What the evidence says about AI before-and-afters

What this is

Surgeons have altered patient photos for decades. Canfield's Mirror software has done it in consultation rooms since long before anyone called it AI, and there was Photoshop and a pencil before that.

The mockup used to look like a mockup. Now it looks like a photograph.

So we went looking for whether that change does anything for the decision a patient is making. We pulled every product we could find that generates a patient-specific outcome image, priced them, and read the peer-reviewed literature on whether the images work.

Vendor and manufacturer numbers are labeled as such throughout and were not treated as evidence. Where we couldn't verify something, we say so.

One thing worth doing before you buy any of this: run an audit of your own digital presence, so you can see what patients are already seeing. A render is one small piece of that picture, and it is easier to judge once you know what the rest of it looks like.

Do patients actually decide on the render?

Short answer: it came third, behind your photos and behind you.

A 2026 study followed 75 women through rhinoplasty, breast augmentation, mastopexy, augmentation-mastopexy, and reduction, all of whom received preoperative 3D simulations. A year after surgery, researchers asked what had driven the decision.

Photos of previous patients took 23 of the 75. Talking to the surgeon took 22. The simulation took 14. The remaining sixteen picked something the study doesn't list.

The top two have something in common. Photos show what a surgeon produces, and conversation shows whether that surgeon understands what a patient wants. Both carry information about the person who'll be holding the instruments.

The simulation carries none. It shows a possible result and says nothing about who would produce it.

Hold it lightly. The gap between the simulation and the consultation is 8 people, and the authors call their own study preliminary.

Patients arrive at this decision nervous, which is part of why the images matter at all. A 2025 study of 122 plastic surgery patients found more than 90% carried preoperative concerns into the procedure, and 16.4% didn't sleep the night before. Asked to rank 11 kinds of content, 100 aesthetic patients put before-and-after photos first.

A much older study puts all of it in proportion. A 2013 conjoint analysis of 150 patients decomposed the choice of surgeon into weighted attributes.

Referrals from a GP, the primary care physician, carried weight. Referrals from TV, radio, and internet forums had a negative effect.

We sell digital marketing, so we'll say the uncomfortable part out loud. The entire online layer came to 9%, and surgeon experience was 4 times that. Your reputation outranks anything we can build for you.

That is the frame a simulator belongs in. It optimizes the smallest of the six attributes and gets sold as a differentiator.

What are renders actually good at?

Short answer: comparing options you get to choose between.

Where an outcome is largely set by something you select, simulation has real material to work with. Filler volume and placement. Implant size and profile. A lift alongside augmentation versus augmentation alone. Two feasible reduction targets.

The evidence is specific about where this applies. In that same 2026 study, patients rated how much the simulation influenced them on a 1 to 10 scale. Breast augmentation scored 8.4. Breast reduction scored 3.8.

Same body part, and the gap tracks how much of the result you can pick off a shelf. Augmentation is largely set by an object with known volume and projection. Reduction is set by tissue quality, pedicle choice, and how the scars behave.

That suggests a principle, and we offer it as our reading of two data points rather than as something the literature states: simulation earns its place in inverse proportion to how much biology mediates the result.

The clearest published example gets this right. A 2025 pilot at MedStar Georgetown used soft tissue simulation to help seven breast reduction patients choose between two feasible cup sizes. Cup sizing is unreliable enough that documented gaps between perceived and measured size run as high as 70 to 100%.

Patients rated that choice 9.6 out of 10 for difficulty, and every respondent called the images extremely helpful. The authors framed it as a decision aid and said plainly it wasn't a prediction.

Seven patients, and nobody has replicated it since. It is the right use of the tool resting on the smallest possible evidence base.

Why can't anyone tell me how accurate it is?

Short answer: because the patient hasn't finished healing yet.

A prediction can only be checked once the patient it was made for has healed. That window runs weeks for a neuromodulator, a few months for filler or a laser series, several months for breast surgery, and 12 to 18 months for a nose.

So validation is a per-procedure question, not a per-product one. A filler simulator that launched in 2024 has had time to compare its predictions against real outcomes. A rhinoplasty simulator that launched the same year has not.

