Trump, Twitter, and Truth Social: how Trump used both mainstream and alt-tech social media to drive news media attention

ABSTRACT

Former President Donald Trump is well-known for dominating the attention-driven hybrid media system through his controversial tweets, which spurred social media user engagement and news media attention. The removal of Trump from Twitter/X raises questions about his continued ability to drive news attention through his alt-tech platform, Truth Social. Our analysis shows that the retweets and “retruths” Trump gained predicted related news attention across the political spectrum. However, our results also point to shifts in news attention, including diminished direct embedding of his “truths,” a lower proportion of stories on his Truth Social activity among all stories about him, and greater partisan media attention to his Truth Social activity in 2022, as compared to 2016. These findings advance our understanding of Trump’s social media based communicative power, changing journalistic practices, and the place of alt-tech in the media system.

KEYWORDS:

• Social media
• alt-tech
• social media engagement
• news attention
• hybrid media
• attention economy

A bombastic and media-savvy populist politician, former U.S. President Donald Trump combined his aggressive anti-establishment and for-the-people rhetoric with the technological affordances of Twitter (rebranded as X in 2023) to gain political power (Ott & Dickinson, 2019; Pain & Masullo Chen, 2019). During the 2016 U.S. presidential election cycle, Trump earned substantial free media coverage (Francia, 2018) via his adept and tactical use of Twitter (Clarke, Grieve, & Danforth, 2019). His tweets’ retweets predicted more stories about him by media outlets across the political spectrum (Wells et al., 2016, 2020). While in office, he used Twitter to speak about and to various groups, attack opponents, and broadcast policies (Coe & Griffin, 2020; Lazarus & Thornton, 2021). His ability to transform social media buzz into news media spotlight hinges on a combination of media system dynamics, including the interlink between social and news media and the need to attract audience attention (Chadwick, 2017; Wells et al., 2020).

Trump’s presence on Twitter ended on January 8, 2021, when Twitter suspended him for inciting violence at the U.S. Capitol on January 6, 2021.¹ After a year-long social media hiatus, he resurfaced on Truth Social, a platform he founded. His supporters and a variety of right-wing extremists flocked to the platform. Truth Social, like other niche platforms like Gab, Parler, and Rumble, is an “alternative technology” (alt-tech) platform. Alt-tech platforms brand themselves as the bastion of free speech and the safe haven for free thinkers (Kor-Sins, 2021). Such sites have become popular among right-wing users who believe their voices are censored by mainstream platforms’ content moderation policies (Jasser, McSwiney, Pertwee, & Zannettou, 2021), such as deplatforming (i.e., suspending or removing) actors who engage in behaviors violating platform policies (Rogers, 2020).

In the evolving media system, can Trump wield Truth Social to obtain comparable levels of news attention as he could with Twitter? Focusing on two periods when Trump was not the President and when there was heightened campaign activity, we compare his use of Twitter to drive news attention during the 2016 presidential primary election with his use of Truth Social to do so during the 2022 midterm election. We collected news media, social media, and event data for each time period. We then applied time series modeling to investigate the predictive power of user engagement with Trump’s Twitter and Truth Social posts in shaping news attention toward him, controlling for major offline events. Our results underscore Trump’s ability to attract news attention via both Twitter and Truth Social, while also revealing shifts in how media outlets responded to his social media buzz in 2022 vs 2016.

This inquiry makes two theoretical contributions. First, it expands on research into Trump’s use of social media to influence media attention (e.g., Lewandowsky, Jetter, & Ecker, 2020, Wells et al., 2016; Wells et al., 2020), exploring the extent to which he has been able to use Truth Social in ways similar to Twitter. Second, it advances our understanding of alt-tech platforms in the attention-driven hybrid media system and how journalists respond to the evolving social media landscape. In what follows, we first discuss features of the media system and political behaviors and media practices within it, followed by how news media responds to politicians’ use of mainstream and alternative platforms.

Literature review

Hybrid media and attention economy

Politicians’ communication practices are shaped by the media system they inhabit (Ferree, Gamson, Rucht, & Gerhards, 2002; Hilgartner & Bosk, 1988). A media system can be defined as a fluid and evolving assemblage of content creators, audiences, institutions, and practices (Jungherr, Rivero, & Gayo-Avello, 2020). One characteristic of the contemporary media system is its hybridity, evidenced by the coexistence, competition, and cooperation between older and newer media (Chadwick, 2017). The interrelation between the older legacy news media and the newer social media challenges politicians to campaign on both fronts. The “ubiquitous presidency” thesis contends that presidential communication must be “access(ibble), personal, and pluralistic” to target diverse audiences across various media platforms (Scacco & Coe, 2016). Such interconnectedness also provides fresh opportunities for politicians: they can use social media to set news media agendas (Conway, Kenski, & Wang, 2015) and share news articles selectively on social media to promote their claims (Heidenreich et al., 2022).

Another feature of the media system is the value of attention, a corollary of the proliferation of media outlets and the explosion of information (Zhang, Wells, Wang, & Rohe, 2018). Couldry (2012) uses the metaphor of supersaturation, from chemistry, to describe such “density of media in contemporary societies” (p. 6). In this environment, quality supersedes quantity such that “any attention may be positive attention” (Wells et al., 2020, p. 665). Accordingly, we define “news attention” as the number of stories about that a candidate is given in news reporting, focusing on the quantitative aspect of news attention in the attention economy. There is no better case to underscore the importance of attention than the 2016 U.S. presidential election. Trump drew “$2 Billion Worth of Free Media” (Francia, 2018), likely because journalists saw him as an attention magnet and his frequently outrageous rhetorical style as a reliable way to attract eyeballs.² The media system’s hybridity and the pursuit of attention are two sides of the same coin. Certain content, like moral and emotional posts that disparage out-groups, is successful at stimulating social media engagement (Brady et al., 2020; Rathje, Van Bavel, & Van Der Linden, 2021), further spurring news media attention (Wells et al., 2020).

