This study involved a retrospective review of publicly accessible YouTube videos and did not involve human participants or identifiable private information. As such, formal ethical approval was not required. This was confirmed by the Imperial College London ethics panel. All data used in the study were publicly available and did not contain any personally identifiable information. Videos were analyzed anonymously, and no attempts were made to contact content creators or link content to specific individuals.

To ensure relevancy of the search strategy for medical cannabis, key terms were derived from an assessment of worldwide Google Trends (Google LLC) data from the 12 months preceding July 8, 2024, to ensure the highest coverage of cannabis-based medicinal products. These included ‘medical cannabis,’ ‘medical marijuana,’ ‘medical weed’ and ‘prescription cannabis.’ These search terms were then entered into the United Kingdom version of YouTube (Alphabet Inc) on July 8, 2024, using incognito mode to prevent third-party cookies and browsing history from affecting the results.

The top 200 videos of each search term were noted using Microsoft Excel spreadsheet with the key attributes: title, number of views, country of origin, duration, age, channel name, number of likes, and number of comments. Exclusion criteria included videos shorter than one minute, older than five years (based on upload date), duplicate videos, not in English, or unrelated to cannabis-based medicinal products, such as those not addressing their biochemistry, effects or social topics.

Data extraction was performed by two independent reviewers (SK and SE), with disagreements planned to be resolved by a third author (SC). Data extraction included video views, likes, channel, and country of origin. The videos were subsequently categorized into the following three categories based on the channel types and purpose: educational channels organized by nonmedical professionals (eg, TED or TEDx Talks); education channels organized by medical professionals (eg, Mayo Clinic or Curaleaf Clinic); and independent nonmedical users and groups (eg, independent content creators).

The primary outcome was the quality and reliability of the video content. This was determined using the DISCERN Grade and Health on the Net (HON) code. These tools have been used previously to assess health care information based on social media influence [14]. The DISCERN Grade consists of 16 questions, with each question scored on a scale of 1 to 5 points, therefore making a total of 80 points. It is divided into three sections, with the first section evaluating the reliability of the paper, the second assessing the quality of information on treatment choices and the final section examining the overall quality of the publication [15]. All sections of the DISCERN grade are included in the sum score. The HON code is used to ensure health and medical websites maintain honesty and serve as reliable sources of information backed up by evidence. There are eight key principles, and each video is scored on a scale from zero to eight, where zero indicates noncompliance, and eight represents full compliance [16]. The videos were assessed independently by SK and SE. Any discrepancies were planned to be resolved through discussion with a senior author (SC). Secondary outcomes included assessment of video reach and engagement through video views and likes.

The YouTube engagement metrics were summarized using median (IQR), while DISCERN grade scores, and HON code scores were summarized using mean (SD). Statistical analysis was performed using PRISM (version 10; GraphPad software). One-way ANOVA was conducted to compare DISCERN and HON scores between different content creators assuming unequal variances.

Pairwise comparisons were conducted using Tukey test to identify specific differences between groups. The Tukey method was applied for multiple comparisons to control for Type I error and was conducted when a statistically significant finding was present on ANOVA. Pearson correlation coefficient was used to determine the association between the DISCERN Grade and the number of views. Statistical significance was defined as P value <.05.

This study aimed to evaluate the quality of information presented in YouTube videos about medical cannabis and assess whether popularity correlates with content reliability. The findings reveal a notable disconnect between viewer engagement and video quality, with independent users gathering the highest views but offering the least reliable content, as indicated by lower DISCERN and HON scores. These results emphasize the persistent challenge of misinformation on digital platforms, even in an era of increasing regulation.

The overall mean DISCERN score across all videos was 34.63, which, on the 80-point scale, suggests generally poor reliability. In comparison, other studies examining health topics such as varicoceles and testicular torsion exhibited similar concerns [17,18]. The analysis by Hong et al [18] expressed frustration as from the nonduplicated videos, an overwhelming majority were expressed to be educational but were rated as poor on the DISCERN score. A comparable pattern was observed in the present analysis with only 12.2% of the videos originating from medical or expert sources, with the overall mean DISCERN score. This suggests that misinformation about medical cannabis is not an isolated issue but part of a broader trend of unreliable health communication online.

