User Interface Designs to Address Cancer Misinformation: A Randomized Clinical Trial

Misinformation about false and potentially harmful cancer treatments is common on social media. Strategies are needed to encourage reporting to reduce exposure to inaccurate treatment claims.

To evaluate the impact of 2 user interface designs (prompts on posts) on intervening and sharing cancer misinformation on social media.

This 3-group, parallel-group, randomized clinical trial assessed outcomes at 5 time points from March 31 to May 12, 2025. US adults (ages 18-65 years) who had a cancer diagnosis or identified as a cancer caregiver and used social media were recruited from Prolific. Participants interacted with cancer misinformation posts on a simulated realistic social media platform. Data were analyzed from May 13, 2025, to July 27, 2026.

Participants were randomized to cancer misinformation posts shown with social cue-enhanced prompts that specify others' actions and a review policy, standard prompts, or no prompts (control).

The primary outcome was a participant flagging a post to intervene. Primary and secondary outcomes included recorded social media behavior (flagging, disliking, liking) and self-reported willingness to flag, dislike, like, comment, and share about the posts. Intervening outside the study platform was also assessed. A separate generalized estimating equations model was used for each outcome.

A total of 294 individuals (mean [SD] age, 39.3 [11.7] years; 179 [62%] women) with cancer or who were caregivers for individuals with cancer were randomized, with 88 participants receiving social cue prompts, 92 receiving standard prompts, and 114 receiving no prompts. Posts with inaccurate claims were flagged 2 to 3 times more often with social cue (224 of 592 posts [38%]) and standard (180 of 689 posts [26%]) prompts than those without prompts (84 of 675 posts [12%]) (social cue vs control: rate ratio [RR], 3.12 [95% CI, 1.79-5.42]; standard vs control: RR, 2.15 [95% CI, 1.20-3.88]). Prompts led to less sharing behavior (ie, liking: 20 posts [3%] with social cue prompts; 10 posts [1%] with standard prompts) compared with posts without prompts (65 posts [10%]) (social cue vs control: RR, 0.07 [95% CI, 0.04-0.15]; standard vs control: RR, 0.13 [95% CI, 0.04-0.42]). Prompts also led to greater willingness to flag and dislike (intervening) and less willingness to like and share compared with no prompt. Standard prompts led to more flagging on existing social media (outside the study platform) than no prompts.

In this randomized clinical trial of user interface designs, simple prompts encouraged responsible behavior online. Prompts on posts led to more intervening and less sharing of cancer misinformation.

ClinicalTrials.gov Identifier: NCT07588607.