In the attention-driven hybrid media system, staging identity performances and generating social media buzz to attract news attention has been leveraged and proven effective by populist politicians (Dai & Kustov, 2022; Ernst et al., 2019). Research shows that Trump’s skillful use of Twitter allowed him to spark news attention when it was waning (Wells et al., 2020) and counterbalance potentially harmful news exposure by tweeting non-threatening topics (Lewandowsky, Jetter, & Ecker, 2020).

When amplifying politicians who engage in strategic performance on social media, different media outlets likely exhibit different tendencies due to the normative, cultural, and structural differences between the left and right. Right-wing media outlets, especially far-right media, position themselves as a counterforce to what they see as a hegemonic, threatening, left-leaning mainstream media (Dowling, Johnson, & Ekdale, 2022; Nadler, 2022), potentially resulting in greater responsiveness to social media metrics (Mukerjee, Yang, & Peng, 2023). However, left-wing outlets are more closely “tied to mainstream core journalistic institutions in ways the far-right newcomers do not” (Wells et al., 2020, p. 665). Rallying behind a core conservative identity, right-wing outlets are more internally coherent and cloistered than their left-wing counterparts (Benkler, Faris, & Roberts, 2018; Faris et al., 2017). Resultingly, right-wing outlets have amplified conservative Russian IRA disinformation Twitter handles more than left-wing accounts (Zhang et al., 2021).

Mainstream vs alt-tech social media

Advancing the above research on political and media behaviors in the attention-driven hybrid media system, we focus on how changes in the social media landscape might impact politicians’ communicative power and journalistic practices. Existing research has documented how politicians leverage mainstream platforms, especially Twitter, to drive news attention (e.g., Gilardi, Gessler, Kubli, & Müller, 2022, Wells et al., 2016; Wells et al., 2020). However, the recent rise of right-wing alt-tech platforms (Kor-Sins, 2021; Zannettou et al., 2018) and Parler (Aliapoulios et al., 2021) begs the question of whether they can serve as vehicles for drawing media spotlight, particularly as conservative media personalities have bemoaned mainstream platforms moderation policies. Proposed as an alternative to mainstream social media, alt-tech platforms have garnered a niche, loyal user base among right-wing groups (Aliapoulios et al., 2021; Dehghan & Nagappa, 2022; Zannettou et al., 2018).³ The chief difference between alt-tech and their more popular counterparts is in their content moderation policies. While platforms generally want to be viewed as neutral spaces for content production, they must inevitably moderate content; the moderating decisions of a platform defines its character and, in part, its audience and business model (Gillespie, 2018). Alt-tech platforms, which promote themselves as defenders of “free speech” through minimal content moderation (Jasser, McSwiney, Pertwee, & Zannettou, 2021), impose a set of ideologically driven moderation practices, thereby inviting violent rhetoric, harassment, and harmful misinformation.

The moderation of right-wing users on mainstream platforms further contributes to the popularity of alt-tech, as content removal or outright suspension of users can encourage them to move or “migrate” to ideologically-aligned alt-tech platforms with fewer or different content moderation policies. Migration as an outcome of suspension not only increases the alt-tech user base (JasserMcSwiney, Pertwee, & Zannettou, 2021), but the discourse on these platforms also becomes more politicized and radicalized (Dehghan & Nagappa, 2022).

Despite the common belief that alt-tech platforms operate in the fringes of the internet to avoid additional moderation or scrutiny (Van Dijck, de Winkel, & Schäfer, 2021), we argue that alt-tech platforms might be well integrated into the broader media system by their amplification of news media and their potential to trigger news attention. First, mainstream news (Peucker & Fisher, 2023) and content from mainstream platforms (Widjaya & Smith, 2023) are circulated on alt-tech platforms. Also, alt-right platforms are amplification machines of far-right news. For example, most of the commonly shared domains on Gab are far-right news sites, unlike Twitter (Zhou, Dredze, Broniatowski, & Adler, 2019). Some of the most linked-to content on BitChute, an alt-tech video-sharing site, also comes from far-right news sources like Infowars and x22report (Childs et al., 2022).

Furthermore, right-wing activists and extremists have a longstanding interest in garnering news attention (Baugut & Neumann, 2019; Freelon, Marwick, & Kreiss, 2020). Politicians and pundits suspended from mainstream platforms, such as Marjorie Taylor Greene and Alex Jones, bemoan the censorship of “big tech” but seek out news attention through other mechanisms.⁴ Their claims can be readily amplified by right-wing media, and they put out controversial claims to provoke reaction from mainstream media (Freelon, Marwick, & Kreiss, 2020).

As such, alt-tech platforms are not on the fringe; rather, they might play a non-negligible role in the media system that, paradoxically, their users criticize. We focus on one critical aspect of this role: news coverage quoting or referencing alt-tech content. As alt-tech content, like that from LiveLeak and Steemi, can spread to mainstream platforms like Twitter (Wilson & Starbird, 2021), it can also reach newsrooms and make its way to news stories. Journalists themselves have suggested that people are likely to consume Truth Social content indirectly through news coverage.⁵

Trump, Twitter & Truth Social

The desire for alt-tech to enter mainstream media discourse may be especially apparent with Truth Social, a platform conceived in the wake of Trump’s Twitter suspension. Prior to Truth Social, Trump’s activity on Twitter could be characterized by its negativity, willingness to attack his opponents (Ross & Caldwell, 2020), and his use of ethnocentric and nationalist populist rhetoric (Ott & Dickinson, 2019; Pain & Masullo Chen, 2019). Such strategic performance often generates feverish Twitter reaction (Lazarus & Thornton, 2021; Zhang, Wells, Wang, & Rohe, 2018), which could signal newsworthiness (Lischka, 2021; Walters, 2021).