There's a related distinction the marketing blurs. Training a model on historical cases isn't the same as validating what it predicts. EntityMed says its facial simulator draws on more than 50,000 clinical cases, which describes training data. It says nothing about whether any render was later held up against how that patient actually healed.

A vendor can say both things truthfully, and only the second is evidence.

Where accuracy has been measured, it lands on approximate. The numbers below are not comparable with each other, since each was produced by a different method on a different procedure, and two of them are patient self-assessment rather than blinded measurement. In the 2026 study, patients rated the resemblance between render and result at 7.5 out of 10 for rhinoplasty and 7.9 for breast augmentation, by self-assessment rather than blinded measurement. A separate study of 38 rhinoplasty patients found accuracy correlated with satisfaction at 0.66. Canfield VECTRA XT, across 78 patients, predicted postoperative volume within 10% in 73% of cases.

There's also a limit no amount of time fixes. Open versus closed rhinoplasty is an access strategy, and the external result can be identical either way. The real differences sit in the exposure, the dissection, the grafting, and the revision profile.

A render of the outside of a nose shows none of it. Septoplasty versus septorhinoplasty is sharper still, because septoplasty is done for airflow.

We found no vendor and no published workflow comparing two surgically distinct techniques on the same patient. What separates two real surgical options usually sits under the skin, where a camera cannot reach it.

Who is selling this, and what are they claiming?

Short answer: two very different groups, split by where the image gets delivered.

We looked at sixteen products that generate a patient-specific outcome image.

Three more names came up and we couldn't verify them. RealFaceValue, New Look Now, and AesthetIQ AI either have no current product documentation or turned out to be something other than a simulator. Canfield Mirror, oVio360 and TouchMD sit outside the table on purpose, since none of them generates a prospective image.

The first group sits in the room with a clinician. The output gets discussed rather than delivered, which is the right posture for an approximation.

Breast augmentation has a wrinkle. Allergan publishes an 86% consultation-to-treatment figure for Crisalix on its Natrelle materials, sourced to physician user data. Johnson & Johnson publishes 87% for the Mentor-branded Arbrea app, sourced to a conference abstract. Two competitors landed a percentage point apart, neither figure rests on a study, and each tool shows its own sponsor's implants. Device makers have always funded the tools surgeons use, and that 86% came from the implant maker.

The second group never meets a clinician. It generates a preview for an anonymous website visitor and sells that to practices as lead generation, and it's priced accordingly. PreviewMD charges a dollar per preview past plan, which means revenue scales with images generated rather than with whether the image was any good.

We could not do basic diligence on that tier, and that is itself a finding. Across the website group we repeatedly could not find a legal entity, named founders, a launch date, or which underlying image model is doing the work. Where a site claimed photos are never stored, we often could not retrieve a privacy policy or a data processing agreement to check it against.

Some of this group also assert in their own marketing that no business associate agreement is needed. Whether one is needed is a determination for you and your counsel, and it is not one a vendor gets to make on your behalf in a FAQ.

What surprised us

Nobody has tested whether photorealism does anything. There is no published experiment comparing photorealistic renders against rough morphs on how firmly patients hold onto them. Looking more real is the entire commercial argument, and it is unmeasured.

Render time separates the two tiers cleanly. The MedStar decision aid takes a week or more. PreviewMD takes three to four minutes, ClinicOS advertises under five seconds, and the Mentor app captures and simulates in sixty seconds. Generating an image doesn't require a week. A clinician reviewing it does.

The adjacent literature is not encouraging. A 2026 review of AI-generated medical images found clinical fidelity problems in 47.2% of the studies it examined. A 2025 review of AI beauty filters found believable changes to nasal tip rotation, lip fullness, and jaw prominence produced while anatomy gets ignored.

Most of what a simulator buys is a signal. This is a judgment rather than a finding, but the pitches lead with differentiation and lead generation. If the value were clinical accuracy, you'd expect accuracy data in the marketing, and what you get instead is conversion data.

