Vaccination has long been recognized as a crucial preventive measure against diseases and infections, but opposition to vaccines has endured [1]. HIV vaccination has been regarded as a potential preventive measure to help combat the HIV epidemic in the United States, with research and progress dating back to the mid-1980s but without success thus far [2]. An estimated 1.2 million people were living with HIV in the United States by the end of 2021, with 36,136 new HIV diagnoses reported in 2021 [3].

On January 27, 2022, the biotechnology company Moderna announced the initiation of clinical trials for an HIV vaccine using messenger RNA (mRNA) technology [4]. In March 2022, the National Institutes of Health announced the start of clinical trials for 3 mRNA HIV vaccines [5]. The mRNA technology had previously been used in the Pfizer-BioNTech and Moderna COVID-19 vaccines, which protected individuals against severe symptoms and fatalities during the pandemic [6]. Following the successes of mRNA COVID-19 vaccines, which led to the Nobel Prize in Physiology or Medicine being awarded to 2 scientists in October 2023 [7], researchers have been investigating the potential of mRNA vaccines for various other diseases, including HIV [8,9]. The announcements of clinical trials for mRNA HIV vaccines revived public discussion on the prospect of vaccines to combat HIV [9] despite >3 decades of unsuccessful research [2]. Meanwhile, these announcements were made against the backdrop of intense vaccine debates during the COVID-19 pandemic, with misinformation and conspiracy theories fueling vaccine hesitancy [10-12].

The X platform, formerly known as Twitter, has been a significant social media platform and a vital source for text-based public discourse. Posts on X have been studied to understand public discourse about vaccines in general [13-15] and about specific vaccines, such as COVID-19 vaccines in recent years [12,16,17]. However, there is a dearth of research about public discourse on HIV vaccines on social media. Given the advancement in mRNA technology in COVID-19 vaccines and heated vaccine debates, it has become especially important to gain insights into public discourse and reactions regarding potential new vaccines.

This study is grounded in the growing field of infodemiology and infoveillance, which investigates the “distribution and determinants of information in an electronic medium,” specifically on the web, by analyzing unstructured text with the aim of informing public health practices or serving surveillance objectives [18]. In recent infodemiology and infoveillance studies, machine learning algorithms have been increasingly used to examine substantial amounts of social media content, such as posts on X related to COVID-19 vaccines [12,16,17] and HIV prevention [19], to extract insights into public discourse and reactions. These algorithms automatically analyze extensive volumes of posts and capture latent textual information such as topics and sentiments. This study aimed to investigate how users used different post types to contribute original content to topics and valence identified through machine learning algorithms and how these topics and valence affected user reactions on X regarding HIV vaccines. In addition, by manually coding the most engaged posts, similar to an approach used in previous infodemiology and infoveillance research [20], the study intended to identify salient aspects of HIV vaccines related to COVID-19 as well as prominent anti–HIV vaccine conspiracy theories. Analyzing posts on X about HIV vaccines can shed light on public discourse and information diffusion. These findings have implications for shaping public health communication strategies about HIV vaccines [18]. Furthermore, the findings may help in understanding the acceptability of the HIV vaccine upon its successful development in comparison with adherence to existing HIV prevention measures. Previous infodemiology and infoveillance research effectively increased the forecast accuracy of COVID-19 vaccine uptake by leveraging insights derived from posts on X [21].

To understand public discourse and reactions surrounding HIV vaccines on the X platform, we put forward the following research questions (RQs):

We collected English-language original posts about HIV vaccines on the X platform from January 1, 2022, to December 31, 2022, using Netlytic [56]. The selected time frame began in January 2022 with the initiation of mRNA HIV vaccine clinical trials fueling public discussion and concluded in December 2022, a significant month for HIV and AIDS awareness marked by World AIDS Day on the first day of the month. Posts, excluding reposts, that contained both keywords (case insensitive)—“HIV” and “vaccine”—were extracted, resulting in a total of 36,424 posts across 365 days. Posts were collected weekly. Posts published from the last ending time point to at least 24 hours before each collection time point were included in the data set, allowing for a substantial reaction time.