Following the Capitol attack on January 6, 2021, Trump was suspended from Twitter and Facebook. He was not alone – Twitter removed over 70,000 QAnon-linked accounts, and Meta removed content using the term “stop the steal” on Facebook and Instagram. Trump tried various ways to communicate with his audience, including his page “From the Desk of Donald Trump” on his website.⁶ These efforts culminated in the creation of a new platform, which Trump announced on October 21, 2021.⁷ Four months later, Truth Social launched on Apple.⁸ Since its inception, the site had 1.7 million unique visitors from the U.S. in September 2022. As of early 2024, Trump’s account has more than six million followers.⁹

To understand Trump’s communicative power in the changing social media landscape and larger media system, we explore his ability to garner news attention through Truth Social in 2022 vis-à-vis Twitter in 2016. The first one spans the 2016 primary election season, between June 16, 2015, when Trump announced his bid for the presidency, and July 19, 2016, when he officially became the nominee of the Republican Party. The second timeframe covers the overlap between Trump’s Truth Social presence and the 2022 midterm election cycle, from May 6 (though Trump started posting to Truth Social regularly from April 28, our news media data stream was not stable until May 6) to November 8, the 2022 election day.

Since “Trump enjoyed substantial news coverage regarding his campaign in general, and his use of Twitter specifically” in the 2016 election (Wells, Zhang, Lukito, & Pevehouse, 2020, p. 8), we distinguish between two types of news attention to Trump: a) news attention to Trump’s social media activity specifically and b) news attention to him in general, with the former being a subset of the latter. This choice is also driven by conceptual considerations. It allows us to assess news attention to Trump’s social media activity against some baseline (the proportion of stories on his social media activity in all stories about him); it also allows us to understand the boundary of the power of Trump’s social media activity (whether his social media activity can predict news stories about it only or general news attention to him as well). First investigating the amount of attention, we ask: Did news media pay more attention to a) Trump’s social media activity and b) him in general in the 2016 primary election season than they did in the 2022 midterm election cycle? (RQ1)

Next, we compare how user engagement with Trump’s posts on Twitter vs Truth Social relates to news attention. Existing research reveals that the number of retweets Trump received did predict news attention toward him in 2016, regardless of the partisan slant of the news organization (Wells et al., 2020), to the extent that Trump can even use Twitter to diffuse threatening news coverage (Lewandowsky, Jetter, & Ecker, 2020). Extending previous work, we propose the following hypothesis: user engagement with Trump’s tweets predicted attention to a) his Twitter activity and b) him in general across media outlets on the left, center-left, center, center-right, and right in the 2016 primary election season (H1). However, we have yet to know Trump’s ability to garner news attention through Truth Social. We thus ask the following question: Did user engagement with Trump’s “truths” predict news attention to a) his Truth Social activity and b) him in general across media outlets on the left, center-left, center, center-right, and right during the 2022 midterm election cycle? (RQ2)

Method

Data and measures

We collected events, social media, and news media data during the two periods. Major offline events during the first time period came from Wells and colleagues’ (2016) study, complemented by a primary election timeline from the National Conference of State Legislatures. Event variables include Trump’s press conferences, town halls, planned media appearances (debates and scheduled media appearances), unplanned media appearances (call-ins to news programs), miscellaneous events (e.g., restaurant/store drop-ins, get-out-the-vote non-rally events), and the primary elections.

Major offline event variables during the second time period include Trump’s speeches (collected from c-span.org), his statements (released on his website), candidate endorsements (released on his website), primary elections, and the January 6 committee hearings.

Trump’s tweets were collected from a GitHub archive.¹⁰ Trump’s Truth Social posts were collected using its API. Four variables were generated: Trump’s daily number of tweets, the daily number of total retweets he received on Twitter, his daily number of “truths” (the Truth Social equivalent of tweets), and the daily number of total “retruths” (the Truth Social equivalent to retweets). The level of user engagement with social media posts is measured by the number of retweets and “retruths” on Twitter and Truth Social, respectively.

News media data for both time periods were collected via Media Cloud, an open-source data corpus and platform for studying web-based news media and media ecosystems (Roberts et al., 2021). For this project, we used the most current set of Media Cloud collections used to study political partisanship, produced by Faris et al. (2020) to study public discourse around the 2020 U.S. presidential election.¹¹ In their study, they included any link tweeted by the studied accounts. Two graduate students manually coded the 13,656 domains from this study to identify media outlets, resulting in a list of 5,359 classified into left, center-left, center, center-right, and right categories. We narrowed the list down to 441 outlets with stable data streams during the two time periods (see Appendix A), covering all major outlets in the U.S.

We used this media list to pull media data in six queries. To collect news attention to Trump’s social media activity, we applied two stringent queries, intended to capture journalists’ direct embedding of Trump’s posts (Oschatz, Stier, & Maier, 2022) and referred to as “realdonaldtrump Twitter” (“realdonaldtrump” AND “twitter” for the 2016 period) and “realdonaldtrump Truth Social” (“realdonaldtrump” AND (“truth social” OR “truthsocial”) for the 2022 period).

