EDITOR’S NOTE This report was produced as an independent analytical exercise. It does not represent the position of any political party, government body, electoral commission, civil society organisation, or media outlet. Thraets has no institutional affiliation with, financial interest in, or advisory relationship to either of the individuals examined in this study. No funding was received from any source with a direct stake in Uganda’s electoral process. The subject matter of this report, Uganda’s 2026 general elections and the political figures associated with them, is, by its nature, politically sensitive. The comparative analysis presented here examines publicly observable social media behaviour using quantitative and computational methods. Wherever possible, findings are presented descriptively, grounded in the data, and qualified where interpretation is required. The research does not seek to validate or invalidate the official election results, nor to make claims about the authenticity of the political positions held by either figure studied. Readers are encouraged to engage critically with the findings and to consult the methodology section and appendices before drawing conclusions. The dataset, variable definitions, and analytical choices are documented in full. Where limitations are known — including constraints on sentiment analysis accuracy, the exclusion of deleted content, and the inability to verify organic versus coordinated engagement using public data alone — these are disclosed explicitly within the report. All images included in this report are sourced from publicly available material and are used solely for illustrative and analytical purposes. No copyright ownership is claimed over any third-party imagery. Screenshots have been reproduced for the purposes of analysis and commentary only. All graphs and charts were generated by the Thraets Investigation team from an original dataset of 2,448 publicly accessible tweets collected between January 2024 and December 2025. 1 2 TABLE OF CONTENTS Editor’s Note..........................................................................................................................................................1 ABSTRACT.............................................................................................................................................................3 INTRODUCTION................................................................................................................................................. 4 METHODOLOGY.................................................................................................................................................6 RESULTS.................................................................................................................................................................8 Data Collection and Corpus Overview ......................................................................................................10 Engagement......................................................................................................................................................12 Thematic Distribution....................................................................................................................................16 Sentiment......................................................................................................................................................... 19 Hashtag Infrastructure................................................................................................................................. 21 International Orientation..............................................................................................................................24 Political Concern Documentation...............................................................................................................26 Temporal and Posting Patterns....................................................................................................................27 Named Figure Co-occurrence......................................................................................................................31 Vocabulary Profile..........................................................................................................................................33 DISCUSSION...................................................................................................................................................... 36 Strengths and Limitations..............................................................................................................................39 INFORMATION OPERATIONS ASSESSMENT.......................................................................................... 