ChatGPT is good at skincare in general and blind to your skin in particular. It can define ingredients, explain treatment types and describe what suits most people. It cannot see the context your skin sits in — your environment, your history, what you have already used, and what changed recently. Its answers are population-level. Your skin is a specific case.
This is not an argument against the tool. It is a description of where the tool stops. A new-guest consultation at Glow in Springvale is $150 and runs for 90 minutes, and most of that time is spent on exactly the material ChatGPT cannot access.
Yes. Ask it for a routine and it will produce a sensible one: a cleanser, SPF, an active or two, an order to apply them in. The steps will be reasonable. Most of them appear in every routine it writes, because they are what suits most people, most of the time.
That is not a flaw. It is what the tool is. A language model generates answers from patterns across everything it has read. Ask a general question and you get the aggregate answer — the most common advice, the most common ingredients, the most common sequence. None of it is about your skin, because the model has never seen your skin.
It does not know your skin feels tight by mid-afternoon in a heated office. It does not know you used retinol for two months last year and your skin objected. It does not know whether your congestion arrived last month or has been there for years — and those are different situations, even though they produce the same one-line prompt.
Because a routine is an answer, and the question is different for every face. Underneath any recommendation sits context: environment (heating, hot showers, the Melbourne winter, the commute), history (what you have used, what you reacted to, what you gave up on after a week), timing (what changed in the month before your skin did), and load (what you are already applying, and whether those actives belong together).
"What should I use for congestion?" cannot be answered well without that context, and AI cannot reach any of it from a short prompt. So it answers for the average case. A generic answer applied to a specific face lands generically. This is how bathroom shelves fill up: reasonable products, chosen from reasonable answers, none of them matched to the skin they were bought for.
It can describe what is visible in one image, under one light, at one moment. That is the limit. A photo carries no history — how long pigmentation has been there, what the redness followed, what you applied the night before, how the same skin looks across a normal week. Two very different skins can photograph almost identically.
The research so far points the same way. A 2024 review in the Journal of Clinical Medicine found ChatGPT reliable on common, general dermatology queries and weaker on complex, specific presentations, and concluded it should support clinical expertise rather than stand in for it. That matches how the tool is built: strong where the answer is general, weak where the answer is you.
Quite a lot, used at the right stage. It can define a term you have seen on a label. It can explain the difference between an AHA and a BHA. It can describe what a treatment involves before you sit in front of anyone. It can help you draft the questions you want a clinic to answer — and a client who arrives with clear questions is harder to sell to, which is a good thing.
That is the honest division of labour. Use AI before a decision, to understand the vocabulary. Do not use it as the decision, because the decision needs context it does not have. Glow's before-you-book notes exist for the same stage of thinking, and they are written to be read alongside anything an AI tells you.
What is going on with your skin, in its actual context. A new-guest consultation at Glow Cosmetique in Springvale is $150 and runs for 90 minutes. It is a full assessment of what your skin is doing and why — environment, history, timing and current products included — and it ends with a written plan: what suits your skin, in what order, over what period. The plan is yours to keep whether or not you book anything, and it works at another clinic or at home.
An assessment also has limits, and they are worth stating plainly. Sometimes the recommendation is to wait, or to do nothing yet. Glow describes skin behaviour; it does not diagnose. Anything that looks medical belongs with your GP, and a consultation will say so directly.
A generic answer is a starting point. An assessment is an answer about you.
---
Written by [RN — pending assignment], reviewed [date].
Originally published in a different form at sigmasync.com.au/thinking/what-asking-ai-wont-fix. Adapted for skin.
1. Goktas P, Grzybowski A. Assessing the Impact of ChatGPT in Dermatology: A Comprehensive Rapid Review. Journal of Clinical Medicine, October 2024. 2. Skin conditions. Healthdirect Australia (Australian Government–funded health information service). Accessed July 2026.