The unit of analysis was a post. For each post, automated extraction produced data for user reactions (the number of likes and reposts) as well as account features (account verification status and follower count). All 36,424 posts underwent topic modeling using latent Dirichlet allocation (LDA) to identify latent topics, as well as sentiment analysis using Valence Aware Dictionary and Sentiment Reasoner (VADER) to access valence. LDA generated topic-specific loadings and identified the dominant topic for each post. VADER generated a valence compound score for each post, which was also categorized as positive, neutral, or negative based on standard VADER classification values.

LDA revealed 3 topics. As the topic of HIV and COVID-19 dominated in a large proportion of posts, we manually coded the 1000 most engaged posts containing the words “HIV” and “COVID” to uncover the salient aspects of HIV vaccines related to COVID-19. To develop coding for subtopics, 2 researchers initially reviewed and coded the top 200 posts with the most reactions. Subtopics were categorized by adapting existing categories from the literature [16,34] and integrating newly identified subtopics from the posts. The Scott π was 0.80 for categorizing subtopics. Subsequently, each researcher independently coded half of the remaining 800 posts.

We then conducted cross-tabulation analyses among all posts to examine the distribution of topics and valence among different types of posts. Furthermore, we conducted linear regression analyses among all posts to assess the influence of content and account features on these 1-click reactions. Of all 36,424 posts, 19,284 (52.94%) received ≥1 like, and 9155 (25.13%) received ≥1 repost. We added a constant value of 1 to all data points for likes and reposts before applying the natural logarithm. This was done to include posts with 0 likes or reposts and to mitigate the skewness of the data distribution.

Of the 28,439 posts that received likes or reposts, 6176 (21.72%) were negative. We manually coded the top 1000 negative posts with the most reactions to uncover prominent anti–HIV vaccine conspiracy theories. To develop coding for conspiracy theories, 2 researchers initially reviewed and coded the top 200 negative posts that received the most reactions. Posts containing conspiracy theories were identified based on expressions of postulated causal connections between people, objects, or events with malevolent intent [38,39]. Conspiracy theories were then classified based on the existing ones from the literature [50,51] and the emerging ones observed in the posts. Coding discrepancies were resolved through a further review of questionable posts and refinement of the conspiracy theories following the approach used in previous social media content analyses [40,57]. The procedure identified conspiracy theories and established intercoder reliability. The Scott π was 0.83 for identifying conspiracy theories and 0.81 for categorizing them. Each researcher then independently coded half of the remaining 800 negative posts.

We used cross-tabulation analyses to investigate the distribution of topics and valence across different post types, in which the dominant topic and valence category for each post were entered, respectively, alongside the post type. We used linear regression models to examine the message-level drivers of user reactions among posts that received likes or reposts. In the linear regression models, a constant value of 1 was added to all data points of like and repost counts. The natural log-transformed values for each post were then regressed on 3 topic-specific loadings generated from LDA, the valence compound score generated from VADER, and 2 autoextracted account features—account verification status and follower count. The “plus one” technique was used to include posts that received 0 likes or reposts and to address the skewness of the data distribution.

Following Long Island University’s institutional review board determination process, an institutional review board review was deemed unnecessary for this study, which collected and analyzed publicly available social media data. All referenced posts were paraphrased to avoid association with any particular user on the X platform.

RQ 1 asked about the topics present in all the posts. We trained a topic model using LDA exploring topic numbers ranging from 2 to 20. The optimal number of topics (k) was selected considering both the coherence score (C_v) and the topic model visualization in a Python library called pyLDAvis, as done in previous research [16,66]. C_v is a metric that reflects the semantic coherence of topics by evaluating the word co-occurrence likelihood within topics [67]. A higher C_v indicates a better classification achieved by the topic model. In this study, the model with 2 topics (k=2) yielded the highest C_v (0.42), whereas the model with 3 topics (k=3) yielded the second highest C_v (0.35). The pyLDAvis chart depicts each topic as a circle. Overlapping areas between circles suggest similarities in topics. Thus, a chart without overlapping circles is preferable for k. The pyLDAvis chart for this study showed that, when the value of k was 2 or 3, the circles did not overlap. However, when k reached 4, the circles began to overlap, and overlapping circles persisted for values of k ranging from 4 to 20. Between the k values of 2 and 3, we opted for a model comprising 3 topics (k=3) considering that a smaller number of topics tends to result in overly broad meanings for each topic [68].

