that use an informal writing style and are humorous or satirical are more likely to be considered fake (Flintham et al. 2018). Audiences also feel more invested in determining the veracity of news articles that deal with issues of relevance to their lives. News that reiterates readers’ own views—such as the ones they have learned repeatedly from their primary media outlets—is likely to be considered more credible than those that challenge pre-existing views (see Metzger et al. 2010).

When a news source is unfamiliar, readers use cues derived from the design, layout, and navigation features of a website to determine source realism (Flanagin and Metzger 2007). Images are particularly useful cues; for example, audiences tend to believe information conveyed via images without fact-checking them (Wineburg et al. 2016), and this is especially true for images published on digital platforms (Kasra et al. 2018). However, this effect on perceived realism is moderated by external elements such as audience perceptions of credibility about the image source and the media source displaying the image. When audiences encounter news images on social media, their perceptions of realism are affected by their digital skills, media literacy, and prior knowledge rather than the actual content of the image (Shen et al. 2019).

Media Trust. People with low levels of media trust believe that the media are biased in their treatment of sources and coverage of issues and that news content is solely driven by profit motives (Ardèvol-Abreu and Gil de Zúñiga 2017; Tsfati and Cappella 2003; Wilner et al. 2022). People using mainstream media sources are likely to have high levels of media trust compared to those who use fringe sources of information (Tsfati and Ariely 2014). Relatedly, people who distrust the media seek out alternative sources and are less likely to be exposed to mainstream news (Tsfati and Cappella 2003). Thus, people with high levels of media trust are likely to have high perceptions of realism for the news they consume on mainstream sites as compared to alternative sites.

Media Literacy. Media literacy is often seen as an antidote to fake news. The expectation is that if people consume news from a variety of sources, and if they have the critical thinking skills to analyze information, they will be able to differentiate between real and made-up stories (Jeong et al. 2012). Digital media literacy includes information literacy, i.e., locating and managing information online, digital photography, the ability to search databases, and being able to identify opinions from facts (Jones-Jang et al. 2021). It also includes being able to adapt to the constantly evolving features of websites and social media and being aware of internet-related terms (Shen et al. 2019). Lastly, news literacy, which focuses on people’s understanding of the role and functions of media, and their ability to analyze news and produce it, is also known to affect fake news perceptions (Ashley et al. 2013). People with high levels of media literacy presumably have low perceptions of realism toward fake news and are able to distinguish between real news and fake news.

1.3. Hypotheses

Putting these considerations together in the context of international news, we expect that mainstream and familiar news sources and topics will invite attention and be considered legitimate compared to foreign news sources and topics that are unfamiliar. Articles that contain images are also more likely to be considered real than those that rely on extensive arguments to make a point. These effects will be moderated by factors such as media trust and media literacy. Specifically, we hypothesize the following:

H1. 

Readers’ rating of the perceived realism of a false story would vary by news source (H1a) and the presence of news image (H1b).

H2. 

The effect of the news source on the perceived realism of a false story is conditional on the presence of a news image (H2a) and the news topic (H2b).

H3. 

The effect of the news source on the perceived realism of a false story is conditional on their general trust in domestic media (H3a), general trust in foreign media (H3b), and media literacy (H3c).

2. Materials and Methods

Data were collected from a web-based survey experiment1 fielded in 2019, with a sample of 555 students recruited from a large public university in China. This survey experiment was conducted in Chinese. Through a 2 × 2 × 3 between-subjects design, this study manipulated the news wire the respondents were exposed to (Associated Press vs. Xinhua News Agency), whether an image was used in the news text, and the topic of the story (“No more Chinese students accepted by MIT due to trade war”, “China cancels U.S. pork due to trade war”, “China recalls giant pandas from U.S. zoo amid bitter trade war”).