41 CONCLUSION AND RECOMMENDATION................................................................................................43 ABSTRACT Uganda's January 2026 general elections concluded with President Yoweri Museveni declared the winner with 71.65% of the vote. The election was conducted amid a government-imposed internet blackout and opposition allegations of fraud. As the 2026 general elections approached, social media platforms such as X (formerly Twitter) became central arenas for political communication and commentation, mobilisation, and advocacy. This study presents a comparative analysis of Twitter/X activity by two prominent Ugandan political figures, Robert Kyagulanyi Ssentamu (Bobi Wine, @HEBobiwine) and General Muhoozi Kainerugaba (@mkainerugaba), during the pre-election period from 1st January 2024 to 31st December 2025. Using a dataset of 2,448 tweets, the research analyses posting frequency, thematic content, engagement patterns, and narrative framing. Key findings indicate that Bobi Wine maintains a substantially higher output volume (1,504 vs 944 tweets) and a 3.43× higher engagement rate (4.08% vs 1.19%). The findings reveal distinct communication strategies. Bobi Wine's content was anchored in democratic mobilisation, human rights documentation, and regime criticism, while Muhoozi's content centred on institutional military identity, religious affirmation, and regional diplomacy. These patterns illustrate how social media platforms are used differently by opposition movements and political elites within hybrid political systems. The study contributes to the literature on digital political communication in Africa and offers insights into the role of social media in shaping political discourse ahead of Uganda’s January 2026 elections. The most operationally significant asymmetry concerns hashtag infrastructure. Bobi Wine deployed 1,624 hashtags across 122 unique tags, using a coordinated campaign architecture. Muhoozi used hashtags on two occasions. International orientation further distinguishes the two figures. Bobi Wine tended to cultivate Western liberal democratic audiences, and Muhoozi positioned himself as an African regional statesman with documented engagement with non- Western powers. These findings show how social media platforms, particularly X, supported both grassroots mobilisation and elite political communication in the lead-up to the 2026 elections in Uganda. Keywords : social media intelligence, Uganda, political communication, Twitter/X, information operations, pre-election monitoring, sentiment analysis, digital mobilisation 3 INTRODUCTION Uganda's January 2026 general elections took place in a political environment shaped by nearly four decades of uninterrupted rule under President Yoweri Kaguta Museveni, who has governed since 1986. The 2021 electoral cycle established the immediate precedent: Bobi Wine officially received 35.6% of the vote against Museveni's 58.6%, though the voting and counting periods were accompanied by a complete internet shutdown, widespread pre-election violence against NUP supporters, and Bobi Wine's subsequent detention under de facto house arrest. That pattern was repeated in 2026. Elections were held on 15 January under a government- imposed mobile internet suspension. On 17 January, the Electoral Commission declared Museveni the winner with over 71% of the vote. Bobi Wine, who received 24.72% of the official count, rejected the results as fraudulent and went into hiding after a police raid on his home. Within this context, Twitter/X has emerged as the primary publicly accessible arena in which competing political narratives are constructed, contested, and disseminated. For opposition movements, social media platforms represent one of the few relatively open channels through which mass communication and political organising are possible. For establishment-aligned figures, the same platforms offer opportunities to project authority, cultivate diplomatic relationships, and shape international perceptions without the accountability constraints of formal institutional communication. The two figures examined in this report occupy diametrically opposing positions within Uganda’s political landscape. Robert Kyagulanyi Ssentamu , popularly known as Bobi Wine, is the President of the National Unity Platform (NUP) party and the widely recognised principal challenger to President Yoweri Kaguta Museveni. Born 12 February 1982, Robert Kyagulanyi Ssentamu rose to national prominence as one of Uganda's most successful popular musicians before entering formal politics in 2017, winning a parliamentary by-election to represent Kyadondo East. His People Power movement, which was later institutionalised as the National Unity Platform, grew into Uganda's largest opposition party. His political career was marked by documented physical abuse during a 2018 detention, multiple subsequent arrests, harassment of his wife, Barbara ‘Barbie' Itungo Kyagulanyi, the killing or detention of hundreds of NUP supporters, and sustained legal obstacles to political organising. He ran against Museveni in both the 2021 and 2026 elections, receiving an official vote share of 24.72% in the latter - a result he rejected as fraudulent. Following the declaration of results on 17 January 2026, he went into hiding after a police raid on his home, and he is now in exile. 