We also applied two broad queries for news attention to Trump’s social media activity. The first, for the 2016 period, searched for news stories about Trump and Twitter in which Trump’s Twitter username (“realdonaldtrump”) was used or in which “trump” was used within 10 words of either “twitter” or “tweet.” The second query, for the 2022 period, searched for news stories in which Trump’s Truth Social username (“realdonaldtrump”) was used or in which “trump” was used within 10 words of “truth” (Appendix A documents the queries). These two queries are referred to as “Trump Twitter” and “Trump Truth Social” respectively.¹²

To collect general news attention to Trump during both periods, we used the query “trump trump”~1000 to return stories in which Trump’s name is mentioned at least twice in the span of a thousand words, ensuring that collected articles did not mention Trump in passing (see Wells et al., 2016, Wells et al., 2020). We refer to these two queries as “Trump 2016” and “Trump 2022.” To remove false positives, we filter out titles with only the name of Trump’s family (Donald Trump, Jr., Ivanka Trump, Eric Trump, Ivana Trump, Barron Trump, Tiffany Trump, Fred Trump, and Melania Trump) in all six queries.

For both time periods, we computed the daily number of stories from outlets on the left, center-left, center, center-right, and right based on the six queries. We applied ANOVA to compare differences in the number of stories between media types. Appendix B shows the visualizations of the event, social media, and news media variables.

Time series modeling

Given the temporal nature of our variables, we conducted a set of time series analyses to investigate the relationship between user engagement with Trump on social media and news coverage about him, while controlling for major offline events. Specifically, we chose Prais-Winsten estimation (Park & Mitchell, 1980, see application in Wells et al., 2016 & Wells et al., 2020). This approach allows the investigation of how changes in the predictors from t-1 to t are correlated with change in the outcome from t-1 to t. Given fast news cycles and journalists’ close attention to Trump, this cotemporaneous model specification is appropriate and consistent with prior research (Wells et al., 2020). For the 2016 primary election period, we ran 15 Prais-Winsten time series regressions for three types of news attention across five media spheres: 1) “realdonaltrump Twitter,” 2) “Trump Twitter,” and 3) “Trump 2016” from news outlets on the a) left, b) center-left, c) center, d) center-right, and e) right. In each model, we included as predictors the total number of retweets that Trump’s tweets received (divided by 10,000 for readability of the coefficient, a measure that does not affect the significance level), the total number of his tweets, and the 2016 event variables. Appendix C1 shows that the predictors are not highly correlated, eliminating concern over multicollinearity.

For the 2022 midterm election period, we conducted 10 Prais-Winsten time series regressions, focusing on (1) “Trump Truth Social” and (2) “Trump 2022” across news outlets on the a) left, b) center-left, c) center, d) center-right, and e) right. In this model, we included as predictors the total number of “retruths” Trump received (divided by 10,000 for readability of the coefficient), the total number of his Truth Social posts, and the 2022 event variables. Appendix C2 shows that the predictors are not highly correlated, eliminating concern over multicollinearity. We did not run models with the number of news stories from the “realdonaltrump Truth Social” query because of sparsity: there were only 81 stories from all the media outlets queried.

Results

To answer RQ1, we provide a quantitative description of news attention (Table 1). During the 2016 election period, the stringent “realdonaldtrump Twitter” query (likely stories directly embedding his tweets) returned 4.93 stories per day and 1,971 in total from all media outlets. The comparable “realdonaldtrump Truth Social” query for the 2022 period yielded just 0.43 stories per day and 81 in total from all outlets. However, despite significantly fewer stories in 2022 based on the stringent measure, the average number of stories mentioning Trump’s Truth Social activity according to the broad measure (27.53 per day and 5,184 in total; the “Trump Truth Social” query) is more than double the average number of stories referencing Trump’s Twitter activity broadly (13.58 per day and 5,433 in total; the “Trump Twitter” query). The imbalance is even greater when it comes to general attention to Trump measured by the “Trump 2016” and “Trump 2022” queries. There were 599.67 stories per day (112,138 in total) mentioning Trump in general in the 2022 period, but only 222.50 stories per day (89,000 in total) in the 2016 period. But when we consider all stories about Trump, proportionally less coverage about Trump was about his Truth Social activity in 2022 (4.6%, 5,184/112,138) than attention to his Twitter activity during the 2016 election period (6.1%, 5,433/89,000).

Table 1. Summary of news attention variables.

	media_type	num_outlets	total_stories	total_stories_ per_day	average_stories per_outlet_ per_day
Timeframe: 6/16/2015–7/19/2016 (400 days)
realdonaldtrump Twitter	left	33	203	0.51	0.02
realdonaldtrump Twitter	center_left	83	712	1.78	0.02
realdonaldtrump Twitter	center	91	599	1.5	0.02
realdonaldtrump Twitter	center_right	46	209	0.52	0.01
realdonaldtrump Twitter	right	17	248	0.62	0.04
Trump Twitter	left	47	391	0.98	0.02
Trump Twitter	center_left	96	1664	4.16	0.04
Trump Twitter	center	111	2166	5.42	0.05
Trump Twitter	center_right	63	971	2.43	0.04
Trump Twitter	right	20	241	0.6	0.03
Trump 2016	left	79	7147	17.87	0.23
Trump 2016	center_left	126	28362	70.9	0.56
Trump 2016	center	130	33745	84.36	0.65
Trump 2016	center_right	75	15396	38.49	0.51
Trump 2016	right	31	4350	10.88	0.02
Timeframe: 5/6/2022–11/8/2022 (187 days)
realdonaldtrump Truth Social	left	5	9	0.05	0.01
realdonaldtrump Truth Social	center_left	20	27	0.14	0.01
realdonaldtrump Truth Social	center	13	20	0.11	0.01
realdonaldtrump Truth Social	center_right	8	15	0.08	0.01
realdonaldtrump Truth Social	right	3	10	0.05	0.02
Trump Truth Social	left	41	1086	5.81	0.14
Trump Truth Social	center_left	91	1332	7.12	0.08
Trump Truth Social	center	105	1336	7.14	0.07
Trump Truth Social	center_right	62	847	4.53	0.07
Trump Truth Social	right	22	583	3.12	0.14
Trump 2022	left	79	20483	109.53	1.39
Trump 2022	center_left	126	31481	168.35	1.34
Trump 2022	center	129	31588	168.92	1.31
Trump 2022	center_right	75	17643	94.35	1.26
Trump 2022	right	31	10943	58.52	1.89