2.1. Participants

Participants were college students aged 16 years and older (M = 21.73, S.D. = 3.73). Most were undergraduates (81.20%), with a small portion being master’s (18.02%) and doctoral students (0.78%). With regards to income level, the majority reported an annual household income of CNY 124,999 and low (69.27%), with a median of CNY 75,000–99,999 per year. This is much higher than the median household income of Chinese residents (CNY 26,532), as reported by the 2019 census data (National Bureau of Statistics of China 2020). The sample also includes a bigger percentage of women (70.93%) than the national population (48.69%), as reported by The World Bank (2019). About 30% reported their English proficiency to be intermediate (can fluently use English for everyday activities) and above.

Though student subjects are often considered non-representative and a limitation for external validity (Benz and Meier 2008), recent research shows that when the experimental study’s treatment effect is heterogenous across various segments of the target population—where a student sample is likely to misestimate the effect—even a representative sample may fail to accurately estimate the treatment effects unless extensive moderation models are performed to control for the heterogeneity related to respondents’ characteristics, the context and time of a study, etc. (Druckman and Kam 2011). Thus, most experimental studies still employ a student sample for causal tests.

2.2. Stimulus Materials

This study administered an online experiment where three factors were manipulated within a news story published on a Chinese social networking site Sino Weibo. To maximize the generalizability of this study’s findings, we referred to Sina’s channel dubbed “Rumor Monitors” (http://piyao.sina.cn/, accessed on 15 August 2019) and chose real-life stories to be our stimulus materials. This channel was established for Sino users to report and filter false posts, and we selected a few posts about the U.S.–China trade war that were published between December 2018 and May 2019.

The selected stories were derived directly from “Rumor Monitors”, with minor edits to language.2 These stories had misleading arguments in the main text that lacked factual evidence, which could elicit varied levels of perceived realism among participants. As found in recent research, the dominant type of misinformation involves reconfigurations, “where existing and often true information is spun, twisted, recontextualized, or reworked” (Brennen et al. 2020, p. 3). Relative to wholly fabricated messages, this so-called “reconfigured misinformation” enjoyed more readership and higher engagement (Brennen et al. 2020). Specifically, Brennen and his colleagues (2020) defined misleading content—the most common form of misinformation—as content that “contained some true information, but the details were reformulated, selected, and re-contextualized in ways that made them false or misleading” (p. 15).

Following these criteria, we chose three stories where facts were mixed with false statements. For example, the MIT story focused on the small number of admissions MIT issued to mainland China applicants in 2019, which was attributed to the growing tensions between the U.S. and China under the trade war. As exhibited in

Table 1. Stimulus stories explained.

	True Statements	False Statements	Supporting Evidence *
MIT story	“Amongst the students that were recently announced for MIT’s early admission program, no Chinese students—though the prospective students’ nationality information was not publicized—were accepted by this year’s AE program”.	“Affected by the escalating US-China trade war, no students from mainland Chinese schools were admitted to the Massachusetts Institute of Technology (MIT) this year”.	Two students from the Chinese mainland were accepted by MIT in the regular action period in 2019, Chinese study-abroad agency Tiandao said in an April article. http://piyao.sina.cn/piyao/article/20190514/04d4f5e786851000.html (accessed on 15 August 2019)
Panda story	“China has recalled two cute giant pandas from the San Diego Zoo in California. The two animals, 26-year-old Bai Yun (meaning “white cloud”) and her six-year-old Xiao Liwu (“little present”) were iconic images of the zoo and aided important worldwide research in the protection, breeding, disease control of giant pandas. Both majestic animals were sent back to their native China after the country scrapped its conversation loan agreement with the U.S”.	“Due to the escalating US-China trade war…”	The two pandas were recalled as the loan agreement between the two countries expired, not due to the trade war, according to an article by Daily Mail. http://piyao.sina.cn/piyao/article/20190519/04db0ab815051000.html (accessed on 15 August 2019)
Pork Story	“The same week U.S. President Donald Trump announced sweeping increases on tariffs against Chinese goods, Chinese buyers dropped orders for 3247 metric tons of U.S. pork…”	“… totaled up to $6.5 billion…”	USD 6.5 billion was not the total value of this pork cancellation; instead, it was the total value of the U.S. pork export market estimated for 2019. http://piyao.sina.cn/piyao/article/20190520/04dc729328c51000.html (accessed on 15 August 2019)

Note: The first two columns are true and false statements included in each story, and the last column (with a superscript *) is relevant evidence to refute the false information presented in each story.

