YouTube: recommendations that expose users to pro-tobacco and nicotine content

A recent study conducted by Truth Initiative[1]-[2] This study highlights the structuring role of YouTube's recommendation algorithm in exposing users to tobacco and nicotine-related content. While the platform is a major source of information, particularly for young people, the results show that recommendations can steer viewing paths toward pro-tobacco content and content promoting other nicotine products, even after viewing prevention materials. Based on the analysis of several thousand videos, the study reveals that these algorithmic dynamics promote the dissemination of discourse favorable to nicotine products and generate significantly higher potential exposure to pro-tobacco and other nicotine product content than to public health content.

The study is based on a quantitative analysis of YouTube's recommendation system, focusing on four main themes: e-cigarettes, nicotine, cigarillos, and nicotine pouches. Using 66 queries from Google Trends, researchers collected 671 seed videos and then analyzed the associated recommendations using an automated tool provided by Google (API). This protocol was repeated five times, resulting in the analysis of 2,560 unique videos and over 5,000 combinations of seed videos and recommendations. The content was then coded according to its positioning (pro-tobacco/nicotine, anti-tobacco (prevention), neutral) and its source (media, public health, self-proclaimed experts, organic content).

Increased exposure to pro-nicotine content in an algorithmic environment central to young people

YouTube now plays a structuring role in information practices, particularly among young audiences. Nearly 90% of users report searching for health information on the platform, reinforcing its role as a source of information on tobacco and nicotine products.

In this context, algorithmic recommendation mechanisms play a crucial role: over 70% of videos viewed on YouTube come from automated recommendations. The algorithm thus acts as a true information intermediary, structuring browsing paths and directly influencing users' exposure to content.

The analysis of viewing trajectories highlights the concrete effects of these logics. Out of 5,182 combinations analyzed, 77.1% of the paths do not contain tobacco-related content, but nearly a quarter of the trajectories expose viewers to such content, including 11.9% of pro-tobacco and other nicotine product content and 11% of anti-tobacco and nicotine product prevention content.

The study primarily shows that user pathways are not rigidly separated: 13.3% of recommendations associated with smoking/nicotine prevention content lead to pro-tobacco/nicotine content, compared to only 10.4% in the opposite direction. This asymmetry indicates that users are more likely to switch from prevention content to content favorable to tobacco and other nicotine products than the other way around.

These dynamics reflect a lack of separation between informational registers, contributing to the confusion of public health messages. They also demonstrate the capacity of algorithmic systems to create hybrid informational environments, in which prevention content can be quickly overshadowed, or even neutralized, by promotional or pro-tobacco and nicotine product discourse.

Increased visibility of pro-tobacco content driven by certain news sources

Estimates based on recommendation mechanisms and click-through rates confirm this exposure asymmetry. Between 1.9 and 3.2 million users could be redirected to pro-tobacco and nicotine product content after viewing prevention content, compared to 1.5 to 2.4 million in the opposite direction. More broadly, recommendations could generate between 14.6 and 23.3 million views for content favorable to tobacco and nicotine products, compared to 7.7 to 12.9 million for prevention content.

The study also highlights the structuring role of certain types of actors in these dynamics. Content from public health sources is very poorly represented in the recommendations associated with pro-tobacco content and other nicotine products (6.3%). Conversely, these recommendations are mainly composed of content produced by individual users (35.9%), the media (29.7%), and self-proclaimed medical experts (28.1%).

These individuals play a particularly concerning role: when prevention content leads to pro-tobacco/nicotine recommendations, 81.6% of these videos are produced by people presenting themselves as medical or public health experts. The study also highlights that a few creators can have a disproportionate impact, with a single podcaster responsible for more than 95% of some pro-tobacco/nicotine recommendations stemming from this type of content.

A growing challenge for public health policies in the digital age

These results highlight the limitations of traditional prevention approaches in digital environments dominated by algorithmic logic. The production of public health content does not guarantee its visibility or effectiveness, especially when integrated into systems where recommendations can direct users toward contradictory content.

They emphasize the need for better regulation of recommendation systems, increased transparency of platforms, and restrictions on the dissemination of promotional or ambiguous content related to tobacco and nicotine products. More broadly, these dynamics fall within the framework of the commercial determinants of health, where digital environments play a significant role in exposing populations to influence strategies.

©Generation Without Tobacco

AE

[1] YouTube recommendations expose viewers to pro-tobacco content, Truth Initiative, published April 14, 2026, accessed April 16, 2026

[2] George DH Pearson, Nathan A Silver, Kristiann Koris, Tatum McKay, Barbara A Schillo, Jessica M Rath, Examining the Impact of YouTube's Video Recommendation Algorithm on Pro- or Anti-Tobacco Messaging, Nicotine & Tobacco Research, Volume 28, Issue 4, April 2026, Pages 642–649, https://doi.org/10.1093/ntr/ntaf218