4 Despite growing attention to social media in African politics, there remains a limited systematic comparative analysis of how opposition and establishment actors use these platforms differently. This study addresses that gap by examining patterns of communication, engagement, and content strategy across a two-year pre-election period. The study is guided by three research questions. First, how do the social media outputs, audience engagement patterns, and content strategies of Bobi Wine and Muhoozi Kainerugaba differ? Second, what do these differences reveal about their respective political positioning and mobilisation approaches? Third, what do these patterns suggest about the broader information environment in which the 2026 election took place? In addressing these questions, the study contributes to a broader understanding of how social media functions differently for incumbents and challengers in hybrid political systems. General Muhoozi Kainerugaba, the eldest son of President Museveni and Commander of UPDF Land Forces, has cultivated an increasingly prominent public profile through a Twitter presence characterised by regional diplomatic engagement, religious messaging, and periodic controversial statements that have generated some diplomatic incidents. Born on 24 April 1974, Muhoozi Kainerugaba is the eldest son of President Museveni and First Lady Janet Museveni, Uganda’s current Minister of Education. His military career has included command of the Special Forces Command and promotion to full General in May 2023, followed by appointment as Commander of UPDF Land Forces. His Twitter presence has generated multiple diplomatic controversies, most notably in October 2022 when tweets suggesting the UPDF could capture Nairobi within two weeks precipitated a diplomatic incident and his temporary removal from SFC command. He remains widely regarded as a potential succession candidate within Uganda's ruling establishment. 5 METHODOLOGY Context and Setting The study covers the two years immediately preceding Uganda's January 2026 general elections. The dataset spans from 1 January 2024 through 29 December 2025. This window was selected to capture sustained communication patterns across the full pre-election mobilisation cycle. The election itself, held on 15 January 2026, with results declared on 17 January, falls outside the dataset but provides the outcome context against which the findings are interpreted. Study Design This is a retrospective, cross-sectional comparative analysis of publicly available social media data. No primary data collection involving human subjects was undertaken. All material examined consists of publicly posted content. The comparative design enables systematic characterisation of within-period differences in communication strategy, audience behaviour, and content focus between the two accounts. Data Source and Collection The dataset includes 2,448 tweets from the public Twitter profiles of Bobi Wine (@HEBobiwine; 1,504 tweets) and Muhoozi Kainerugaba (@mkainerugaba; 944 tweets) during the study period. For each tweet, the dataset records the tweet URL, full text content (including mentions, hashtags, and links), creation timestamp (UTC), bookmark count, like count, quote count, reply count, retweet count, and view count. Only publicly accessible posts were included; deleted tweets, protected accounts, and platform-internal analytics were not available. A limitation we found was that inauthentic or bot-generated interactions cannot be distinguished from organic engagement using public tweet data alone. Engagement rates should therefore be understood as rates of observable interaction with each account's content, not as validated measures of authentic audience behaviour. Network-level analysis would be required to assess coordinated amplification reliably. Main Study Variables The primary outcome variable is engagement rate, defined as (favourites + retweets) / views x 100. This controls for reach and isolates the proportion of viewers who actively interact with content. Secondary variables include: 1.tweet volume and posting frequency; 2.sentiment polarity and subjectivity scores; 3.thematic category proportions; 4.hashtag frequency and diversity; 5.named entity and foreign country mention counts; 6.temporal posting distributions by hour-of-day and day-of-week. 6 Analytical Methods Quantitative Metrics Descriptive statistics (mean, median, standard deviation, quartile distribution) were calculated for all engagement metrics. Outlier tweets were identified at the 95th percentile of views for qualitative examination. Engagement rate was computed as described above and compared across accounts and across the time series. Sentiment Analysis Sentiment scoring used the TextBlob natural language processing library, which assigns each text a polarity score from -1.0 (maximally negative) to +1.0 (maximally positive) and a subjectivity score from 0.0 (objective) to 1.0 (subjective). Tweets were categorised as Positive (polarity > 0.1), Neutral (-0.1 to 0.1), or Negative (< -0.1). TextBlob performs well on standard English but has reduced accuracy for Luganda and Swahili vocabulary, political neologisms, irony, and context- dependent meaning; sentiment scores are therefore treated as indicative of directional patterns rather than precise measurements. Topic Modelling Latent Dirichlet Allocation (LDA) was applied independently to each account corpus following standard preprocessing: URL removal, mention removal, hashtag removal (analysed separately), punctuation stripping, lowercasing, and stopword filtering augmented with domain-specific terms. Eight latent topics were extracted per account to enable meaningful cross-account comparison, and topic distributions