On January 5, 2024, a search was performed on YouTube for semaglutide videos in English. Using Google Chrome with Incognito mode, YouTube was searched with the words semaglutide, Ozempic, Wegovy, and Rybelsus, respectively. For each search word, the first 30 full-length videos (videos without a time limit) and 30 shorts (videos that are no longer than 1 minute) resulting from the search, sorted by relevance, were recorded. The number of 30 videos was chosen according to a recent study, which claimed that very few YouTube users searched beyond the 33rd video [27]. After discounting duplicates and excluding unsuitable videos, a total of 96 full-length videos and 93 shorts were analyzed.

We recorded the basic video metrics, such as the duration, number of views, number of comments, number of likes, number of channel subscribers, and the age of the video since upload (number of days until March 14, 2024). The readability and sentiment of the video transcripts were assessed. The quality of video content was evaluated. Further details are described in further sections.

To evaluate the readability and overall sentiment, video transcripts were generated and analyzed. To evaluate the quality of video content, the entire videos were watched and analyzed.

The video transcripts were generated by Whisper (with the “large” version, edition 20231117), an artificial intelligence automatic speech recognition system developed by OpenAI [28]. It has been used in previous research to transcribe educational videos [29] and had the best performance compared with similar automatic transcription tools [30]. The readability of the transcripts was evaluated by the Flesch Reading Ease (FRE) score [31], calculated via an online platform (Readability Formulas website). In brief, the score ranged from 0 to 100, with 90 to 100 being very easy, 0 to 29 being very difficult, and 60 to 69 being standard. Meanwhile, the sentiment of the transcripts was evaluated by ChatGPT 3.5, an artificial intelligence large language model developed by OpenAI that has been demonstrated to be very effective in sentiment analysis across multiple languages [32]. Referring to the method by Fu et al [32], the prompt was set as “Is the sentiment of this text positive, neutral, or negative? Respond with the sentiment label only.” The “temperature” of the ChatGPT model was set at 0 to ensure the consistency of the answers with the least creativity. Temperature is a variable that changes the degree of randomness of the output generated by the model [33].

Next, the quality of video content was evaluated by 3 tools, Global Quality Score (GQS) [34], Modified DISCERN [35], and a custom checklist. Manual evaluations were independently performed by two authors (AWKY and AGA). Disagreements were resolved through mutual discussion. During these evaluations, the overall audiovisual content of the videos, not merely limited to verbal narrative, was examined. The GQS is a 5-point Likert scale designed to evaluate online health information. A score of 1 means “poor quality, poor flow, most information missing, not at all useful for patients,” whereas a score of 5 means excellence and high usefulness for patients [34]. Meanwhile, the Modified DISCERN contains 5 evaluative items and gives 1 point for every positive answer and 0 points for negative answer. It was designed to evaluate YouTube videos on health care information [35]. The 5 items are as follows: (1) Are the aims clear and achieved? (2) Are reliable sources of information used? (3) Is the information presented balanced and unbiased? (4) Are additional sources of information listed for patient reference? (5) Are areas of uncertainty mentioned? For readers’ information, the Modified DISCERN is based on an original version of DISCERN, which was designed to evaluate written health information and used a 5-point Likert scale to answer 15 questions plus an overall rating [36]. Since GQS and Modified DISCERN could only give a more general evaluation of the videos, a custom 12-point checklist was devised to evaluate the video content based on some specific aspects of side effects and safety concerns related specifically to semaglutide.

The custom 12-point checklist was compiled by the authors’ team with reference to Smits and Van Raalte [37]. It recorded whether the following contents were mentioned in the videos or not: (1) form of application (injection, exception: oral for Rybelsus); (2) safe dosage; (3) need for long-term usage, or change in lifestyle and eating habits to avoid rebound back to original weight after drug cessation; (4) serious side effects (eg, retinopathy and pancreatitis); (5) gastrointestinal symptoms (eg, nausea, diarrhea, vomiting, gastric reflux, and gastritis); (6) prevalence or frequency of such side effects; (7) increased risk of aspiration during the induction of anesthesia; (8) contraindications; (9) long half-life (7 days) so that potential side effects persist for multiple days after drug cessation; (10) lack of long-term data; (11) potential alternatives (eg, diet and bariatric surgery); and (12) risk of counterfeit drugs.

