Artificial intelligence-enabled interventions for smoking cessation: a systematic review and meta-analysis of randomised controlled trials.
Yusuff Adebayo Adebisi, Anas Ali Alhur, Isaac Olushola Ogunkola et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis assessed randomised controlled trials of AI-enabled smoking-cessation interventions in adults who smoked combustible cigarettes. Fourteen trials involving 101,331 participants were included, with eight contributing to meta-analysis. Pooled abstinence estimates were imprecise and did not establish a consistent benefit: at six months or longer, RR 1.33 (95 % CI 0.82-2.15; four trials, 2,694 participants), and at approximately three months, RR 1.36 (95 % CI 0.63-2.93; four trials, 715 participants). The authors concluded that evidence is sparse and heterogeneous, and does not show a reliable incremental benefit over non-AI digital support.
Key findings
- At six months or longer, pooled abstinence was RR 1.33 (95 % CI 0.82-2.15; I-squared = 67.6 %; four trials, 2,694 participants).
- At approximately three months, pooled abstinence was RR 1.36 (95 % CI 0.63-2.93; I-squared = 58.4 %; four trials, 715 participants).
- Adding two completed trials reported only in registries moved the six-month estimate towards the null (RR 1.20, 95 % CI 0.91-1.59).
- Certainty of evidence ranged from low to very low; no included trial randomised a generative large language model intervention.
Why it matters
Digital cessation tools increasingly incorporate algorithmic personalisation, but these results do not establish that AI itself improves abstinence beyond digital support more generally. Distinguishing the contribution of an algorithm from the effects of the broader intervention is important when interpreting efficacy claims and designing evaluations. The review highlights the need for clearer reporting and comparisons that isolate AI-specific effects.
What to keep in mind
The evidence was sparse, heterogeneous and imprecise, with low-to-very-low certainty. Only eight of 14 trials contributed to meta-analysis, and comparator type was confounded with other trial characteristics. Biochemically verified sensitivity analyses excluded the largest null trials and were not confirmatory.