Table 1 summarizes the 3 topics and lists their representative posts. Each topic was interpreted by examining the top 10 probable words ranked by frequency, along with sample posts exhibiting high topic-specific loadings and 1-click reactions. Topic 1 was HIV and COVID-19, covering 78% of the tokens [69] and dominating in 92.46% (33,678/36,424) of the posts. Topic 2 was mRNA HIV vaccine trials, covering 14% of the tokens and dominating in 5.91% (2151/36,424) of the posts. Topic 3 was HIV vaccine and immunity, covering 8% of the tokens and dominating in 1.63% (595/36,424) of the posts.

Figure 1 illustrates the daily numbers of original posts about HIV vaccines throughout 2022, in total and categorized into 3 topics. Moderna’s announcement of clinical trials for its first mRNA HIV vaccine on January 27, 2022, likely triggered the initial surge, culminating in a daily peak when the number of posts reached 805 on January 29, 2022. The daily number of posts about mRNA HIV vaccine trials (topic 2) in the week following Moderna’s announcement was higher than on other days throughout the year. Nevertheless, even during that week, there were higher daily numbers of posts about HIV and COVID-19 (topic 1), which remained dominant among the 3 topics during the entire year. The year’s second and highest daily peak occurred on February 8, 2022, recording a total of 1603 posts, most of which focused on HIV and COVID-19 (topic 1). This could be attributed to the emergence of new HIV-related events in early February 2022, including the promotion of HIV tests by public figures [64] and the discovery of a new highly virulent HIV strain [65]. The third highest daily peak, comprising 1085 posts, occurred on May 18, 2022, which has marked HIV Vaccine Awareness Day since 1998. Most of the posts centered on HIV and COVID-19 (topic 1). The remainder of the year did not reach such high peaks, with the largest daily volume of 205 posts occurring on December 2, 2022, the day following World AIDS Day, observed since 1988. Similar to previous daily peaks, most of the posts revolved around HIV and COVID-19 (topic 1).

The results revealed the dominance of HIV and COVID-19 (topic 1) in 92.46% (33,678/36,424) of the posts, with HIV as the most frequent word and COVID as the fourth most frequent word. To gain a deeper understanding of salient aspects of HIV vaccines related to COVID-19, we manually coded the top 1000 posts with the highest total number of likes and reposts that contained both HIV and COVID. Table 2 summarizes the subtopics and their representative posts with like and repost counts.

The first major subtopic, comprising 24% (240/1000) of the posts, focused on the reciprocal influence of HIV vaccines and COVID-19 vaccines on each other’s development. Years of HIV vaccine research facilitated the rapid development of mRNA COVID-19 vaccines, and the success of COVID-19 vaccines might accelerate the development of mRNA HIV vaccines. The second major subtopic, comprising 17.6% (176/1000) of the posts, involved comparisons between HIV and COVID-19 in various aspects. Specifically, the development speed of HIV vaccines compared to COVID-19 vaccines was a major point of comparison. In addition, some posts questioned whether potential HIV vaccines could be comparable to COVID-19 vaccines in terms of cost and accessibility during rollout. Others raised concerns about efficacy, safety, and inequality for both vaccines. The third major subtopic, comprising 26.5% (265/1000) of the posts, connected COVID-19 vaccines with HIV. One issue discussed was whether COVID-19 vaccines contained, caused, or increased HIV. Another issue raised was distinguishing between HIV symptoms and COVID-19 vaccine side effects, such as a fabricated condition called VAIDS, short for vaccine-acquired immunodeficiency syndrome. The fourth major subtopic, comprising 13.6% (136/1000) of the posts, featured conspiracy theories that presented hypothesized patterns linking COVID-19, HIV, and their vaccines with malicious intent. Prominent conspiracy theories in this subtopic included connecting misinformation that COVID-19 vaccines contain, cause, or increase HIV with the ongoing development of HIV vaccines; associating HIV and AIDS symptoms with side effects of COVID-19 vaccines; and claiming that COVID-19 originated from unsuccessful HIV vaccine research. As this study also manually coded the 1000 most engaged negative posts to identify prominent conspiracy theories, additional results pertaining to conspiracy theories will be discussed further in another subsection. The remaining posts related to HIV and COVID-19 included those that generally mentioned research on them or made connections without specifying details.