Following descriptive analysis of the video contents, 2-sample t tests were performed were performed to evaluate if there were significant differences between the full-length videos and shorts in terms of the mean FRE, GQS, and Modified DISCERN scores, as well as the mean number of mentioned points from the custom checklist. To supplement, the same tests were performed among the full-length videos, to evaluate if there were differences between those uploaded by YouTube-verified health source channels and those without this verification.

Ethical approval was not applicable, as this study only analyzed publicly available data from existing datasets, and results were presented in aggregate that did not contain any identifiable information.

This study found that semaglutide videos on YouTube have reached a broad audience. The 189 analyzed videos had a total view count of over 45 million. For comparison, the 98 most viewed YouTube videos on diabetic retinopathy were collectively viewed 1 million times [38]. It implied that these videos might potentially affect the perception and even health care decisions of the general public regarding the use of semaglutide. As expected, full-length videos were generally more informative than shorts. For patient education, shorts would serve better to grab the attention of the patients or make them aware of 1 or 2 particular issues related to semaglutide; whereas selected full-length videos might be more suitable to be incorporated as part of a panel discussion or public forum.

Alarmingly, the videos seldom covered the risk of counterfeit drugs. For instance, the antidiabetic drug, Ozempic, contains semaglutide that is synthesized by yeast fermentation and subsequent synthetic modification [39]. Without proper quality control of the synthetic processes, some falsified Ozempic products were found to contain contaminants such as glass particles and filler substances [39]. Meanwhile, some pharmacies would produce compounded versions of semaglutide to circumvent the patent issue. The high demand, high cost, and limited supply have led to a period of time when some patients and drug providers switched to compounded semaglutide [10]. Subsequently, the FDA stated that the compounded drugs, such as semaglutide sodium and semaglutide acetate, do not have their approval due to lack of testing, may not possess the same drug effect as semaglutide, and may even cause adverse effects (unspecified) [40,41]. Another safety issue of using compounded semaglutide is an increased risk of overdose due to suboptimal drug packaging. A recent case series reported that compounded semaglutide might be dispensed in vials instead of prefilled manufactured injection pens such as those by Ozempic and Wegovy [42]. Vials that contain large volumes of semaglutide and vials dispensed together with subpar syringes might allow for overdose much more easily during self-administration [42]. There seemed to be yet a case series on the overdose of oral semaglutide, but it would be reasoned that compounded semaglutide in the form of tablets for oral intake could have the same increased risk if each tablet did not conform to the amount of semaglutide contained in approved branded semaglutide products such as Rybelsus.

Another not uncommon risk of taking semaglutide that was often neglected by the videos was the risk of aspiration during the induction of anesthesia. One effect of semaglutide is gastroparesis, that is, reduced bowel motility and gastric emptying without any physical obstruction. This would increase the gastric volume and increase the risk of regurgitation and pulmonary aspiration of gastric contents even with the usual recommended fasting time before anesthesia [43,44]. According to a recent study, aspiration of gastric contents accounted for 5% of closed anesthesia malpractice claims in the United States during 2000-2014 [45]. Among these claim cases related to aspiration, 57% (66/115) of patients died and another 14% (16/115) suffered from permanent severe injury [45]. It implied that aspiration during anesthesia was not uncommon, and it could lead to very serious consequences. While medical organizations develop recommendations regarding the use of GLP-1 receptor agonists before operations, the optimal approach to patient data management is still being specified [46,47]. Hence, it is important for researchers and clinicians to conduct subsequent studies to optimize the fasting time and airway management strategy for patients on semaglutide who need to undergo anesthesia.