Signaling that you invest in technology is legitimate. Buying a VECTRA says something about you, the same way a renovated waiting room does. The difference is that the practice captures the benefit of looking current while the patient carries an image of their own face implying a result nobody validated. A renovated waiting room makes nobody a promise.

How we would think about this in a practice

This is how we would work through it, and every practice has context we do not. The first question is which tier you are being sold. The next four are for the vendor, read against what the evidence supports.

If you want it for the consultation room. We would buy it as a comparison tool and use it as one, showing two or three options against each other rather than one image as the answer. The literature that supports simulation studied option comparison, and none of it studied prediction.

If you want it on your website. That is a marketing purchase, so it deserves marketing scrutiny: where the photograph goes, who processes it, and who decided that no business associate agreement was needed.

If you already run one. It is worth checking what the patient-facing page claims. The headline, the imagery, and the disclaimer get read together, by patients and by the FTC.

If your budget is finite. Photos of real patients outperformed simulation as a decision factor and cost nothing beyond the discipline to shoot them consistently. On that evidence, standardized photography returns more than a subscription does.

Who produced the conversion number, and what does it cite? Internal user data and conference abstracts are marketing. Ask what conversion was defined as.

What's in the catalog? If the tool shows one manufacturer's products, know that before it gets described to a patient as a neutral comparison.

Validated against what, over what window? Ask per procedure. Ask also whether they mean training data or outcome comparison, because those get used interchangeably and only one is evidence.

Where does the photograph go? Privacy policy, data processing agreement, subprocessor list, and a written position on business associate status.

The honest summary is that this technology is real, useful in a smaller range than it's sold for, and arriving faster than the evidence behind it. That's a reason to know which tier you're buying from, not a reason to avoid it.

About this review

Method: Sixteen products identified and priced from public vendor materials, plus peer-reviewed literature on simulation accuracy and decision-making retrieved through August 2026, plus public regulatory guidance. Manufacturer-published and vendor-published figures are labeled as such throughout and were not treated as independent validation.

Limits worth stating. We could not access the full text of the 2026 simulation study, so the survey instrument is inferred: the reported shares resolve to whole patients out of 75, which indicates a single-select question, and the fourth category in the chart is derived by subtraction rather than reported by the authors. Pricing is as advertised and was not independently confirmed. There is no public market-share data for this category. And nobody has compared photorealistic renders against rough morphs on expectation anchoring, which is the most important open question here.

Fractional Aesthetics is a growth strategy practice for aesthetic medical businesses.

Sources

S1Cross-sectional study of preoperative concerns in 122 plastic surgery patients, Aesthetic Plastic Surgery, 2025. Peer-reviewed.

S2Survey of 100 aesthetic surgery patients ranking 11 content types, reported by ASPS, 2017. Industry survey.

S3Ersan et al., preoperative 3D simulation and virtual reality in aesthetic surgery, Computer Assisted Surgery, 2026. Peer-reviewed.

S4Lava CX, Li KR, Spoer DL, et al. The role of decision aids and soft tissue simulations in breast reduction surgery: a pilot study. Plast Reconstr Surg Glob Open. 2025;13(9):e7129. Peer-reviewed.

S5Review of clinical fidelity in AI-generated medical images, 2026. Peer-reviewed.

S6Review of AI beauty filters and facial feature modification, 2025. Peer-reviewed.

S7EntityMed public product and customer materials, 2025. Vendor-published.

S8Conjoint analysis of surgeon-selection attributes, 150 patients, Plastic and Reconstructive Surgery, January 2013. Peer-reviewed.

S9Yamamichi et al., simulation accuracy and FACE-Q satisfaction in rhinoplasty, 38 patients. Peer-reviewed.

S10Lorange et al., VECTRA XT breast volume prediction accuracy, prospective cohort, 78 patients. Peer-reviewed.

S11Crisalix and Allergan Aesthetics, Natrelle 3D Breast Visualizer materials, March 2026. Manufacturer-published.

S12Johnson & Johnson MedTech, Arbrea Breast Simulator Surgeon App for Mentor product page, 2026. Manufacturer-published.

S13Vendor product and pricing pages for all sixteen products listed above, retrieved August 2026. Vendor-published.