RQ 2 asked about the sentiment valence present in all the posts. According to the standard VADER classification values, valence is categorized by compound scores as follows: positive (0.05 to 1), neutral (−0.05 to 0.05), and negative (−0.05 to −1) [65]. On average, all posts had a marginally positive score of 0.053. HIV and COVID-19 (topic 1) had a slightly positive average score of 0.055. The mRNA HIV vaccine trials (topic 2) had a neutral average score of 0.040, leaning toward the positive side. HIV vaccine and immunity (topic 3) had a more neutral average score of −0.0008. Moreover, 42.78% (15,584/36,424) of the posts were positive, 25.64% (9338/36,424) of the posts were neutral, and 31.58% (11,502/36,424) of the posts were negative.

Of the 36,424 posts, 18,580 (51.01%) were replies, making up over half of the overall count. Self-composed posts totaled 41.6% (15,151/36,424), whereas the remaining 7.39% (2693/36,424) were quote posts. RQ 3 asked about the distribution of topics and valence among the 3 post types. As Table 3 shows, the distribution of topics varied by post type (N=36,424, χ²₄=2511.4, P<.001). Of the self-composed posts, 85.36% (12,933/15,151) focused on HIV and COVID-19 (topic 1) and 13.21% (2001/15,151) focused on mRNA HIV vaccine trials (topic 2). In comparison, quote posts and replies exhibited a different pattern, in each case >97% of posts centering on HIV and COVID-19 (topic 1; 2616/2693, 97.14% and 18,129/18,580, 97.57%, respectively).

As Table 4 shows, the distribution of valence also varied by post type (N=36,424, χ²₄=911.7, P<.001). The proportion of positive posts was slightly higher among self-composed posts at 44.95% (6810/15,151) compared to replies at 41.09% (7634/18,580) and quote posts at 42.33% (1140/2693). Self-composed posts had a smaller proportion of negative posts at 23.56% (3570/15,151) compared to replies at 37.64% (6994/18,580) and quote posts at 34.83% (938/2693). The proportion of neutral posts was larger for self-composed posts at 31.49% (4771/15,151) compared to quote posts at 22.84% (615/2693) and replies at 21.27% (3952/18,580).

Regarding the distribution of topics and valence among the 3 types of posts, quote posts and replies displayed similarities, whereas self-composed posts diverged. Compared to self-composed posts, which initiate new conversations, there was a higher proportion of HIV and COVID-19-related posts (topic 1) and a greater proportion of negative posts among quote posts and replies, which contribute to existing conversations.

RQ 4 asked about the influence of content and account features on likes and reposts.

Liking is more common than reposting. While 52.94% (19,284/36,424) of posts received an average of 24.83 likes, ranging from 1 to 102,843, a total of 25.13% (9155/36,424) posts received an average of 11.38 reposts, ranging from 1 to 10,572. Table 5 reveals the influence of content features (topics and valence) and account features (verification status and follower count) on the natural log-transformed number of likes and reposts. Both linear regression models were significant at P<.001. The adjusted R² was 0.072 for the like model and 0.090 for the repost model.

Among the 3 topics identified using LDA, HIV and COVID-19 (topic 1) did not affect like counts but decreased repost counts. In comparison, mRNA HIV vaccine trials (topic 2) decreased like counts while increasing repost counts. Positive valence increased like and repost counts. Account verification status and follower count increased like and repost counts.