A noticeable finding in this study was videos from medical education sources were less popular and had lower quality and reliability scores in comparison to nonmedical educational sources. This is a challenge also seen in other areas of health content on YouTube [19]. One reason could be medical sources often use technical language and longer video formats suited for academic purposes but less engaging to the public. In contrast, nonmedical creators may adopt language and entertaining visuals, which enhances the relatability and shareability for lay audiences. Moreover, while the quality and reliability of medical videos was higher than independent creators, it was lower than that of the nonmedical educational content. Many of the nonmedical sources were from news organizations; therefore, the introduction of clear editorial policies for medical channels may help further enhance their quality. While independent creators received lower DISCERN and HON code scores, they had higher engagement. Evidence demonstrates that subjective language and content that falls under entertainment categories are more likely to receive a higher number of views and likes [20]. In addition, posting outside of business hours is more likely to receive higher engagement. Independent creators are less bound by professional standards of objectivity and are also more likely to focus on changing posting frequency and timings to maximize video engagement. YouTube has introduced initiatives to try and help improve the reach of content produced by accredited health care professionals. It is also important that this content continues to be as engaging and meets viewers at their level for effective dissemination of health information on this particular social media platform [21].

Furthermore, a study on COVID-19 vaccine videos found that approximately 11% of the most-viewed videos contradicted the reference standards from the WHO or the Centers for Disease Control and Prevention. These videos had significantly lower modified DISCERN (mDISCERN) scores compared to factual ones, yet they garnered far more likes and engagement [22]. Similarly, this analysis identified several of the most viewed videos were of cannabis content published by nonexpert independent users with low reliability.

The weak positive correlation between engagement metrics (views and likes) and DISCERN scores (r=0.34 and r=0.30, respectively) further confirms that popular videos are not necessarily trustworthy. Moreover, visual analysis of correlation plots highlights that the r value is likely impacted by several outlier videos. This is consistent with research by Stimpson and Ortega [23], which found that a large proportion of social media users (82%) perceive false or misleading health information to be prevalent on these platforms, with 67% admitting they are unable to assess the accuracy of the information they encounter [23]. This increases the risk of misinformed health decisions by patients and the general public, but may also increase anxiety among viewers [24]. Videos that over illustrate the therapeutic benefits of cannabis while downplaying risks, such as adverse effects or drug interactions, may lead individuals to experiment with unproven treatments without medical guidance. For example, some cancer patients believe cannabis can treat their cancer directly, despite a lack of scientific evidence, potentially leading to worse outcomes if they prioritize unproven cannabis use over chemotherapy or immunotherapy [25].

YouTube enforces strict policies on cannabis-related content, prohibiting videos that display or promote the consumption of cannabis and typically permitting educational, documentary, or news-focused discussions if they do not encourage use or provide instructions for consumption [26]. While YouTube’s approach to misinformation prohibits health claims that contradict recognized authorities and removes content with serious risk of harm, the platform does not systematically check creators for medical credentials or require source acknowledgment, resulting in only partial alignment with the more stringent transparency and authority requirements of the HON code.

To combat the spread of misinformation, the WHO is working with the Council of Medical Specialty Societies and the National Academy of Medicine on Phase II of a project on identifying credible sources of health information on social media [27]. The project aims to provide social media platforms with standardized criteria to distinguish reliable health information, addressing challenges like inconsistent vetting processes and the proliferation of unverified claims. However, systemic issues like algorithmic bias and regulatory inertia continue to remain unresolved.

Several limitations of this study should be acknowledged. First, the video selection was restricted to English-language content, which limits the generalisability of the findings to non-English speaking populations and other geographic regions. Additionally, while the DISCERN tool provides a structured framework for assessing information quality, it allows for subjective interpretation, particularly when assigning partial scores between 2 and 4. This introduces some variability in how each criterion is evaluated. The study also did not consider channel-level data or account characteristics, which may have implications for both the reach and perceived credibility of the videos hosted on that channel. Lastly, this study focused solely on YouTube and did not examine other social media platforms such as Instagram, TikTok, or Facebook, which are also social media sites where individuals may be exposed to unverified health information.

In conclusion, the quality of information about medical cannabis among the most viewed videos on YouTube remains poor. This mirrors similar patterns observed during the COVID-19 pandemic and in studies of other health topics. This study contributes to a growing body of evidence suggesting that user engagement does not strongly correlate with informational quality, posing ongoing challenges for public health communication. In the future, it is important YouTube and other social media sites implement policy changes to mitigate the spread of misinformation in all fields, particularly health, and to regulate the content by health care professionals more frequently. Medical professionals should also be encouraged to engage more actively on social platforms and implement strategies to make their content engaging as well as informative to compete with low-quality content.