In short. It cannot examine tissue, take a history, judge expectations or carry responsibility for a decision, and no study shows it improves care.
Key points
- The volume of data a tool collects says nothing about whether the right data was collected.
- A long questionnaire that misses the one relevant question is less useful than a short conversation that finds it, and the difference is judgement rather than completeness.
Claims about software in consultation fail in a consistent way: each one takes something software genuinely does and extends it into something a consultation consists of. Listing them is the quickest way to see where the boundary actually sits.
It cannot examine tissue
An image records appearance in one lighting condition at one moment. It carries nothing about texture, temperature, tenderness, how something moves or how it behaves under a hand. Those are not properties a better camera would capture; they are properties that require contact and a trained observer, and a diagnosis is frequently made because of one of them rather than because of how something looks.
It cannot take a history
A form collects answers. A history is the process of deciding which answers matter and following the ones that do. A clinician notices the pause before an answer, asks a second question because the first one was vague, and asks about something the person did not think to mention. Software asks what it was written to ask, and an unasked question is invisible in the output.
Collecting information is not assessing it
The volume of data a tool collects says nothing about whether the right data was collected. A long questionnaire that misses the one relevant question is less useful than a short conversation that finds it, and the difference is judgement rather than completeness.
It cannot judge expectations
A consultation includes forming a view about whether what someone wants is achievable, whether they have understood what a treatment does and does not do, and whether the honest answer is that they should not proceed. That judgement requires being in the room, and it carries a responsibility that no software accepts.
It cannot carry responsibility
This is the claim that underlies all the others. A clinician who gives advice is accountable for it, and accountability is what makes advice careful. A tool that returns text is accountable for nothing, which is why a plausible answer and a considered answer can be produced by the same software and only one of them has been weighed.
Three claims that do not hold
- That a score identifies a problem — when it identifies a preference in its training data.
- That a predicted image shows a result — when soft tissue behaviour is outside the model.
- That a chat reply is advice — when nothing in the process examined, asked or accepted responsibility.
None of this makes measurement software useless. It locates it: a tool that produces a number, which a clinician interprets alongside an examination and a history, and which substitutes for neither.
Frequently asked questions
Can software take a medical history?
It can collect answers to questions, which is not the same thing. A history is a clinician deciding which answers matter, following up an ambiguous reply and noticing what was not said. Software asks what it was programmed to ask, and a question nobody wrote is never asked.
Can software judge whether my expectations are realistic?
No. Judging expectations requires seeing the person, understanding what they have seen and been told, and being willing to say that what they want is not achievable. That is a conversation with a person who carries responsibility for having it, and software carries none.
Is a tool that measures my face a diagnosis?
No. A measurement describes an image. A diagnosis asserts a condition, and it requires a different regulatory claim and evidence of accuracy in the population where the tool is used. A measurement can be perfectly reliable and still not be a diagnosis.
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.