In a retrospective study of 16,556 adults prescribed semaglutide in an Australian digital weight loss service, clinical and behavioral factors predicted 6-month GLP-1 retention better than an AI support bot: adherence was 53.2% post-Junebot vs 47.3% pre-Junebot, but the post-Junebot era was independently linked to higher attrition (OR 1.178; 95% CI 1.052-1.318). The study found that non-automation factors dominated retention, including high month-1 tracking (>25 tracks; OR 15.753 for attrition), program pauses, and higher costs, while moderate digital engagement correlated with better persistence after matching.

Why It Matters To Your Practice

• Adding conversational AI alone may not improve GLP-1 adherence in real-world Obesity care.

• Retention appears more sensitive to practical and behavioral signals than to baseline AI support availability.

• For clinicians using digital programs, early patient behavior may help predict who is at risk of dropping off therapy by 6 months.

Clinical Implications

• Watch for early high-intensity tracking behavior, which may reflect anxiety or friction rather than healthy engagement.

• Ask about affordability and treatment interruptions, since cost and program pauses were stronger predictors of attrition.

• Consider proactive outreach in the first month for patients showing risk signals, rather than relying on automated support alone.

• Moderate, active engagement may be a more useful target than maximal app activity.

Insights

• The analysis included multivariate logistic regression and three 1:1 propensity score-matched sensitivity analyses.

• Although crude adherence improved after Junebot launched, adjusted analyses suggested the operational era itself was associated with greater attrition.

• That mismatch highlights a common AI evaluation problem: apparent gains in observational rollout data may reflect confounding, not true tool effect.

• For AI in practice, prediction and risk stratification may deliver more value than generic patient-facing automation.

The Bottom Line

• This study suggests AI chat support did not independently improve 6-month semaglutide retention.

• Clinical, behavioral, and financial factors were stronger predictors of persistence.

• For clinicians interested in AI, the near-term opportunity may be using digital signals to identify dropout risk early and intervene with human-centered support.