Table 6 summarizes posts ranked within the top 5 for the number of likes and reposts presented in chronological order. It is worth noting that all posts in the top 5 for likes and reposts were self-composed. One particular post, which garnered the most likes (n=102,843) and reposts (n=10,572), expressed the incredible feeling of witnessing the development of an HIV vaccine within our lifetimes. It was posted by an unverified account on January 28, 2022, the day after Moderna’s announcement of clinical trials for its first mRNA HIV vaccine.

RQ 5 asked about prominent anti–HIV vaccine conspiracy theories. Of the 1000 negative posts that received the most reactions, 227 (22.7%) contained conspiracy theories. As Table 7 shows, we classified these prominent anti–HIV vaccine conspiracy theories into 4 categories and presented their representative posts and the number of reactions.

The first category, comprising 44.9% (102/227) of the posts, formulated conspiracy theories by connecting COVID-19, COVID-19 vaccines, HIV, and HIV vaccines. For instance, 52.9% (54/102) of these posts connected the misinformation regarding COVID-19 vaccines containing, causing, or increasing HIV with the ongoing efforts to develop HIV vaccines. This misinformation may have arisen from past occurrences resurfacing following Moderna’s initiation of its mRNA HIV vaccine trials. One incident occurred at the end of 2020, when an Australian COVID-19 vaccine, which used a small fragment of protein from HIV to clamp SARS-CoV-2’s spike proteins, was abandoned due to false HIV-positive results [70]. Another incident occurred in October 2020, when 4 researchers sent a letter to a medical journal expressing concerns about the potential increased risk of HIV acquisition among men receiving COVID-19 vaccines using adenovirus type-5 vectors without supporting data from COVID-19 vaccines [71]. The misinformation typically interpreted the incidents out of context and generally suggested that COVID-19 vaccines contained, caused, or increased HIV without specifying details. In addition, there were conspiracy theories linking HIV and AIDS to COVID-19 vaccine side effects, including a fabricated condition known as VAIDS. VAIDS falsely suggests that COVID-19 vaccines caused immune deficiency [72]. Furthermore, there were claims that COVID-19 originated from unsuccessful HIV vaccine research.

The second category, comprising 38.3% (87/227) of the posts, suggested that the alignment of concurrent events with Moderna’s start of mRNA HIV vaccine trials in late January 2022 was intentional to manipulate the market for HIV vaccines. These events included the rising HIV discussion and fear; promotion of HIV tests by public figures [73]; the discovery of a new highly virulent HIV strain [74]; and the passing away of HIV researchers, including Luc Montagnier, codiscoverer of HIV with an antivaccine stance during the COVID-19 pandemic [75], all occurring in early February 2022.

The third category, with 11.5% (26/227) of the posts, revealed conspiracy theories based on the distrust of powerholders [76]. Some posts extended existing conspiracy theories, such as the Big Pharma conspiracy theory [50] and the New World Order conspiracy theory [51], into the context of HIV vaccines, emphasizing the intent of powerholders, including major pharmaceutical companies and governments, behind vaccine promotion for financial profits and society control. Other posts created conspiracy theories about the government’s research on HIV vaccines. The remaining posts generally stated that HIV vaccines were a scam. The final category comprised the remaining 5.3% (12/227) of the posts with other conspiracy theories.

It is worth noting that, of the 227 posts containing conspiracy theories, 39 (17.2%) were posted by accounts that had already been suspended at the time of manual coding. For these posts, the X platform displays the following message—“This post is from a suspended account”—and the content of the post is not visible. The X platform suspends accounts that violate its rules [77]. However, specific details of the violations are not accessible on the platform. The invisibility of these posts halted their spread when the suspension was enacted. For our manual coding of these posts, we used the text obtained during the data collection process.

This study investigated the patterns of public discourse and the message-level drivers of user reactions on the X platform regarding HIV vaccines through the analysis of posts using machine learning algorithms. We examined the distribution of topics and valence across different post types and assessed the influence of content features (topics and valence) and account features (account verification status and follower count) on like and repost counts. In addition, we manually coded the 1000 most engaged posts about HIV and COVID-19 to understand the salient aspects of HIV vaccines related to COVID-19 and the 1000 most engaged negative posts to identify prominent anti–HIV vaccine conspiracy theories.