, this story includes a partially true statement that “no student from mainland China was admitted to Massachusetts institute of Technology in 2019”; however, what was omitted from this piece was that this zero-admission rate only applied to 2019’s early admission program. In addition, although this story attributes the falling admission rates to the U.S.–China trade war, no evidence was found to support this argument. Similar patterns were identified in the other two stimulus stories, as detailed in Table 1.

The second treatment, the source of a news story, varied between a Western news agency “Associated Press” and a Chinese news wire “Xinhua News Agency”. To maximize the realism of the stimulus, this study used the exact handles and profile images these two news agencies have for their Weibo accounts, as displayed in Appendix A.

The third factor, visual elements, was manipulated by exposing participants to news posts with or without a news photo. To ensure the stimulus posts looked realistic, each news post was formatted to the same style as every Weibo post, where the Weibo handle was displayed first, followed by publication time/date, path, and textual/visual content. See Appendix A for details.

2.3. Procedure

Participants started with a consent form that informed them of the purpose of this study—which was to understand university students’ opinions on online news, as well as to capture their views on a brief news report they would read in this survey—and that their participation was voluntary and would be compensated with a monetary reward (USD 1.00) upon completion. After digitally signing this informed consent, participants were first asked to answer pre-test questions such as the use of traditional and online media for news about politics, the use of social networking platforms, Internet skills, attention to news, etc. Next, participants were randomly assigned to 1 of the 12 experimental conditions, in which they were asked to read a distinct short news post published on Sino Weibo, where the total amount of time spent on the story page was also recorded (M = 20.87 s, SD = 27.62). Following this, participants answered questions about the story they just read and demographics. At the end of the survey, respondents were assigned a verification code for them to receive the monetary reward. In addition, they were also told that the news reports they read in this study were fact-checked to have included misleading information—which was used for research purposes only—and that they should not share them online.

2.4. Variables and Measurements

The perceived realism of a false story is the primary outcome variable in this study, which was operationalized as the degree to which individuals feel the false story can represent the reality of a related phenomenon (Balmas 2014). Accordingly, participants were asked to assess the question, “To what extent does the story you just read represent the reality of the U.S.-China trade war?” on a 101-point scale ranging from 0 (“not at all”) to 100 (“very much”). This variable (M = 55.10, SD = 22.39) was captured immediately after the participants’ exposure to the news stimulus.

General trust in Chinese media and general trust in U.S. media are the primary moderating variables used in this study, which were operationalized using three items commonly adopted in media and message credibility scales (Flanagin and Metzger 2007) and were adjusted to gauge the overall credibility of Chinese [U.S.] media. Respondents were asked to rate the following three items on a 7-point scale, where 1 represents “strongly disagree” and 7 represents “strongly agree”: “Generally speaking, Chinese [U.S.] news outlets are trustworthy/accurate/tell the whole story”. Responses were computed into a unidimensional scale for analysis (general trust in Chinese media: Cronbach’s α = 0.88, M = 4.40, SD = 1.23; general trust in U.S. media: Cronbach’s α = 0.87, M = 3.79, SD = 1.05).

Media literacy is another moderating variable of interest, which was operationalized as individuals’ personal belief in their understanding of news production (Vraga and Tully 2015). Specifically, participants were asked to rate on a seven-point scale, ranging from “strongly disagree” [1] to “strongly agree” [7], five items: “I have a good understanding of the concept of media literacy”, “I have the skills to interpret news messages”, “I understand how news is made”, “I am confident in my ability to judge the quality of news”, and “I am confident in my ability to fact check news”. Accordingly, these five items were computed into a unidimensional index (Cronbach’s α = 0.85, M = 4.67, SD = 1.04).