Attention from different types of media outlets varied. We found a statistically significant difference in the number of stories in 2016, on Trump’s Twitter activity (both the stringent, F(4, 1995) = 37.14, p < .001, and broad measures, F(4, 1995) = 22.29, p < .001) and on Trump in general (F(4, 1995) = 105.6, p < .001), produced per day per outlet across different media types. Right-wing media outlets produced an average of .04 stories on Trump’s Twitter activity (stringent measure) per day per outlet, compared to other media types (.01–.02 stories per day per outlet). But when referencing Trump’s Twitter activity (broad measure) and Trump in general, center and leaning outlets tended to produce more news articles than partisan outlets.

However, partisan media outlets in 2022, on the left (M = .14) and right (M = .14), paid more attention to Trump’s Truth Social activity (broad measure), on average, than center (M = .07), center-right (M = .07), and center-left outlets (M = .08), F(4, 930) = 16.58, p < .001. Additionally, there was a statistically significant difference in the number of stories about Trump in general produced per day per outlet across different media types (F(4, 930) = 25.31, p < .001), with outlets on the right producing an average of 1.89 stories per day per outlet, which was larger than the amount from any other media types (1.26–1.39 stories per day per outlet) (Appendix D presents a full table with pairwise comparisons).

Time series modeling results for Twitter during the 2016 presidential election provide partial support for H1 (Table 2). The number of retweets is positively associated with the stringent measure of news attention to Trump’s Twitter activity across the spectrum (p < .001 across the board). Unplanned media events also positively predict such attention from center-right outlets (b = 0.15, p < .05) and primary elections negatively predict such attention from center-left outlets (b = −0.88, p < .05). Similarly, the number of retweets of Trump is positively correlated with news attention to his Twitter activity measured by the broad query from all five corners of the media system (p < .001 across the board). But the number of tweets posted by Trump negatively predicts such attention from media outlets on the left (b = −0.02, p < .01) and center (b = −0.08, p < .05). None of the event variables are associated with such news attention. In terms of general news attention to Trump, retweets of Trump’s tweets are only positively associated with such attention from outlets on the left (b = 0.56, p < .01) and right (b = 0.43, p < .001). The number of tweets posted by Trump negatively predicts such attention from media on the right (b = −0.08, p < .01). All predictors can barely explain the variance in center and center-left/right news attention to Trump in general. A comparison of these 15 full models and models with only retweets shows that the variance is mainly explained by retweets (Appendix E1).

Table 2. Summary of the 15 prais-winsten regression models for twitter during the 2016 primary election season.

	News attention to realdonaldtrump and Twitter (stringent measure)					News attention to Trump and Twitter (broad measure)					General news attention to Trump
	Left	Center-Left	Center	Center-Right	Right	Left	Center-Left	Center	Center-Right	Right	Left	Center-Left	Center	Center-Right	Right
Retweets of Trump	0.06 ***	0.19 ***	0.22 ***	0.06 ***	0.08 ***	0.16 ***	0.38 ***	0.67 ***	0.26 ***	0.08 ***	0.56 **	0.70	1.50	0.12	0.43 ***
Trump Tweets	0.00	−0.01	−0.01	−0.01	−0.01	−0.02 **	−0.02	−0.08 *	−0.02	−0.01	−0.03	−0.03	−0.22	0.00	−0.08 **
Press Conferences	0.12	0.33	−0.02	−0.13	0.01	0.13	0.08	−0.94	−0.08	0.07	2.42	3.24	0.06	1.95	0.03
Town Halls	0.00	0.20	−0.03	−0.13	0.10	0.10	−0.03	−0.76	0.72	0.09	1.85	5.90	9.79	3.33	0.61
Planned Media Events	−0.01	−0.03	−0.12	−0.02	−0.04	−0.05	−0.26	−0.27	−0.15	−0.07	0.84	2.57	2.61	1.11	0.09
Unplanned Media Events	0.04	0.03	0.19	0.15 *	−0.01	0.08	0.14	0.72	0.15	−0.03	1.03	1.70	2.73	1.80	0.38
Miscellaneous Events	0.01	−0.28	−0.39	0.22	−0.12	0.17	−0.36	−1.13	0.24	−0.04	2.24	5.93	5.01	2.88	1.01
Primary Election	−0.03	−0.88 *	−0.50	−0.12	−0.27	−0.22	0.42	−2.22	0.47	−0.19	2.24	12.79 *	17.51 *	3.14	0.76
Constant	0.19 *	1.11 ***	0.86 ***	0.38 ***	0.44 ***	0.59 ***	3.01 ***	4.21 ***	1.67 ***	0.43 ***	14.98 ***	66.18 ***	78.85 ***	37.19 ***	10.07 ***
Adjusted R-Squared	0.04	0.07	0.08	0.03	0.06	0.11	0.09	0.07	0.07	0.05	0.05	0.02	0.02	0.00	0.03
F-statistic	3.12 **	4.89 ***	5.11 ***	2.77 **	4.07 ***	7.46 ***	5.69 ***	4.71 ***	5.01 ***	3.90 ***	3.37 ***	2.13 *	1.77	1.08	2.61 **

* < 0.05, **< 0.01, ***< 0.001.