were examined for temporal evolution across the analysis period. Thematic Classification A domain-specific keyword dictionary was developed to reflect the Ugandan political context across nine categories: Democracy/Elections; Human Rights; Military/Security; Foreign Policy; Religion; Regime Criticism; National Unity; Condolences; and Celebrations. Multi-label classification was applied, since individual tweets frequently address more than one theme simultaneously. Full keyword lists are provided in the Appendix. Temporal Pattern Analysis Posting frequency was aggregated at hourly, daily, weekly, and monthly intervals. Hour-of-day distributions (UTC, interpreted as EAT = UTC+3) were examined for evidence of automated posting. Volume spikes were cross-referenced against the timeline of known political events in Uganda during the analysis period. 7 Metric Bobi Wine (@HEBobiwine) Muhoozi Kainerugaba (@mkainerugaba) Total tweets 1,504 944 Analysis period Jan 2024-Dec 2025 Jan 2024-Dec 2025 Retweets of others 0 (0%) 0 (0%) Average views per tweet 147,942 155,942 Average favourites 4,404 1,823 Average retweets 706 239 Engagement rate (mean) 4.74% 1.35% Mean sentiment score 0.071 0.131 Positive tweets (%) 32.80% 37.00% Negative tweets (%) 12.10% 6.20% Total hashtag uses 1,624 2 Unique hashtags used 122 1 RESULTS Data Collection and Corpus Overview The final dataset comprises 1,504 tweets from @HEBobiwine and 944 from @mkainerugaba across the full 24-month window, with both accounts producing content throughout the period without extended gaps. Neither account retweeted any third-party content within the dataset. Summary statistics are presented in Table 1, with Figure 1 providing a visual overview of the eight key comparative dimensions. Table 1. Summary statistics for both accounts, January 2024 - December 2025. 8 Graph 1: Key Performance Metrics: Total followers, tweet volume and average per-tweet engagement for the 24-month analysis period. Graph 2: Avg. Engagement Composition per Tweet: Breakdown of average favourites, retweets and replies per tweet. 9 Engagement Despite comparable average view counts, Bobi Wine's engagement rate of 4.08% is 3.43 times higher than Muhoozi's 1.19%. This differential is consistent across every sub-metric: average favourites favour Bobi Wine by 2.4:1 (4,404 vs. 1,823); average retweets by 3.0:1 (706 vs. 239). Quote-tweets, replies, and bookmarks follow the same pattern. The consistency of this advantage across all interaction types rules out metric-specific explanations and establishes a structural difference in audience behaviour. The figures below provide a detailed engagement analysis. The engagement rate trend (top left) shows Bobi Wine's rate rising sharply in the final months of the analysis period, reaching nearly 7% by December 2025 - the immediate pre-election window. The views vs. favourites scatter (top right) shows Bobi Wine's cluster with a steeper conversion slope. The top 10 tweets by views (bottom left) show broadly comparable peak reach for both accounts, with Muhoozi's single highest-viewed tweet marginally exceeding Bobi Wine's. The engagement rate distribution histogram (bottom right) confirms a systematic rightward shift in Bobi Wine's distribution, with a mean of 4.74% versus 1.35%. Graph 3: Engagement Metrics Comparison 10 Bobi Wine's cluster exhibits a steeper conversion slope, where more views lead to a proportionally greater number of likes. Graph 5: Engagement Rate Distribution Graph 4: Views vs. Favourites There is a systematic rightward shift in Bobi Wine's distribution (mean 4.74% vs 1.35%). The gap is structural and not driven by outliers. Graph 6: Bobi Wine Top Tweets by Total Engagement 11 Controversy-driven peaks dominate the highest-engagement posts. Despite comparable average view counts, Bobi Wine's engagement rate of 4.08% is 3.43 times higher than Muhoozi's 1.19%. This differential is consistent across every sub-metric. The consistency of this advantage across all interaction types rules out metric-specific explanations and establishes a structural difference in audience behaviour. Graph 7: Muhoozi’s Top Tweets by Total Engagement Thematic Distribution Bobi Wine's content is anchored in four dominant themes: National Unity (74.5% of tweets), Democracy and Elections (38.0%), Regime Criticism (33.3%), and Human Rights (29.7%). Military/Security content accounts for a further 25.2%, almost entirely in a critical frame documenting alleged security force conduct. Condolences comprise 18.2% of his output, reflecting the volume of tweets mourning supporters killed or detained during the period. Religion accounts for 11%. Muhoozi's thematic distribution is relatively balanced. The theme of National Unity appears at a significantly lower rate of 22.4%. The second most prominent theme is Military/Security, which accounts for 19.6% and is framed in a positive and institutional context. Religion, at 16.3%, and Foreign Policy, at 14.8%, are similar in their levels of prominence. However, Democracy (1.8%) and Human Rights (1.7%) are nearly absent from the distribution. The core topics identified from 1,504 tweets are weighted by their frequency of mention, with a strong emphasis on human rights and democracy. Top-performing tweets are ranked based on the total number of favourites, retweets, and replies. 