2.5. Manipulation and Randomization Check

This survey experiment also included a set of demographic factors for a randomization check, as well as two other questions intended for a manipulation check.

As displayed in Appendix B, each of the three treatments was regressed against age, gender, education, household income, English fluency, general media credibility, habitual SNS usage, internet skills, media literacy, and prior exposure to news about the trade war and the U.S. ban on Huawei. In all models, none of these predicting variables were statistically significant, indicating that the assignment of participants to the 12 experimental conditions was independent of respondents’ demographic characteristics. This lends support for a successful randomization procedure.

We then proceed to examine the validity of the two manipulated treatments, including the source and topic of the assigned news posts, among participants across all conditions. About 64.60% correctly recalled the source of the news post they were asked to read, and 82.85% correctly recalled the assigned story’s topic. This provides moderate support for the successful injection of the two treatments.

3. Results

We first evaluated whether the respondents’ rating of the perceived realism of the false story varies by news source (H1a) and the presence of a news image (H1b). As stated above, the outcome variable “perceived realism of false story” was measured on a 101-point scale ranging from 0 (“not at all”) to 100 (“very much”). A log transformation was used to curve for the skewed distribution of this variable. Accordingly, we ran an ordinary least squares (OLS) regression, where the perceived realism of false information was regressed against the news wire treatment (AP vs. Xinhua), the news image treatment (presence vs. absence), and the news issue treatment (pork import cancellation, panda recall, and MIT admission).

As shown in Model 1 in

Table 2. OLS regression models predicting perceived realism (logged) as a function of manipulated treatments.

	Model 1 (H1a and H1b)	Model 2 (H2a)	Model 3 (H2b)
AP (vs. Xinhua)	−0.09(0.05) #	−0.11(0.08)	0.13(0.10)
News image (vs. W/O image)	0.02(0.06)	−0.01(0.08)	0.01(0.06)
Pork story (vs. MIT story)	0.15(0.07) *	0.15(0.07) *	0.24(0.10) *
Panda story (vs. MIT story)	−0.20(0.07) **	−0.20(0.07) **	0.05(0.10)
News Wire X Pork story	—	—	−0.16(0.14)
News Wire X Panda story	—	—	−0.49(0.14) ***
News Wire X News image	—	0.04(0.11)	—
Constant	3.94(0.06) ***	3.95(0.07) ***	3.83(0.07) ***
F	7.03 ***	5.65 ***	6.99 ***
Adjusted R2	0.04	0.04	0.06

Note: N = 555. The dependent variable was logged to curve for skewness. Entries are coefficient estimates with standard errors in parentheses. News wire is a dichotomous variable, with “Xinhua” coded as 0 and “AP” coded as 1. ^# p < 0.10, * p < 0.05, ** p < 0.01, and *** p < 0.001.

, the use of a news image had no main effect on the perceived realism of a false news story, when holding the other two treatments constant. The news source treatment yielded a statistically significant effect: relative to stories that were said to be published by Xinhua News Agency, those from the AP (b = −0.09, p = 0.06) were perceived as less reflective of the U.S.–China trade war reality. Thus, we found support for H1a but not for H1b.

We then tested whether the effect of news source is moderated by the presence of a news image (H2a) and news topic (H2b), using two OLS regression models. As displayed in Model 2 in Table 2, the coefficient estimate on the “News Wire X News image” interaction term (b = 0.04, p = 0.692) was not statistically significant. Thus, H2a was not supported. Model 3 in Table 2 revealed support for H2b, as the coefficient estimate on the “News Wire X Panda story” interaction term (b = −0.49, p < 0.001) was statistically significant. For a more intuitive interpretation of this interaction effect, Figure 1 plots the difference in perceived realism of false news between the AP and Xinhua across three news topics: the contrast was not statistically significant for the MIT and “pork import cancellation” stories; yet, on the topic of “panda recall”, the fake story was perceived to better represent the reality of the U.S.–China trade war when reported by Xinhua (M = 3.89) than when the same story was reported by the AP (M = 3.53).