For RQ2, the results for Truth Social during the 2022 midterm season paint a similar picture (Table 3). The number of Trump’s retruths significantly and positively predicts attention to his Truth Social activity (broad measure) and general attention to Trump from media across the political spectrum. General news attention across the spectrum is also positively predicted by Trump’s public statements, the Jan 6 hearings, and midterm elections. A comparison of these 10 full models and models with only retruths indicates that the variance in the full models for attention to Trump’s Truth Social activity (broad measure) is mainly explained by the number of retruths, but not for attention to Trump in general (Appendix E2).

Table 3. Summary of the 10 prais-winsten regression models for truth social during the 2020 midterm election season.

	News attention to Trump and Truth Social (broad measure)					General news attention to Trump
	Left	Center-Left	Center	Center-Right	Right	Left	Center-Left	Center	Center-Right	Right-
Retruths of Trump	0.75 ***	1.13 ***	0.79 **	0.92 ***	0.56 ***	2.37 *	5.76 **	4.25 *	3.43 **	1.69 **
Truth Social Posts	−0.21 **	−0.36 **	−0.21	−0.28 ***	−0.22 ***	−0.63	−1.90 *	−1.59	−1.34 *	−0.52
Speeches	−0.26	−0.32	−1.12	−0.77	−0.07	−4.80	−6.94	−5.77	4.55	−4.90
Public Statements	0.54	0.19	0.35	0.65	0.35	9.94 **	16.72 **	14.66 *	11.78 **	4.49 *
Endorsements	−0.02	0.26	0.19	0.02	0.08	−0.82	−0.08	−0.84	−0.07	−0.31
Jan 6 Hearings	0.67	0.88	0.92	−1.58	0.05	47.92 ***	107.45 ***	88.29 ***	31.68 **	13.45 *
Midterm Elections	−0.43	0.08	1.08	−0.15	−0.76	17.04 *	54.66 ***	41.15 **	19.58 *	9.79 *
Constant	2.23 *	2.26	3.18 *	0.52	1.21 *	91.84 ***	130.41 ***	143.90 ***	75.44 ***	49.14 ***
Adjusted R-Squared	0.14	0.11	0.04	0.15	0.19	0.19	0.28	0.20	0.14	0.14
F-statistic	5.30 ***	4.26 ***	2.05 + (p = .051)	5.58 ***	7.21 ***	7.07 ***	11.36 ***	7.74 ***	5.41 ***	5.20 ***

*< 0.05, **< 0.01, ***< 0.001.

Since Prais-Winsten models are concurrent, we conducted a supplementary analysis that considers a lag of one (i.e., moving the independent variables backward by one day) to entertain the assumption that news attention takes longer to materialize. The patterns largely hold for Trump’s retweets in the models with a lag of one: a change in retweets of Trump at t-1 is positively associated with a change in news attention to his Twitter activity at t measured by the broad and stringent queries across the media system (Appendix F1). While external events in the contemporaneous 2016 models are not highly associated with news attention, it is more so in the lagged 2016 models. This observation suggests that news attention to the engagement level of Trump’s posts is more immediate than that to related events. However, “retruths” of Trump at t-1 are positively associated with attention to his Truth Social activity (the broad query) at t only from center-left and center-right outlets (Appendix F2), suggesting that news attention is only related to the engagement level of Trump’s Truth Social posts and external events in a contemporaneous way and dissipates quickly.

Discussion

In this article, we ask whether Trump was able to use Truth Social, an alt-tech platform, to spark news media attention during the 2022 midterm election cycle as he could with Twitter, a mainstream platform during the 2016 primary election season. The results suggest that news media covered Trump’s Truth Social activity with greater fervor than they did his Twitter activity, a pattern mostly driven by partisan media on the left and right. Furthermore, media outlets across the political spectrum invariably responded to user engagement (retweets and “retruths”) that Trump stirred up on Twitter in 2016 and on Truth Social in 2022. The more he stimulated Twitter or Truth Social user engagement, the more stories news outlets across the media system ran about his Twitter or Truth Social activity. In addition, though the number of retweets was associated with general media attention to Trump only from news outlets on the left and right in 2016, the number of “retruths” predicted general media attention to Trump from the entire media system in 2022.

These findings about Trump’s use of Twitter echo previous research (Wells et al., 2020) while making important methodological advancements. First, our measures of news attention to Trump are more nuanced and precise, ranging from general news attention to him and specific attention to his Twitter or Truth Social activity, which is further measured by broad and stringent queries. In the 2020a study by Wells et al., media coverage of Trump was quantified as the number of stories mentioning Trump twice, without excluding stories focusing on other Trump family members. Second, we were able to substantially expand the number of news media outlets across the political spectrum by utilizing meticulous human coding and validation of the MediaCloud collection. These methodological updates allow us to confirm “surprising uniformity” in the media ecosystem’s response to Trump’s ability to generate engagement on Twitter (Wells et al., 2020, p. 676). Our study also demonstrates that when we introduce an alt-tech platform, such consistency is more reflected in media’s attention to his social media activity, which is a subset of media’s general attention to Trump.

Thus, whenever Trump succeeded in generating feverish support or incandescent rage on Twitter and whenever his “truths” resonated widely with users on Truth Social, journalists might be compelled to write stories about it or mention it when covering other aspects of Trump. Given the pressure to write clickable stories in this attention economy – aptly characterized as an “epic scramble to get inside our heads” (Wu, 2017) – it is not surprising that journalists might gravitate to either the politics-heavy mainstream platform Twitter or the new alt-tech platform Truth Social. Our finding that Trump’s Truth Social activity was significantly related to news media coverage points to broader implications for journalism practice: while journalists have to attend to alt-tech platforms because of their watchdog obligation, this reporting can amplify alt-tech platforms and their users.

