In short. Sort the words that commit to something from the ones that only signal novelty, and know which claims can be checked at all.
Key points
- A new method is not better for being new, and a familiar one is not worse for being familiar.
- The words that signal novelty are doing the work that evidence would do if evidence existed, and a page that leans on novelty is usually a page with nothing else to lean on.
Marketing about software in aesthetics leans on a small set of words, and most of them are chosen because they cannot be checked. The useful skill is not scepticism in general but knowing which sentences commit the writer to something verifiable and which only signal that something new is involved.
Words that commit to nothing
Intelligent, advanced, personalised, next-generation, precision, powered by AI. Each of these sounds like a claim and asserts nothing that could be tested. They are not lies, which is precisely why they survive scrutiny: there is no proposition in them to examine. When a paragraph relies on them, the paragraph is not describing a product.
Words that do commit
A statement that a product holds a clearance for a named task is checkable. A statement that a study measured a particular outcome is checkable. A statement about how images are stored and how consent works is checkable. These are the sentences to look for, and their absence from a page that is otherwise full of confident language is informative in itself.
Novelty is not evidence
A new method is not better for being new, and a familiar one is not worse for being familiar. The words that signal novelty are doing the work that evidence would do if evidence existed, and a page that leans on novelty is usually a page with nothing else to lean on.
Photographs before and after
Paired photographs are the most persuasive and least controlled evidence in this field. Lighting, angle, expression, makeup, posture and camera position all change between two images, and a difference produced by any of them is indistinguishable from a difference produced by a treatment. Without standardised conditions and an account of who was included, a pairing is an illustration rather than a result.
Testimonials and demonstrations
A person describing a good experience is describing their experience, which cannot be generalised to anyone else and cannot be checked. A demonstration shows software running, which nobody disputes and which says nothing about accuracy. Neither is evidence about whether the output is right.
Three questions that sort a claim
- What named task is this product cleared for, and does the sentence stay inside it?
- What outcome did the study measure, on whom, and against what comparison?
- What happens to my photograph, under which terms, and can I have it removed?
Ask those three and most AI claims resolve into one of two groups: a measurement tool with a defined task, or a page of adjectives. That sorting is the whole of the exercise.
Frequently asked questions
What does the word AI commit a clinic to?
In practice, very little. It usually signals that software was involved somewhere without saying what it does, what it is cleared for, or how it was evaluated. A claim that cannot be checked cannot be relied on, and novelty language is rarely checkable by design.
Which claims in this area can be checked?
Three: whether a product holds a clearance, and for what named task; whether a study exists and what outcome it measured; and what the terms say about your images. Those are answerable. Whether a tool is intelligent, advanced or personalised is not.
Is a demonstration convincing evidence?
No. A demonstration shows the software running, which nobody disputes. It does not show that the output is accurate, that it changes what happens to anyone, or that the training data resembles the people it will be used on. Seeing something work is not the same as seeing it work correctly.
Risks & Limitations
- Software measures an image; it does not examine tissue, and an examination is what a diagnosis usually turns on.
- A rating reflects what a model was trained on rather than a standard of beauty, and it is least reliable for faces least represented in its training data.
- Agreement between software and clinicians is measured against an opinion, not against correctness.
- A predicted image is a drawing of a plan expressed in a body the model does not describe.
- No algorithm takes a history or carries responsibility for the advice it gives.
When to Seek Care
- A new or changing lesion, a mole changing in size, shape or colour, or a sore that is not healing — have it examined by a clinician rather than assessed from a photograph.
- A skin change you are worried about, whether or not a tool has told you it is nothing.
- A reaction, infection or worsening after treatment — contact the service that holds your records.
- If you have uploaded a facial image and want to know how it is held or want it deleted, ask the service that holds it under the terms you accepted.
References & Further Reading
- U.S. Food and Drug Administration (FDA) — guidance and authorisations for software as a medical device and for image analysis
- European Commission and notified bodies — Medical Device Regulation and artificial intelligence guidance
- National Medical Products Administration (NMPA, China) — approved software and its named claims
- World Health Organization (WHO) — guidance on the ethics and governance of artificial intelligence for health
- Cochrane Library — systematic reviews of diagnostic and decision-support software
This page is patient education, not medical advice, and it does not cite individual studies. It reviews evidence and regulation; it is not a booking page and does not recommend a treatment. Discuss your own case with a qualified clinician. To ask a question about this review, see Contact & Consultation.