Although there have been many clinical trials on the efficacy and safety of semaglutide, most of them (if not all) had a study period of up to 2 years (104 weeks) only, such as the STEP, SUSTAIN, and PIONEER trials [37,48]. Since the use of semaglutide could last beyond 2 years and there could be a possibility of life-long usage, a lack of long-term data could mean potential risks yet to be elucidated, such as the discovery of more side effects, and potential development of drug dependence or drug resistance [49,50]. The COVID-19 vaccines may illustrate the situation of lack of long-term data. After the COVID-19 pandemic began near the end of 2019, pharmaceutical companies put huge efforts into creating vaccines that could lower the infection rate and reduce the symptoms or morbidity. By the end of 2020, at least 10 vaccines have already been introduced into the global market and authorized by governments for emergency use [51]. After several years in use, millions of people vaccinated and billions of doses administered, data have accumulated. By analyzing the retrospective data, researchers have recently found 2 new rare but potentially severe side effects of COVID-19 vaccines, namely acute disseminated encephalomyelitis and transverse myelitis [52]. This new finding may influence the decision-making of some people in the public on whether they should be vaccinated, take a booster, or choose which type of available vaccines to use. In the case of semaglutide, more data besides its weight loss and antidiabetic properties would facilitate better-informed decisions from clinicians and patients. For instance, there were cases reported on the development or recurrence of depression 1 month after taking semaglutide, which was subsequently relieved after discontinuing the drug [53]. On the other hand, a retrospective study that covered over 1.5 million patient records reported that semaglutide users had a lower risk of incident and recurrent suicidal ideation [54]. At the same time, an analysis of over 40,000 user comments posted on social media platforms has found that users of GLP-1 receptor agonists (including semaglutide) felt that the drugs have mixed effects on their mood, anxiety, and insomnia conditions [14]. Patients should be aware of the fact that many weight-loss drugs, mainly appetite suppressants, were withdrawn from the market in the past due to adverse drug reactions [55]. Therefore, in the future when long-term data become available, the safety and side effects of semaglutide could be better assessed and established.

Findings from this study echoed previous studies on Reddit content on Ozempic or semaglutide, in the sense that there is generally a lack of discussion or elaboration on the potential health risks and hazards associated with the use of Ozempic, not to mention its off-label use [12,13]. However, the risks and hazards indeed exist, such as concerns with depression and anxiety raised by social media users on YouTube and TikTok [14]. Therefore, actionable recommendations included better public health awareness campaigns to educate the public on the proper use of Ozempic or semaglutide including the potential side effects and how to manage them. Pharmaceutical companies, governments, and health authorities can organize exhibitions in shopping malls, provide easy-to-understand information on their web pages, and develop user-friendly mobile apps to engage members of the public who are interested. Future research should evaluate the effects of such activities on the knowledge level of the public on semaglutide.

This study has several limitations. First, only the first 30 full-length videos and 30 shorts were initially screened for each search word, rendering a final analysis of 189 videos. This might only represent a small proportion of the entire relevant video collection. Second, only videos in English were analyzed. The themes and foci could be different for videos produced in other languages. Third, errors might exist during computational and manual evaluations. Besides, there was not a detailed analysis of video sources, which might further enhance the findings, for example, whether videos produced by clinicians were more informative than those from general influencers. It was said that there was a tendency for the YouTube algorithm to place more “reliable” videos at the top of the search results [56], hence the presented results in this study might be biased.

The 189 analyzed YouTube videos on semaglutide have attracted more than 45 million views. Full-length videos were much more informative than shorts in terms of side effects and safety issues. The analyzed videos lacked coverage of several important aspects, including the increased risk of aspiration during the induction of anesthesia, the persistence of side effects due to the long half-life of semaglutide, the risk of counterfeit drugs, and the lack of long-term data. Patients should be aware that these videos may not be comprehensive enough even if they were uploaded by YouTube-verified health source channels, and hence cannot replace the consultation from the physician, who may make tailor-made recommendations and treatment plans for each patient.