Forced Feed (2026)

Forced Feed 2026 is a national UK report exploring how social media algorithms shape what users see and how much control they have, or want, over their feeds. The report combines two complementary research approaches to give a complete picture of both algorithmic behavior and public attitudes.

First, a structured behavioral exposure study, led by Cybersmile volunteers and behavioral scientists, examined how TikTok and Instagram algorithms deliver content in real time. The study revealed how quickly harmful material can appear in users' feeds—even without searching for it—highlighting the risks inherent in algorithm-driven content delivery.

Second, a nationally representative survey of 2,000 participants explored public understanding of social media algorithms, experiences of content exposure, and demand for more control over personal feeds. The findings showed that many users feel unaware of how algorithms decide what they see, while a majority want greater insight and agency over their online experiences.

By combining these two approaches, Forced Feed 2026 provides a dual perspective: the real-world risks of algorithmic content delivery and the public's desire for transparency and control. Together, these insights inform Cybersmile's PROTECT framework, which recommends ways to empower users, promote safer online spaces, and ensure that people, not algorithms, decide what they consume online.

Behavioral Exposure Study

The study found concerning levels of harmful content being fed to users within minutes. It took as little as 16 seconds for adults and 8 minutes for children to be served harmful content without their consent.

Racism, misogyny, and violence were the most common themes observed in the study, with a wide range of graphic and harmful content being served to users. This included hateful speech towards different groups and communities, content depicting and encouraging extreme violence, suicidal ideation, content demeaning people with disabilities, and content encouraging dangerous behaviors.

Key study findings include:

The volume of harmful content increased over time. By day three, nearly one in five posts on children's accounts and more than one in three on adult accounts contained harmful material. Within three days of testing:

• 90% of the accounts were exposed to racist content
• 60% of the accounts were exposed to misogyny
• 60% of the accounts were exposed to violence

Adult accounts were exposed to more harmful content overall – ranging from 3–34% on day one to 5–38% on day three. Child accounts were exposed to less extreme content overall, but the volume ranged from 0–12% on day one to 1–18% on day three.

Nationally Representative Survey

• 65% would like more control over their algorithm and the content they see
• 58% use social media every day
• 40% do not understand algorithms
• 36% feel in control of their algorithm and the content they see
• 69% would like more insight into how the algorithm decides what they see

Forced Feed 2026 used two complementary research approaches to understand both the behavior of social media algorithms and public attitudes toward algorithmic control.

Behavioral Exposure Study

Cybersmile volunteers, all aged over eighteen and guided by behavioral scientists, conducted a structured experiment on TikTok and Instagram to observe real-world algorithmic content delivery. Adult participants created new adult and child accounts on factory-reset devices to ensure a neutral starting point. The study included:

• Accounts: 8 total accounts — 4 child accounts (aged 15; 2 male, 2 female) and 4 adult accounts (ages 25–45; 1 male, 3 female).
• Procedure: Participants scrolled TikTok and Instagram Reels for 45 minutes per day over three days, screen-recording the content shown to them. They did not like or comment on videos but watched harmful content 2–3 times if it appeared.
• Data Collected: Time to exposure to harmful content, the types of harmful content, and the proportion of content that was harmful each day.
• Content Categories: Harmful material included racism, misogyny, violence, suicidal ideation, content demeaning people with disabilities, and content encouraging dangerous behaviors, categorized based on Ofcom's priority harms in the Online Safety Act.

This study provided insight into how quickly and frequently harmful content can appear in users' feeds without active searching, demonstrating potential risks in algorithmic content delivery.

Note: No children participated in this experiment. All volunteers were over eighteen years of age. Any child accounts created as part of the experiment were created by adults posing as children.

Nationally Representative Survey

To complement the behavioral study, we commissioned a national survey conducted by Censuswide, involving 2,000 UK participants aged 16+. The survey explored:

• Public understanding of social media algorithms
• Experiences of content exposure
• Demand for control and personalization of algorithmic feeds

The research was conducted by Censuswide with 2,000 nationally representative consumers (16+) between 10.27.2025 and 10.29.2025. Censuswide abides by and employs members of the Market Research Society, which is based on the ESOMAR principles, and is a member of the British Polling Council.

The Cybersmile Foundation (2026). Forced Feed Report (2026): A national UK report exploring social media algorithms.