Overview
The authors examined how AI is used across the stages of care in facial cosmetic surgery, which surgeries it is applied to, what kinds of models are used and where they come from. Following PRISMA 2020 and a registered protocol, they searched PubMed and Web of Science on August 25, 2025, for English studies with at least 4 patients. Two reviewers screened and extracted data, and bias was rated with ROBINS-I and RoB 2.
Findings
- 1832 records screened; 29 studies met criteria.
- Most studies came from Asia or the United States.
- Most studies focused on validating, developing or testing AI tools; few were interventional.
- Most AI models used convolutional neural networks and were commercially available.
- AI helps with treatment planning and outcome assessment, especially in facelift, rhinoplasty and blepharoplasty.
- The authors call for standard reporting, updated regulations, and education so surgeons help design AI models.
- Level of evidence: 4.
What it means for a patient
- AI tools are being used to plan facelifts and to measure results, but few studies test whether they change care.
- Most AI models studied are commercial products.
- The review does not show whether AI improves results for patients.
- Limits: 29 mostly early-stage studies, and the abstract gives no pooled accuracy or outcome figures.
Why this paper matters
It maps where AI is used in facial aesthetic surgery and shows the evidence is mostly tool development rather than tests of patient benefit. It points to reporting standards and regulation as next steps.
Terms
- Artificial intelligence (AI): computer programs that learn patterns from data to make predictions.
- Convolutional neural network: a type of AI model built to analyze images.
- Systematic review: a study that searches for and combines published research on one question using set rules.
- PRISMA 2020: a standard checklist for reporting systematic reviews.
- PROSPERO: a public registry for systematic review protocols.
- Interventional study: a study that tests the effect of doing something to patients.
- ROBINS-I and RoB 2: tools for rating risk of bias in non-randomized and randomized studies.