To investigate discussions surrounding COVID-19 treatment on social media, we collected data from various social media platforms using a comprehensive search query (Table S1 in Multimedia Appendix 1). Data collection spanned from September 2019 to September 2022 and focused on capturing data from the United Kingdom, specifically English language content. Below, we outline our data sources and search strategy.

This approach yielded a dataset of 709,634 posts.

Following the search, the generated dataset of 709,634 posts covering September 2019 to September 2021 was refined via a multistep process.

The steps are listed in Textbox 1.

To prepare the data for analysis, we performed several text processing steps using a Python pipeline [19]:

For data categorization, we used Named Entity Recognition (NER) to classify unstructured text into predefined categories. We then identified key terms (~5000 in total) based on frequency (eg, “COVID” or “patient”). The list was manually reviewed and structured into a 3-tier hierarchy:

Themes included symptoms, infection, testing, diagnosis, and treatment. Synonyms and related terms were grouped accordingly. Through this process, 89.7% of posts were categorized.

With all posts in the dataset categorized, we analyzed the data using 2 approaches listed in Textbox 2:

To cross-validate social media trends, we examined Google search data using AnswerThePublic [21], Google Trends [22], and Ahrefs [23]. These platforms provided insights into search volume and query types related to COVID-19 treatment, allowing us to compare social media discussions with public search behavior.

This study did not require human participant research review, as it did not meet the criteria for such oversight. All data were sourced from publicly accessible social media platforms. However, due to the inclusion of health-related content concerning COVID-19, additional privacy safeguards were implemented. Personally identifiable information was protected by using obfuscated identifiers; social media post IDs and usernames were redacted during data collection.

Following the search, the generated dataset of 709,634 posts covering September 2019 to September 2021 was refined via a multistep process (Figure 1). This comprehensive cleaning process resulted in a final dataset of 31,319 posts.

The final dataset is primarily composed of data from Twitter (subsequently rebranded X), with a minimal contribution from forums and other social media platforms such as Facebook, Instagram or YouTube.

Trends in social media discussions closely aligned with Google search trends. Social media posts emerged in December 2019, with the highest peak (107,796 posts) in March 2020, coinciding with the first UK lockdown [4]. Discussion volume decreased in late 2020 but spiked in response to COVID-19 waves, new variants, and treatment announcements (Figure 2). Synthesio data trends closely mirrored Google search patterns (Figure 2).

Notably, the Omicron variant was more prominent in social media discussions (3824 posts, 12.0%) compared to Delta (894 posts, 2.9%) (Figure S1 in Multimedia Appendix 1).

Thematic analysis was performed on the cleaned dataset (n=31,319) to further examine the search trends and volume on predetermined stages of the treatment pathway, that is, symptoms → testing → diagnosis → treatment.

Our study examined public perceptions of COVID-19 symptoms, testing, diagnosis, and treatment through social media between September 2019 and September 2022. Public interest fluctuated, with peaks occurring during lockdowns, variant outbreaks, and treatment availability. The analysis of public perception of COVID-19 over time showed that the public attention fluctuated, with peaks occurring when lockdowns were introduced and when new variants emerged. The COVID-19 pandemic created an unparalleled sense of uncertainty and widespread health anxiety across the globe [26]. Despite or perhaps due to the vast amount of available information, uncertainty remains regarding the characteristics of the virus, as well as the emergence of new waves and variants. Studies show that information-seeking and knowledge-sharing practices increase rapidly in response to emergencies: since infectious diseases such as HIV or AIDS, acute respiratory syndrome, and Asian bird flu have emerged, the need for information by the public continues to increase [27-30]. Since social media is more frequently used than other media sources to acquire information on COVID-19, the government should have guidelines in place to ensure that reliable and accurate digital health information reaches diverse populations and those with lower levels of education.

Our findings show that COVID-19 symptoms represented 6% of social media expressions. People shared and sought validation for their symptoms, possibly reflecting anxiety, uncertainty, or a need for peer assurance. Posts related to symptoms ranged from less severe, like a cold, to potentially more severe, such as high fever, difficulty in breathing, and fatigue. Searches related to symptoms were closely linked to severity and age. In another study, a survey of over 2000 adults highlighted the awareness of main symptoms but not of others including skin rash, muscle and body aches, diarrhea, headache, nausea, and vomiting. Younger age groups in particular were least aware of COVID symptoms [31]. This lack of awareness may have led to increased anxiety and fear among the public, and to some extent a positive change in people’s health behaviors toward preventive measures [32].

Discussions about testing were common. The sentiment toward testing was neutral through the pandemic but turned positive from July 2021 and then became negative in the United Kingdom in March 2022. Diagnostic tests played a crucial role in COVID-19 management worldwide [33]. The public encountered challenges in accessing tests for diagnosis of COVID-19 infection particularly in the earlier days of the pandemic. Public frustration with testing delays was further aggravated as the pandemic progressed and test sites became overwhelmed. Major challenges encountered by health authorities included addressing the ever-increasing demand for testing and real-time PCR limitations (including being resource-intensive and time-consuming), which may have been the main source of public frustration with the test. Despite anxiety with regards to testing, social media activity eased during mid-2021, and public sentiment to COVID-19 testing also started turning positive. This aligned with the second phase in the evolution of the COVID-19 testing strategy which occurred several months into the pandemic, making diagnosis more affordable, easier, and faster than molecular tests [34].