The results revealed that COVID-19 plays a substantial role as a context for public discourse and reactions regarding HIV vaccines. Of the 3 topics identified using LDA, the leading topic was HIV and COVID-19, covering 78% of tokens and dominating in 92.46% (33,678/36,424) of the posts. Furthermore, on each of the top 4 days with the highest post counts, most of the posts were about HIV and COVID-19. This comprehensive topic included important subtopics that linked HIV vaccines with COVID-19 vaccines, as demonstrated through the manual coding of the 1000 most engaged posts about HIV and COVID-19. These subtopics encompassed the reciprocal influence of HIV vaccines and COVID-19 vaccines in advancing each other’s development; comparisons in their development speed; inquiries about the possible alignment of HIV vaccines with COVID-19 vaccines in terms of cost and accessibility during distribution; and concerns about efficacy, safety, and equality for both vaccines.

COVID-19 positioned HIV vaccines in both a positive and negative context. On the one hand, the success of mRNA technology in COVID-19 vaccines [6] potentially cast mRNA HIV vaccines in a positive light. The topic of HIV and COVID-19 had a marginally positive valence score of 0.055. Moreover, 3 (60%) out of the 5 most liked posts and 2 (40%) out of the 5 most reposted posts expressed excitement about advancements in HIV vaccines that were based on the experience with COVID-19 vaccines. On the other hand, antivaccine discourse, including conspiracy theories, heated up during the COVID-19 pandemic [10,11,27,44,48,49], which posed challenges to HIV vaccines. Of the 1000 most engaged posts about HIV and COVID-19, a total of 136 (13.6%) featured conspiracy theories. Of the 1000 most engaged negative posts, 227 (22.7%) contained conspiracy theories, with 102 (44.9%) of them revolving around HIV and COVID-19. For instance, a prominent conspiracy theory connected the misinformation about COVID-19 vaccines containing, causing, or increasing HIV infection [55] with the initiation of clinical trials for mRNA HIV vaccines [4,5], implying a malevolent intent behind the deliberate connection. The results indicate that conspiracy theories tend to elicit an approach-oriented response, as evidenced by people engaging in liking and reposting, as opposed to an avoidance-oriented approach [39]. This underscores the need to intensify efforts to counter conspiracy theories in public health communication about HIV vaccines.

According to a study conducted by the Pew Research Center, irrespective of the subject matter, replies constituted the largest portion of original posts on X, followed by self-composed and quote posts [28]. Specifically, the number of replies was 3 times greater than that of self-composed posts. In this study, although replies constituted slightly more than half (18,580/36,424, 51.01%) of the posts, it is worth noting that the subject of HIV vaccines elicited a higher proportion of self-composed posts at 41.6% (15,151/36,424). Specifically, the number of replies was 23% higher than that of self-composed posts. Moreover, the topic of mRNA vaccine trials was most evident in self-composed posts compared to replies and quote posts. In comparison, there was a higher proportion of focus on the topic of HIV and COVID-19 and a greater proportion of negative posts among quote posts and replies, which contribute to existing conversations. This suggests that users were more likely to initiate new conversations rather than joining existing conversations about mRNA HIV vaccines. In contrast, they were more likely to join existing conversations rather than starting new conversations about HIV and COVID-19. In addition, users were less likely to initiate new conversations negatively but more likely to contribute negatively to existing ones.

As the primary topic, HIV and COVID-19 had no impact on like counts but had a negative impact on repost counts. In comparison, the topic of mRNA HIV vaccine trials had a negative impact on like counts and a positive impact on repost counts. The results should be interpreted while considering that, as revealed in previous research [16,34] and this study, most posts on the X platform are unlikely to receive likes and even less likely to receive reposts. In this study, among the total of 36,424 posts, approximately half (n=19,284, 52.94%) received likes, and approximately one-quarter (n=9155, 25.13%) received reposts. To include all posts and mitigate the data distribution skewness in the linear regression analysis, we applied the “plus one” technique. This involved adding a constant value of 1 to all like and repost data points before taking the natural logarithm. Although most posts were not liked or reposted, it is noteworthy that the topic of mRNA HIV vaccines led to an increase in repost counts, highlighting its positive influence on social sharing. In addition, 2 (40%) out of the 5 most reposted posts were about mRNA HIV vaccine trials. These results correspond to the findings of previous research that suggested the diffusion of novel useful information [12,16,32,36].