Finally, we explored whether the effect of news source on the perceived realism of the fake story was moderated by readers’ general trust in Chinese media (H3a), general trust in U.S. media (H3b), and news literacy (H3c). To test these hypotheses, we ran three OLS regression models with an interaction term included. As displayed in Model 4 in

Table 3. OLS regression models predicting perceived realism (logged) as a function of news wire, moderated by general media trust and media literacy.

	Model 4 (H3a)	Model 5 (H3b)	Model 6 (H3c)
AP (vs. Xinhua)	−0.03(0.21)	−0.47(0.22) *	0.55(0.34)
General Trust in Chinese Media	0.08(0.03) *	—	—
News Wire X GTCM	−0.01(0.05)	—	—
General Trust in U.S. Media	—	−0.03(0.04)	—
News Wire X GTUM	—	0.10(0.05) *	—
Media Literacy (logged)	—	—	0.39(0.14) **
News Wire X Media Literacy (logged)	—	—	−0.43(0.22) *
News image (vs. W/O image)	0.03(0.06)	0.03(0.06)	0.03(0.06)
Pork story (vs. MIT story)	0.15(0.07) *	0.16(0.07) *	0.15(0.07) *
Panda story (vs. MIT story)	−0.20(0.07) **	−0.18(0.07) *	−0.21(0.07) **
Constant	3.58(0.15) ***	4.04(0.16) ***	3.36(0.22) ***
N	546	539	535
F	6.51 ***	5.17 **	6.28 ***
Adjusted R2	0.06	0.04	0.06

Note: The dependent variable was logged to curve for skewness. Entries are coefficient estimates with standard errors in parentheses. News wire is a dichotomous variable, with “Xinhua” coded as 0 and “AP” coded as 1. * p < 0.05, ** p < 0.01, and *** p < 0.001.

, the coefficient estimates on “News Wire X GTCM (general trust in Chinese media)” were statistically insignificant (b = −0.01, p = 0.784). Thus, H3a was not supported.

Consistent with our assumption in H3b, the effect of news source on the perceived realism of a false story was conditioned by respondents’ general trust in U.S. media (Model 4, Table 3: b = 0.10, p < 0.05). Figure 2 plots this interaction effect for easier interpretation: given their confidence regions, the perceived realism of the AP and Xinhua stories appear to be statistically different over the 1 to 3 range of the “general trust in U.S. media” scale. Specifically, when respondents had low trust in the U.S. media (GTUM = 1), the story was perceived to represent the reality of the U.S.–China trade war less accurately when reported by the AP (M = 3.65) than when reported by Xinhua (M = 4.02). This difference minimized as GTUM increased. In addition, for respondents who had a general high trust in U.S. media (GTUM >= 5), the stories reported by the AP exhibited a higher perceived realism than those reported by Xinhua, though these differences were only marginally significant. H3b was thus supported.

Next, we tested H3c, and the results revealed a statistically significant coefficient estimate on the “News Wire X general media literacy” interaction term (b = −0.43, p < 0.05 in Model 6, Table 3). As plotted in Figure 3, for respondents with low to moderate media literacy (0 <= logged media literacy <= 1.5), false stories reported by the AP were perceived to better represent the reality of the U.S.–China trade war than those reported by Xinhua News Agency; however, these contrasts were statistically insignificant, as suggested by their confidence regions. In contrast, for participants with a high media literacy (logged media literacy >= 1.6), false stories reported by the AP were perceived to represent the reality of the U.S.–China trade war less accurately than those reported by Xinhua News Agency, and these differences were statistically significant. These findings indicate that the impact of news wire on the perceived realism of false news is not the same across all participants; instead, this relationship is conditioned by their general media literacy, which thus provides support for H3c.