The increase in negative sentiment in March 2022 could have been due to stopping free rapid COVID-19 tests in the United Kingdom. Ending free access to COVID-19 rapid tests could have contributed to health inequality, disproportionately affecting low-income and marginalized communities who may struggle to afford testing. This has the possibility of creating barriers to early detection and isolation, increasing the risk of transmission, particularly in high-risk settings [35]. Ensuring affordable, widespread testing is crucial for early intervention and outbreak control. Investing in community-based diagnostics, digital health solutions, and stronger public health data integration can improve disease monitoring and promote health equity, supporting better preparedness for future crises.

The diagnostic-related posts were less frequent and mostly involved sharing test results, which may reflect the normalization of COVID-19 infection over time. As COVID-19 became an ongoing part of daily life, public concern around diagnosis likely diminished, especially after the removal of widespread testing programs. With fewer policy changes and declining case severity, individuals who may have viewed COVID-19 as less urgent led to reduced engagement in discussions about diagnostic tools. This shift suggests that as disease becomes endemic, public interest in testing decreases, which has implications for maintaining awareness and ensuring continued access to diagnostic resources for effective disease monitoring.

Our analyses showed a large volume of queries about COVID-19 treatment, including its availability, effectiveness, and eligibility criteria. Initially, the public had a negative attitude toward possible side effects of antiviral treatments for COVID-19, and “rebound” was found to be the main concern for patients. Patients who tested positive for COVID-19 were also anxious regarding their eligibility to receive the available treatment. This concern was highlighted in the discussions due to the limited availability of the antiviral drugs, particularly for patients with mild-to-moderate symptoms. Difficulty accessing treatment and lack of access to information about treatment were also documented elsewhere as being particularly challenging for vulnerable communities in the United Kingdom throughout the pandemic, which may have influenced these sentiments [36].

Our findings align with previous studies showing that social media trends mirror Google search behavior [37]. Similar studies have confirmed that digital platforms effectively capture real-time public interest in health topics [38-41]. However, social media also contributed to misinformation, highlighting the importance of public health communication strategies [42]. Of note, the UK government and WHO had collaborative campaigns and created a series of social media infographics and messages to explain the safety profile of COVID-19 vaccines [43]. In addition, the British Broadcasting Company launched a successful project called “Stop the Spread” during May and June 2020; the purpose of which was to increase the public’s awareness of the magnitude of COVID-19 misinformation and motivate them to limit the spread and associated consequences.

One key limitation of this study is the demographic representation of social media users. Twitter data do not fully reflect the general population, and user characteristics, such as age, gender, and location could not be inferred. Although some geographic analysis was conducted, the findings did not show significant differences and were therefore not included in the study. Another limitation is the observational nature of the study. The comparison between social media trends and Google searches is descriptive rather than a statistical correlation analysis. To ensure clarity, we explicitly state this in the study to avoid any causal implications. Finally, there are constraints related to the data source. The use of Synthesio, a proprietary tool, limits the reproducibility of the study. In addition, while efforts were made to filter noise and spam, some irrelevant content may persist due to the inherent limitations of natural language processing techniques.

The nature of internet-based conversation about COVID-19 is rapidly changing, providing insights regarding the public’s thoughts and needs. Governments and public health authorities can leverage these insights use these insights to enhance crisis communication, combat misinformation, and promote vaccine uptake. By monitoring social media discussions, policy makers can identify emerging concerns and misconceptions, allowing them to tailor public health messaging accordingly. For example, during COVID-19 peaks, public uncertainty about symptoms, testing, and treatments could have been addressed through targeted social media campaigns to provide accurate information and alleviate confusion.

In addition, our findings highlight the role of social media in both spreading reliable information and amplifying misinformation. Health authorities should proactively track trending misinformation, implement fact-checking initiatives, and collaborate with platforms to promote credible sources. The success of initiatives like the WHO’s “Stop the Spread” campaign underscores the potential impact of coordinated efforts to counteract health misinformation and enhance public understanding.

Social media insights can also help inform vaccine communication strategies by identifying public concerns and hesitations regarding treatments and vaccinations. By analyzing discussions and trends, policy makers can develop targeted campaigns that address specific fears, correct misunderstandings, and promote informed decision-making. Integrating social media insights with traditional public health surveillance systems can further strengthen early detection of emerging health threats. By leveraging these real-time insights, governments can proactively engage with the public, reduce uncertainty, and build trust during health crises.

Future research could also build upon and deepen the types of analysis we conducted for this study. For example, a more comprehensive sentiment analysis could be conducted delving into a deeper range of emotions beyond the positive, neutral, and negative categorization used in this research [44,45], or further exploring how the different proportions of post related to each journey theme changed over time in line with previous studies [45,46]. We also note that while we did not analyze social media platform content individually against Google activity, future similar types of research could further investigate differences by social media platforms, Google activity, or both to clarify which platforms bear the most similarities in trends, particularly regarding conversations around health and health emergencies.

Throughout our social listening analysis on several social media platforms, we had a promising overview of patients with COVID-19 journey from diagnosis to treatment. The study highlights the significant role of social media in shaping public perceptions of COVID-19 and underscores its potential as a tool for public health interventions. By leveraging social media analytics, policy makers can gain valuable real-time insights into public sentiment, allowing for more effective crisis communication, misinformation management, and targeted health campaigns. Future public health strategies should incorporate social listening tools to identify trends, address misinformation, and foster trust through clear and responsive communication.