Our study reveals that Trump’s ability to attract news attention via social media is not predicated on a specific platform, but on an ability to engage social media users generally. Even after his removal from Twitter, Trump continues to pique journalistic interest through an alt-tech platform (one which he founded) by harvesting attention from his loyal user base. As a recently launched platform, Truth Social has limited exposure to the public, with only 27% of Americans having heard of it (Stocking et al., 2022). Despite its comparatively small size and audience number, news outlets across the spectrum still track it, responding to conservative and far-right users’ enthusiasm toward Trump’s “truths.” The fact that the positive relationship between Trump’s engagement on social media and media-system-wide attention to Trump’s social media activity holds for both Twitter and Truth Social speaks to the center of gravity that Trump retains in American politics via social media. The alt-tech label that Truth Social carries is simply outweighed by the “newsworthiness” of Trump’s statements on this platform. This finding explains why Trump, despite making virtually no profit from Truth Social,¹³ is compelled to maintain this site. Of course, talking about Truth Social in terms of profit misses its purpose. It is public-relations-as-platform: a means for Trump to generate news attention with no filter.

This finding further provides important insights into social media platforms and moderating policies. Critically, such policies are not the be-all and end-all solution to disturbing political communication problems. Content moderation decisions are ultimately compromises between users’ values and platforms’ profit goals (Gillespie, 2018). Thus, platforms often “err on the side of encouraging as many people to stay as possible, imposing rules with the least consequences, keeping troublesome users if they can, and bringing them back quickly if they can’t” (Gillespie, 2018, p. 201). Twitter’s decision to ban Trump and Elon Musk’s decision to reinstate Trump both might be both motivated by profit: Twitter moderated content because they knew that there are large segments of the population who do not want to be inundated by attacks, misinformation, and spam, and that advertisers do not want their brand images tainted; Musk’s decision is based on some combination of ideological reasons and a particular opinion about what’s best for profit. Additionally, even if platforms are genuine about content moderation, those efforts rely on volunteer labor or are shrouded in secrecy, without providing a system that empowers users (Gillespie, 2018).

Furthermore, while deplatforming presupposes that harmful speech can be contained by cutting off access to a large audience, consensus on its impact remains uncertain. For example, even if high-profile politicians are removed from mainstream platforms, they may be able to exert similar levels of influence using alt-tech platforms like Truth Social or Rumble. In those spaces, they might become more unrestrained in their speech and interaction, further helping them court attention in the media system. After all, in this attention-driven economy, whatever turns heads begets further attention. Therefore, alt-tech platforms are integrated into the media system, playing a non-trivial role in shaping information flows.

Given this, it is important to recognize that Trump’s success with alt-tech should be seen as a potential ceiling rather than a generalizable case. As a populist politician with significant wealth and countless controversies, Trump has capitalized on his political and economic power to adapt to the evolving social media landscape. Given that it is unrealistic for other politicians or users of mainstream platforms to have comparable resources or audiences to finance, build, and promote an alternative platform, our analysis highlights a best-case scenario.

Relative to Trump, it is uncertain how other politicians, pundits, or activists would fare in terms of attracting news attention via an alt-tech platform. Nevertheless, Trump’s actions, both offline and on social media, might provide a playbook for other candidates. To put it another way: though Trump is unique in terms of wealth, political clout, and prestige, the strategy of leveraging an alt-tech platform to advance one’s ideology is one that can be readily adopted by other politically active individuals.

However, our results also point to ways that an alt-tech platform may not garner as much news attention as a mainstream one. While Trump’s Twitter or Truth Social activity was significantly related to media coverage across all types of media outlets, the type of coverage between 2016 and 2020 was different. Though news media outlets directly and profusely embedded and amplified Trump’s tweets in 2016 (which can directly take readers to his Twitter page and facilitate engagement with his tweets), they no longer did so with the same level of intensity in 2022 (0.43 per day vs 4.93 per day, using the stringent measure). The same can be said of attention to Trump’s social media activity according to the broad measure (in the form of embedding his “truths” and describing what he said on Truth Social): news stories referencing Trump and Truth Social account for a smaller percentage (4.6%) of total news stories referencing Trump in 2022, compared to the percentage (6.1%) of news stories mentioning Trump and Twitter in all stories mentioning Trump in 2016, despite that his launching of Truth Social was newsworthy. These findings show that media coverage on Trump’s Truth Social activity in 2022 relatively declined in the broad context of news coverage on Trump, in comparison with news attention to his Twitter activity in 2016.

This change may indicate a shift in editorial choice: journalists and publishers may be more aware of the consequences of indiscriminately amplifying social media posts and thus more reluctant to embed posts directly. During the 2016 presidential election, by embedding tweets of Russian troll accounts (Lukito et al., 2020), media outlets directly or indirectly fueled the follower growth of those initially obscure accounts (Zhang et al., 2021). Given that indiscriminate media amplification has resulted in the mainstreaming of disinformation actors, populist politicians, and extremist groups alike, Donovan and Boyd (2021) call for journalists to practice strategic silence before they know how to strategically amplify content in a socially responsible way. Mainstream media outlets may have turned to exercise their amplification power more cautiously to guard against exploitation by potential bad actors (Abubakar, 2020).

Also, notably in 2022, media on the right still paid significantly greater attention on average to Trump than any other media outlets on the political spectrum. The insulated and fragmented communication ecology of the right, centering around influencers and amplifiers, provided a channel for Trump to easily insert his presence into right-wing media. Also, media on the left and right were significantly more likely to cover stories about Trump’s Truth Social activity than center and leaning media, showing partisan media’s uptake of the alt-tech platform, either cementing or refuting alt-tech ideologies. This contrasts with 2016, when center and leaning outlets tended to pay more attention to Trump and his Twitter activity in a broader sense. In the contemporary hybrid media ecology, partisan media, especially right-wing media, remain hooked to alt-tech platforms and populist figures.