4. Discussion

This study tested how source cues and visual cues affect readers’ perceptions about the realism of false news that arises amid international disputes such as the U.S.–China trade war. Though this study focuses only on Chinese readers, its findings corroborate the idea that the news sources that people are firmly affiliated with and trust can affect their attitude toward false information. Chinese readers perceived a false story to better represent the reality of the U.S.–China trade war if it was published by Xinhua News Agency, which can be attributed to their frequent use of domestic Chinese media. On the other hand, as Chinese readers have fewer opportunities to use Western media and may question their credibility, they tend to rate the false stories published by the AP as less representative of reality. This is in line with the speculation that familiar news organizations are considered credible, whereas newer sources are seen as biased and untrustworthy (Flintham et al. 2018).

This relationship between source cues and perceived realism of false stories was conditional on respondents’ general trust in U.S. media, such that respondents who have a general high trust in U.S. media rated the stories reported by the AP to be better representative of the trade war reality than those reported by Xinhua. This again provides support for the idea that source cues affect readers’ perceptions about news through their trust in the news outlet (Metzger et al. 2010). The AP is a storied news organization with a long history and a global reputation for publishing balanced news, which might have contributed to the audience’s perceptions of its credibility. In addition, the AP’s membership-based structure of revenue generation gives the impression of non-biased coverage. As a result, foreign Chinese readers who are aware of the AP’s merits may tend to consider it as more credible than their peers who lack such knowledge.

The effect of source cues on readers’ perceptions about the level of realism in false information is also conditional on media literacy. Specifically, variation in source cues—with the story to be associated with either the AP or Xinhua News Agency—yielded insignificant impact on readers with low to moderate levels of media literacy. But those with a greater understanding of the process of news-making tended to perceive stories by the AP to be less representative of the U.S.–China trade war reality than those reported by Xinhua News Agency. Thus, Chinese readers’ confidence in disentangling the complex news-making process seems to have lessened their trust in Western media but strengthened their affinity to Chinese media. But without further investigation by splitting media literacy into two distinct strands—namely, self-reported capacity in understanding Chinese media’s news production or U.S. media’s news production—we are unable to look much into the possible mechanisms.

Our examination of whether the news topic moderates the effect of the news source on the perceived realism of a fake story produced partial support. While the stories on “MIT admission” and “pork order cancellation” did not affect readers’ perceptions about the realism of these stories, the story on “the pandas being recalled from the zoo in California” was successful in affecting readers’ perceptions. A possible explanation for this finding is that given the news value of oddity, this story is likely to have been extensively covered by the Chinese media, and it is possible that the respondents were already familiar with this issue but were unable to recall the reasons behind why the pandas were recalled. This finding shows a need for more research in the area. Additional experiments with a range of topics having different news values could cast more light on which topics affect perceived realism and why.

Interestingly, we did not find any impact of images on the perceived realism of a story. In our study, respondents were more likely to rely on cues such as the news source and the topic of the news to evaluate a false story’s perceived realism. As such, our findings echo the existing research (for example, Shen et al. 2019) that finds audiences rely more on their digital skills, media literacy, and prior knowledge to judge perceived realism. This finding is encouraging because it indicates that when images are used in conjunction with text, audiences do not rely on visual cues to determine the realism of the content, despite these cues being easier to process. Rather, they look for other cues, such as the information source, to determine the credibility of the content.

While this present study assessed people’s media literacy through self-reported measures, future work may consider pre-test techniques, such as asking participants to complete given tasks (i.e., searching information, cross-checking sources, etc.) to evaluate media literacy. Future research may explore the potential influence of fake news stories on Western users and compare it to the reactions of Chinese readers. Due to the escalating tensions between the two super-powers and many differences in their media systems, users from different countries are likely to treat news differently. Therefore, it would be enlightening to see whether the source cue effect varies by readers’ perceived relevance to international issues, their feelings toward the other nation, and so on. Future studies may also test the effect of visuals on perceived realism in more nuanced ways. For example, do certain types of visuals inspire more credibility than others? Are static data visualizations and interactive charts more compelling and realistic than images? How do combinations of these visualizations with text affect perceptions of realism? These questions are worthy of further exploration.