The overall valence of the posts about HIV vaccines was marginally positive. The positivity aligns with the positive sentiment found in posts on X about vaccines in general [13-15] and COVID-19 vaccines in particular [12,16,17]. However, the positivity about HIV vaccines was not apparent as the average score of 0.053 placed it on the edge of the neutral range, which goes from −0.05 to 0.05 according to the standard VADER classification values. Positive sentiment had a favorable impact on like and repost counts, partially consistent with findings of previous research on COVID-19 vaccines [16]. The post that achieved the most likes conveyed the incredible feeling of witnessing the development of an HIV vaccine in our lifetimes. This could be attributed to the psychological rationale that social transmission of positive content fulfills people’s motivation to present a positive image [35,37]. In alignment with the findings of previous research [13,16,34], account verification status and follower count increased like and repost counts.

This study has implications for public health communication related to HIV vaccines and potentially other vaccines. Given the massive scale of the COVID-19 vaccination campaign, it is understandable that people will draw comparisons with other vaccines. Topic modeling identified HIV and COVID-19 as the primary topic, and manual coding revealed various intertwined aspects. Leveraging the advantages observed in the COVID-19 vaccine campaign, such as its widespread accessibility, could be valuable. Furthermore, addressing common concerns such as efficacy, safety, and inequality could also prove beneficial.

In the case of HIV vaccines, it is essential to tackle concerns associated with COVID-19 vaccines, especially those related to HIV vaccines. A major subtopic of HIV and COVID-19 involved suspicions about COVID-19 vaccines containing, causing, or increasing HIV. Another major subtopic was the confusion between HIV symptoms and the alleged side effects of COVID-19 vaccines, such as VAIDS. Misinformation concerning both subtopics has been woven into conspiracy theories, further complicating this situation. To combat misinformation and conspiracies that have these elements, efforts could focus on promoting evidence-based factual information [45-47].

Another notable technique in the conspiracy theories was linking concurrent COVID-19 and other HIV-related events in unsubstantiated relationships to create false perceptions, suggesting that these events were intentional to manipulate the market for HIV vaccines. These HIV-related events included rising HIV discussion and fear, promotion of HIV tests by public figures [73], the discovery of a new highly virulent HIV strain [74], and the passing away of HIV researchers, all occurring in early February 2022. These findings suggest that refuting false connections among such concurrent events can be an effective strategy to counter these conspiracy theories [45-47]. These occurrences, frequently entwined within conspiracy theories, could be specifically addressed in public health communication efforts.

This study has several limitations. Because we used autoidentified content features (topics and valence) and autoextracted account features (verification status and follower count) in the regression models to predict the autoextracted number of user reactions (likes and reposts), the results were mostly limited to the examined autoidentified and autoextracted factors. For instance, political polarization, which manifested in a wide range of issues, including response to vaccines [78], could be a factor worth investigating in future studies. Furthermore, manual coding of conspiracy theories revealed a prevalent technique of twisting concurrent events into false relationships. This underscores the significance of refuting unfounded associations among these incidents to counter such conspiracy theories. It will be interesting for future research to assess the impact of this technique on user reactions to conspiracy theories. These findings could provide further insights into public health communication strategies to combat conspiracy theories.

The results highlight COVID-19 as a significant backdrop for public discourse and reactions on the X platform regarding HIV vaccines. COVID-19 situated HIV vaccines in both a positive and negative context. The success of mRNA COVID-19 vaccines shed a positive light on HIV vaccines. However, COVID-19 also situated HIV vaccines in a negative context, as evident in anti–HIV vaccine conspiracy theories falsely linking HIV vaccines to COVID-19. The findings provide implications for public health communication strategies concerning HIV vaccines.