12 Graph 8: Bobi Wine’s Topic Frequency Two fundamentally different content universes. Bobi Wine's topics cluster around democratic accountability and state violence, while Muhoozi's centre on military identity, religion and regional diplomacy, with no thematic overlap. Graph 9: Muhoozi Kainerugaba’s Topic Frequency Graph 10: Narrative Architecture 13 Sentiment Bobi Wine's mean sentiment score of +0.071 is lower than Muhoozi's +0.131. His negative tweet proportion (12.1%) is nearly twice Muhoozi's (6.2%), whereas his positive tweet proportion (32.8%) is marginally lower than Muhoozi's (37.0%). Both accounts are majority-neutral, consistent with high-volume public-figure accounts that mix informational and opinion content. Temporal analysis shows Bobi Wine's most negative monthly averages coincide with documented crisis events, major arrests, violence against supporters, and significant electoral manipulation allegations, and his most positive periods correspond to campaign successes and international advocacy moments. Graph 11: Bobi Wine’s Weekly Sentiment: Share of positive, neutral and negative tweets per week. Graph 12: Muhoozi Kainerugaba’s Weekly Sentiment: Share of positive, neutral and negative tweets per week. 14 Graph 13: Overall Sentiment Breakdown Bobi Wine carries almost twice the negative sentiment load (12.1% vs 6.2%), indicating his critical messaging against the opposition. Hashtag Infrastructure Bobi Wine utilised hashtags 1,624 times across 122 unique tags, averaging more than one hashtag per tweet. His ten most frequently used tags were: #PeoplePowerOurPower (361 uses), #ANewUgandaNow (312), #ProtestVote2026 (278), #FreeUgandaNow (136), #FreeUganda (103), #FreeAllPoliticalPrisonersInUganda (62), #BringBackOurPeople (55), #ProtestVoteUG2026 (53), #FreeAllPoliticalPrisoners (31), and #FreeKizzaBesigye (12). In contrast, Muhoozi only used hashtags twice across 944 tweets, both related to Rwandan Liberation Day, specifically #Kwibohora31. Notably, his dataset contains no campaign hashtags, movement identity tags, or calls for mobilisation. 15 Graph 14: Bobi Wine’s Top Hashtags by Usage #PeoplePowerOurPower leads at 361 uses. The top 3 hashtags alone account for 951 coordinated deployments. Graph 15: Cumulative adoption (normalised) of Bobi Wine's major hashtags over the campaign period. 16 Metric Bobi Wine (@HEBobiwine) Muhoozi Kainerugaba (@mkainerugaba) Media freedom 211 24 Abductions/disappearances 158 2 Arrests/detention 153 14 Torture/abuse 149 9 Education 86 4 Healthcare 80 1 Corruption 79 9 Unemployment 40 8 Economy 29 4 International Orientation Bobi Wine's mentions of foreign countries concentrate heavily in Western liberal democracies: the EU/UK (450 mentions combined) and the United States (68). Regional African neighbours receive comparatively little attention: Kenya (41), the DRC (3), and Rwanda (1). Muhoozi's international engagements are distributed across the following regions: Rwanda (47), DRC/Congo (29), Kenya (15), South Sudan (12), Sudan (11), Ethiopia (7), Somalia (5). Russia appears 11 times in Muhoozi's feed, including explicit commentary on the reported recruitment of Ugandan nationals for service in the Russia-Ukraine conflict. In contrast, it is absent from Bobi Wine's feed. Political Concern Documentation Across all political concern categories, Bobi Wine produces substantially higher tweet counts (Table 2). The largest gaps are in human rights-specific categories. Even development-oriented concerns follow the same directional pattern. 17 Temporal and Posting Patterns Bobi Wine's output exceeds Muhoozi's in nearly every month of the analysis period, with spikes in output correlating with identifiable political events: campaign tour launches, major arrests or incidents of violence, electoral registration periods, and international advocacy moments. Muhoozi's volume is more irregular and shows less consistent correlation with external events. Both accounts post predominantly within the 06:00-20:00 UTC (09:00-23:00 EAT) window, consistent with human daytime activity. Neither account exhibits the mechanical regularity or overnight posting patterns associated with automated content generation. Graph 16: Bobi Wine’s Daily Tweet Activity from 2024–2025 Graph 17: Muhoozi’s Daily Tweet Activity from 2024–2025 Each individual cell corresponds to a single day, with the intensity of the colour reflecting the volume of tweets generated on that day. A darker shade signifies a higher number of tweets, while a lighter shade indicates fewer tweets. 18 Bobi Wine tends to gain the most traction and visibility for his content around midday, as this is when his audience is most engaged. In contrast, Muhoozi often shares his posts later in the day, particularly in the afternoon, when followers are likely to be more active online. Graph 18: Combined tweet volume across all 24 hours. Graph 19: Aggregate tweet count per week. Bobi Wine's public attention experiences noticeable spikes that correlate closely with significant events in his political campaign and various human rights incidents. These fluctuations indicate a strong relationship between his visibility and crucial moments, reflecting how both his political activities and the broader human rights landscape capture public interest and media coverage. 19