Red Flags That Reveal an Online Skin Platform Is Skipping Human Review Before Prescribing

AI-assisted telehealth dermatology with clinician review means AI may help assess a case, but a licensed clinician reviews the patient’s photos and history and makes the prescribing decision before treatment is issued. A clinician’s name on a prescription does not prove that happened. Patients can look for five warning signs, including missing prescriber details, incomplete intake, weak safety screening, treatment menus without escalation, and hidden formulation details.

Why a clinician's name on a prescription no longer guarantees a human reviewed your case

You assumed the doctor read your chart. That is a reasonable expectation when a prescription arrives with a clinician’s name attached. Asynchronous, or “store-and-forward,” care allows the patient and clinician to participate at different times, but it still requires a clinical evaluation before prescribing. The concern is that human review can now be automated or deferred without that difference being clear to the patient.

Utah’s 2026 acne pilot illustrates the gap. Each AI-issued prescription bears a clinician’s name even though no person writes it. The staged model requires physician review for the first 100 patients, then permits autonomous prescribing with weekly retrospective review for the next 400, followed by monthly spot-checks of at least 10%. LaRocco et al. (2026) found that clinicians place most of the liability on the AI company and regulatory body when autonomous systems prescribe without escalation, while patients still bear the harm to their health.

What does genuine clinician review in AI-assisted online dermatology look like?

You are entitled to expect a clear sequence. You send photos and your health history, a licensed clinician reviews them on their own, and that clinician decides whether to prescribe before the prescription goes out. The FDA’s January 2025 draft guidance says supportive AI tools “rely on the human to interpret the AI outputs and ultimately make clinical decisions”. No fully autonomous AI prescription service has been cleared by the FDA.

A clinician who reviews an output must also spot where it is uncertain before approving it. A clinician looks at visual features along with medications, skin history, and systemic conditions that an algorithm may not have been trained to flag. LaRocco’s study identifies three minimum conditions for safe autonomous prescribing, namely confidence thresholds that stop uncertain decisions, communication that distinguishes model uncertainty from clinical ambiguity, and reasoning transparent enough to audit. The clinicians surveyed would not permit autonomous prescribing without confidence-based escalation and would accept additional liability only when transparent reasoning allowed substantive judgment.

Red flag 1 — No named, credentialed clinician disclosed before or at prescribing

Who is responsible for the treatment you receive? A platform should disclose the name and credentials of the licensed clinician responsible for prescribing before or when the prescription is issued. The rules on what services must disclose vary by state, so missing details should not be treated as the same breach of law everywhere. They are nevertheless a practical warning about accountability.

The first check belongs before the intake form. Search the website for the prescriber’s name, state license, credentials, and a way to reach them directly. An undisclosed prescriber’s identity is a gap to resolve before you proceed. But identity is only half the question. As the Utah pilot shows, a name on the label cannot establish that its owner reviewed your case, so the platform must also explain what the named clinician actually does before prescribing.

Red flag 2 — Questionnaire intake with no image review or clinical follow-up

A symptom questionnaire cannot supply all the information dermatology requires. Texture, distribution, and how the skin around it looks help tell one condition from another. Written answers serve a different purpose than photographs. They can describe history and symptoms, but they cannot replace the visual evidence a clinician needs to assess them. A platform prescribing from a questionnaire alone leaves that part of the assessment incomplete.

Even the Utah pilot calls for standardized, multi-angle photos as part of a structured intake that takes about 10 to 15 minutes. A service can collect images without a human ever reviewing them. An almost immediate prescription with no stated review window and no explanation of who assessed the images is a stronger warning. The practical check is whether the intake description requires photos, promises clinician review before prescribing, and gives a timeframe. It should also explain how to send additional images or report symptom changes, because a dead end after purchase leaves no clear human pathway for addressing treatment failure or adverse reactions.

Red flag 3 — Safety screening for high-risk prescriptions is missing or superficial

Even an intake that is easy to use must ask hard questions about safety. For tretinoin, minimum screening includes pregnancy and breastfeeding status, current medications and topical products, especially photosensitizing agents, and active skin conditions such as eczema or severe dermatitis that may contraindicate use. A platform that dispenses tretinoin without asking them has skipped minimum safety review. A product recommendation is not enough if the intake has not established whether the patient can safely use it.

Isotretinoin has a separate, legally enforceable safety framework. The FDA restricts its distribution through the iPLEDGE REMS, which requires prescribers to be enrolled and activated. Patients who can become pregnant must have a confirmed negative pre-treatment pregnancy test in a medical setting and use two forms of contraception consistently for at least one month before starting, throughout treatment, and for one month after stopping. These requirements are not optional platform policies. An online service offering isotretinoin without the required enrollment verification, testing, and documented two-form contraception counseling violates the REMS.

