What the research shows
Photos must be taken the same way. Tilting the head by a few degrees or a slight smile changed facial measurements by large amounts, enough to misrepresent a result (Riml 2011). Published standards set the views, lighting and position for each procedure so that before and after photos can be compared (Henderson 2005). Most women preferred a mirror-reversed photo of themselves, the face they see each day, and facial surgery patients more so (de Runz 2016).
3D imaging measures what the eye sees. 3D cameras showed that a vertical deep plane lift restored midface volume that lasted at least a year, by lifting fallen cheek fat back into place (Jacono 2015). Measured during surgery, the best direction of lift differed between patients and between the two sides of one face, and was more vertical in women than in men (Talei 2024).
How much younger patients look. Age-estimating AI judged patients about 5 to 6 years younger a year after SMAS lifts and about 3 years younger after skin-only lifts, with about 2 more years when fat was added (Gibstein 2021). Patients themselves felt about 7 years younger, more than the AI found, and the AI estimate tracked how satisfied they were (Zhang 2021, Du 2024). Patients who looked older than their age before surgery came back to looking their age, while those who already looked younger than their age looked younger still (Du 2024). In one practice the measured gain was larger at three months than at a year (Elliott 2023). Across many surgeons, combining face, neck, brow and eyelid surgery gave the largest gain, and patients who looked older beforehand gained the most (Rames 2026).
Around the eyes and in expression. A brow lift lowered apparent age, while eyelid surgery alone did not show a separate effect (Kreh 2025). Expression software found faces at rest looked less angry and happier after a brow lift (Boonipat 2021).
Chatbots as advisers. For questions before surgery, ChatGPT gave clearer and safer answers than Bard, but neither was accurate enough to rely on without a surgeon (Abi-Rafeh 2024). A newer model gave no unsafe advice but left out less common risks and did not tailor answers to the person (Mangal 2026). Faced with simulated complications after a facelift, ChatGPT usually named the likely problem but chose the right next step only about half the time, most often sending patients in more urgently than needed (Abi-Rafeh 2023). Readers could barely tell ChatGPT-written patient reviews from real ones (Knoedler 2024).
Where surgeons disagree
- Does technique change the AI result. One study found SMAS lifts took more years off than skin-only lifts (Gibstein 2021); others found no single approach better, including deep plane against SMAS lifts (Elliott 2023, Tan 2026).
- How far to trust the AI. On standardized front-facing photos AI estimated age closely (Du 2024); on profile views from surgeon websites it was off by well over a decade (Tan 2026).
What is new
- From photographs alone, ChatGPT-5 agreed with plastic surgeons on who is a facelift candidate in nearly every case in a small pilot (Saeed 2026).
- AI outcome studies moved from single practices to before-and-after photos from many surgeons across North America (Rames 2026).
What is still unknown
- Whether AI tools change care. Most studies validate tools rather than test them in practice, and reporting is not standardized (Colceriu 2026).
- How results look in unselected patients; the larger photo sets came from websites and society galleries chosen by surgeons (Tan 2026, Rames 2026).
Key papers
- Henderson 2005: standard photographic views for facial surgery.
- Riml 2011: small photo errors distort results.
- Jacono 2015: 3D proof that a vertical deep plane lift restores midface volume.
- Zhang 2021: AI age reduction tracks patient satisfaction.
- Gibstein 2021: AI compared facelift techniques by years taken off.
- Abi-Rafeh 2023: ChatGPT is not yet safe for judging complications.
- Knoedler 2024: AI-written reviews pass as real.
- Rames 2026: combined procedures gave the largest measured benefit.