Our study has a few limitations. One limitation with the stringent search queries is that we cannot determine how frequently media included images of tweets/”truths” without the embedding feature that both Twitter and Truth Social provide. It is likely that, although journalists are less inclined to directly embed Trump’s “truths,” they might be more willing to insert screenshots of these posts. Relatedly, Media Cloud’s database is limited to web-based media. Some outlets like Fox News have a large web presence, but do not reliably publish transcripts for their television programs, making their broadcast content potentially different from their online content.

Although our focus on the quantity of attention led us to investigate how Trump’s social media activity was associated with the number of related news stories in an expansive collection of media outlets, it is equally important to study specific media portrayals of his social media activity. For example, when Trump used Twitter or Truth Social to ridicule journalists and media outlets, how was such an attack on media portrayed in news stories? The impact of attention might not be fully understood without considering specific media framing. We also hope future research to expand upon our findings by examining relationship between alt-tech platforms and media in comparative contexts, such as investigating prominent politicians or groups in other countries as well as exploring the dynamics of activity across multiple alt-tech platforms. Lastly, we caution that the 2016 primary election and the 2022 midterm election may not be directly comparable, due to different election types and Trump’s different political roles (presidential candidate vs former president). Nevertheless, these two periods represent the times when Trump’s attention-seeking and mobilization-oriented social media activity was heightened before becoming a formal candidate for the party. It is worth noting that the focus of this analysis is to understand Trump’s ability to garner attention within each period and relative to the baseline, rather than to directly compare the absolute value of attention from each time period. We therefore accept the inherent limitations of this natural experiment and encourage a longitudinal observation for future scholarly inquiry.

Overall, by replicating and improving upon previous research, this study sheds light on the relationship between Trump’s social media strategy, the role of alt-tech platforms in the evolving social media landscape, and the response of news media. While Trump’s political power and financial resources will likely remain formidable forces shaping news attention in the evolving media system, we provide a deeper understanding of his use of social media to garner attention and journalistic practices in response to this political communication behavior, while offering a window into understanding similar practices from other politicians in the U.S. and around the globe. After all, benefiting from and challenged by a deeply interconnected media system, politicians, journalists, and platforms are all swept up in the mad scramble for attention.

Acknowledgement

We thank our reviewers, David Karpf for the feedback on our manuscript, and Chris Wells for sharing the 2016 event data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19331681.2024.2328156

Additional information

Notes on contributors

Yini Zhang

Yini Zhang (Ph.D. University of Wisconsin-Madison) is an assistant professor in the Department of Communication at the University at Buffalo, State University of New York. She studies social media, media system, and political communication, using computational methods.

Josephine Lukito

Josephine (“Jo”) Lukito (Ph.D. University of Wisconsin-Madison) is an Assistant Professor at the University of Texas at Austin’s School of Journalism and Media and Director of the Media and Democracy Data Cooperative. She is also a Senior Faculty Research Affiliate for the Center for Media Engagement.

Jiyoun Suk

Jiyoun Suk (Ph.D. University of Wisconsin-Madison) is an assistant professor in the Department of Communication at the University of Connecticut. She studies the role of networked communication in shaping social trust, activism, and polarization, using computational methods.

Ryan McGrady

Ryan McGrady (Ph.D. North Carolina State University) is a researcher with Media Cloud, the Initiative for Digital Public Infrastructure, and the Media Ecosystems Analysis Group, based at the University of Massachusetts Amherst.

Notes

1. Twitter made a statement on January 8 2021 about Trump’s suspension. His Twitter account was restored on November 20, 2022, following Elon Musk’s purchase of the platform.

2. See “$2 Billion Worth of Free Media for Donald Trump” from the New York Times.

3. For comparison, in 2022, Twitter has over 400 million users, whereas Gab had 1.1 million registered accounts in 2020.

4. See “Marjorie Taylor Greene and Big Tech’s never-ending censorship loop” from Vox and “Twitter permanently bans Alex Jones, website Infowars” from Reuters.

5. See “Trump’s Truth Social Really Is a (Tiny, Conservative) Phenomenon” in New York Magazine.

6. See “Trump launches ‘From the desk of Donald J. Trump’ as potential Facebook ban looms” from USA Today.

7. See “Former U.S. president Donald Trump launches ‘TRUTH’ social media platform” from Reuters.

8. See “Exclusive: Trump’s Truth Social app set for release Monday in Apple App Store, per executive” from Reuters.

9. See “Truth Social’s Influence Grows Despite Its Business Problems” from the New York Times.

10. The GitHub archive can be located at https://github.com/bpb27/trump_tweet_data_archive. We checked that the data from this archive is the same as the Trump Twitter Archive.

11. Similar to their work on the American political media landscape during the 2016 election (Benkler, Faris, & Roberts, 2018), Faris and a team at the Berkman Klein Center analyzed about 15,000 Twitter users active between January and June 2019 who self-identified as liberal/democrat or conservative/republican in their profile. They collected links shared by these users, assigned them a score (on a −1 to 1 scale) based on how frequently people on the left or right linked to them, and sorted the domains into five collections based on whether accounts on the left or right tweeted them significantly more often than the other side, somewhat more often, or evenly (Faris et al., 2020).

12. Our stringent measure quantifies direct embedding. Embedding a Twitter or Truth Social post requires a snippet of html that includes the username. If a page lacks a username, it lacks the embed code. Our broader measure of attention to Trump’s Twitter or Truth Social activity includes not only direct embedding, but also mentioning Trump and Twitter/Truth Social together in some other fashion, whether quoting him, paraphrasing, or something else.

13. See “Trump reports little income from Truth Social, $1 M from NFTs” from the Associated Press.