The FDA’s February 2026 provisions allow pregnancy tests during and after treatment outside a medical setting, including at home, if the prescriber permits it. The pre-treatment test must still occur in a medical setting. That distinction is important when evaluating a platform’s description of remote care. For patients, the check is whether the service explains the requirements accurately and verifies iPLEDGE enrollment before prescribing, rather than treating a general safety questionnaire as sufficient.

Safety also extends beyond the first prescription. Depression, psychosis, and suicidal ideation have been reported during isotretinoin therapy, and a questionnaire-only intake cannot reliably detect or monitor these concerns. A clinician who keeps seeing the patient can watch for these concerns over time. AI confidence offers no substitute, particularly when training images under-represent darker skin tones and less common conditions. Such gaps can leave an algorithm confident in a case where a human clinician would pause.

Red flag 4 — A formulary so narrow it fits every patient, or so broad it has no guardrails

A formulary is the set of treatments a service offers. A short list is not inherently a problem, especially when the service treats a limited range of conditions. The warning appears when the same menu is presented as suitable for every skin type, condition, and severity. What happens to the patient who does not fit? Without an escalation pathway, the choice can come down to what products are on offer at the expense of what the patient needs.

The Utah pilot provides a useful contrast. Its AI is restricted to physician-approved topical treatments and predefined pathways, cannot improvise, and must hand cases to a physician when it cannot choose confidently. Those limits distinguish a bounded service from either a universal product shortlist or an unconstrained prescribing system. They also acknowledge that some cases need something else. Dermatologists who discussed the pilot said topical-only options may not be enough for patients who need systemic therapy or more advanced treatment, especially for moderate-to-severe acne that risks scarring.

The practical check is the platform’s explanation of exceptions. It should describe what happens when a condition is complex, a contraindication appears, or the standard treatments do not fit. AAD president Murad Alam, MD warned that reduced physician oversight increases the risk of reviewers deferring to AI output.

Red flag 5 — Ingredient concentrations and formulation rationale are hidden or absent

Knowing the product’s name is not enough for informed consent. Patients need to know the active ingredients, their concentrations, and why that formulation was chosen for their condition. Branded language cannot answer those clinical questions. If details are withheld or buried in marketing copy, patients have less ability to flag interactions, recognize side effects, or make meaningful treatment decisions. Disclosure is part of prescribing.

This is especially important for compounded medications. These custom-mixed formulations are common on skin platforms, and the prescribing clinician must specify what was ordered rather than leaving patients to infer it from a standard product label. The practical check is whether ingredients and concentrations are disclosed before purchase and whether the treatment rationale relates to your skin type and condition. A claim about what a treatment can do in general does not explain why it was prescribed for one patient. If the platform cannot explain the choice, the clinical reasoning remains hidden and informed consent incomplete.

Why the access argument does not resolve these concerns

Faster, more affordable skin care is a real benefit. There is the strongest case for making care easier to get when patients need simple, low-risk topical treatment and have no contraindications. That is the bounded situation the Utah pilot was designed around, with a restricted formulary and explicit escalation requirements. It is not a description of everyone seeking online skin care. The access argument weakens when patients take multiple medications, have active systemic conditions, or need treatment with greater safety demands.

Training-data gaps make the trade-off more difficult for patients with darker skin tones, while retinoid and isotretinoin prescribing require particular attention to screening. Speed cannot resolve those gaps in assessment. Dr. John Whyte, CEO of the American Medical Association, said AI could do a great deal but cannot take the place of physician judgment. The AMA’s position is that AI should assist rather than act as an autonomous decision-maker, with transparency, accountability, and meaningful physician oversight.

The alternative is not necessarily a lengthy in-person wait. A well-structured asynchronous consultation can be fast, affordable, and supervised by a clinician who reviews the case before prescribing. For most patients, its review step need not impose a meaningful delay compared with AI-only prescribing.

How to use this checklist before you submit a photo or fill out a form

To apply these checks, start with the public website before creating an account. First, look for a named prescriber, their credentials and state license, and a direct contact route. Then read the intake description. It should require photos and health history, state that a clinician reviews them before prescribing, and explain the review timeframe.

The platform should explain how a human clinician becomes involved when a case is complex, a contraindication appears, or the formulary does not fit. Then examine the stated safety screening requirements. For retinoids, look for questions about pregnancy, breastfeeding, medications, and existing skin conditions, and for isotretinoin, ask directly about iPLEDGE enrollment before proceeding. If the website does not provide these details, the missing information is something to ask about before sharing clinical information.

Finally, check what the service promises to tell you before you buy. Active ingredients, concentrations, and a reason for choosing the formulation are all necessary to understand the proposed treatment. Your exact prescription cannot be explained before assessment, but the platform can state its disclosure policy in advance. Without those details at the prescribing stage, you do not have enough information for informed consent. Telehealth remains a legitimate first step for everyday skin concerns, but the decisive question is whether a clinician makes the treatment decision before the prescription reaches you.