Most people picture telemedicine as a video call. A patient on one side, a clinician on the other, a conversation that resembles an office visit conducted through a screen. That model exists, but it is not how the majority of online weight-care prescribing happens.
The dominant model is asynchronous. A patient completes a structured intake at whatever hour suits them, the submission queues, and a licensed clinician reviews it later without the patient present. No appointment, no simultaneity. It is sometimes called store-and-forward, and it is the engine underneath a market that Fact.MR valued at 763.8 million dollars in 2026 for GLP-1 telehealth platforms alone, with projected growth to roughly 3.28 billion by 2036.
Asynchronous care is not a lesser version of a video consultation. It is a different instrument with different strengths, and understanding where it is strong and where it is weak is more useful than treating it as a compromise.
What the structured intake is doing
A well-built asynchronous intake is not a form. It is a clinical decision pathway rendered as questions, and its quality determines everything downstream.
The intake has to capture medical history, current medications, relevant family history, contraindications and the specific factors that would make a particular drug class inappropriate for this person. It has to do this without a clinician present to notice hesitation, ask a follow-up or observe anything. Every piece of information a clinician would normally elicit conversationally has to be anticipated and converted into a question that a non-clinician can answer accurately.
This is harder than it sounds and it is where providers diverge most sharply. Two intakes can ask a similar number of questions and differ enormously in what they would catch.
The advantages are real and underrated
Asynchronous review has genuine clinical merits that the video-call framing obscures.
A clinician reviewing a written submission has the full history in front of them in a consistent format, with time to check interactions and consider the record rather than responding in real time. Patients, for their part, complete intakes more honestly than they report in person on certain topics. Weight history, alcohol consumption and prior treatment attempts are all subjects where the absence of a face across a desk tends to improve disclosure rather than degrade it.
There is also an access argument. A patient who would have to take time off work, arrange transport and sit in a waiting room faces a barrier that has nothing to do with their clinical need. Removing it reaches people who would otherwise not present at all.
The failure modes are specific
The weaknesses are equally concrete. A structured intake captures what it was designed to ask. Anything outside that design is invisible, and a patient cannot volunteer what they do not know is relevant.
There is no physical examination. For weight-care prescribing this matters less than in many specialties, but it is not nothing, and providers handle it differently. Some require recent vitals or laboratory work. Some do not.
The most significant risk is structural rather than clinical. An asynchronous pathway can be designed so that approval is the default and declining requires an affirmative act against the commercial grain. Nothing about the technology requires this, but the incentive exists wherever an intake is also a checkout.
Why the regulatory layer makes the intake load-bearing
The clinical review step carries more weight in 2026 than it did two years ago, for reasons that have nothing to do with telehealth itself.
The FDA removed tirzepatide from its shortage list in December 2024 and declared the semaglutide injection shortage resolved in February 2025. That closed the broad enforcement discretion under which large-scale compounding had operated. A 503A pharmacy compounding these molecules now works from a prescriber-documented, patient-specific clinical rationale rather than a shortage designation.
Compounded medications are prepared by a licensed pharmacy for an individual patient, are not approved by the FDA, and do not go through its premarket review for safety, effectiveness or quality. In April 2026 the FDA proposed permanently excluding semaglutide, tirzepatide and liraglutide from the 503B Bulks List, with a comment window extended into July 2026.
The consequence is that the prescriber’s individualised determination is no longer merely good practice. For compounded preparations it is the basis on which the whole arrangement rests. An intake that cannot produce a documented clinical rationale is not performing the function the current framework assigns to it.
How it works in practice
HealSend, a US telehealth service, runs an asynchronous model with the review step stated explicitly. Patients complete an online health profile, and the company states that every patient is reviewed by a US-licensed physician before any prescription is written, with eligibility for GLP-1 medications determined by that physician. Treatment is open to adults aged 18 to 75 and is not for use during pregnancy.
Where a prescription is appropriate, the company lists compounded semaglutide at 299 dollars per month and compounded tirzepatide at 399 dollars per month, each as a once-weekly injection, shipped via FedEx in discreet, temperature-controlled packaging. Patients looking into semaglutide online can see the clinical review step and the price before beginning an intake rather than after. Plans are stated to be HSA and FSA eligible for reimbursement, and the company states that if its provider determines a GLP-1 medication is not right for a patient, the patient receives a full refund. The service operates across most, though not all, US states.
That refund term is the part worth isolating. In an asynchronous model the single most important safeguard is that a declined prescription is an acceptable outcome for the business as well as for the patient. A provider that refunds in that case has removed one of the pressures that would otherwise bear on the review step.
How a good intake differs from a weak one
Because the intake carries so much of the clinical load in this model, it is worth being concrete about what separates a well-constructed one from a form that exists to collect an address.
A strong intake asks about contraindications specifically rather than generally. A question inviting a patient to list any relevant medical history relies on the patient knowing what is relevant, which is precisely the expertise they do not have. A question naming particular conditions and asking directly does not.
It handles medication interactions by capturing a full current list rather than asking whether the patient takes anything that might interact. The second phrasing outsources a clinical judgement to someone unqualified to make it.
It includes internal consistency checks. A submission where stated history conflicts with stated medications should surface for closer review rather than passing through. This is straightforward to implement and frequently absent.
And it has a mechanism for the clinician to request more information rather than choosing between approving and declining on an incomplete picture. Without that middle path, ambiguity resolves toward whichever outcome is operationally easier, which is rarely the cautious one.
What a patient can observe
A prospective patient completing an intake can notice several of these directly. Whether the questions name specific conditions. Whether a full medication list is requested. Whether the process can pause for clarification or only moves forward. Whether anything in the flow suggests a submission could result in a declined prescription.
An intake that takes four minutes and asks nothing a patient has to think about is telling them something about how it will be reviewed.
Friction that has not been resolved
State licensure fragments the map
Telehealth prescribing authority is granted state by state. A clinician licensed in one state cannot prescribe into another without appropriate licensure, so provider availability reflects a licensing footprint rather than clinical capacity. Patients in excluded states are excluded for administrative reasons that are invisible from the outside.
Follow-up is weaker than intake nearly everywhere
Enormous effort goes into the first interaction because it is the one that converts. The eighth month receives far less design attention. A patient whose circumstances change partway through treatment needs a channel back to a clinician, and the quality of that channel varies far more between providers than the quality of the intake does.
Comparison is difficult by construction
A prospective patient cannot inspect the clinical rigour of an intake before completing it, and cannot easily compare two providers’ review processes from their marketing. The observable signals are indirect: whether the provider names the licensure of its reviewers, whether it states that prescriptions can be declined, and what it commits to in writing about refunds and cancellation.
The useful reframe
Asynchronous prescribing is frequently discussed as though the question were whether it is as good as seeing someone in person. That framing produces unhelpful answers, because the realistic alternative for a large share of these patients is not an office visit. It is nothing.
The better question is whether a given asynchronous pathway is well constructed: whether the intake asks what a clinician would ask, whether a real licensed prescriber reviews it, whether declining is a genuine outcome, and whether anything exists after the first shipment.
Those are answerable questions, and the answers differ substantially between providers wearing identical marketing language. The technology is not the variable. The design is.