Non-Commercial Research

GLP-1 Telehealth Prescribing Practices: 2026 Data, Screening Questions, and Standards

Not medical or legal advice. The RX Index is an independent research and reference resource. This page carries no advertising, affiliate links, sponsored placements, provider comparisons, or treatment recommendations.

Between August and December 2025, a Yale researcher using one clinically eligible simulated-patient profile applied for a GLP-1-based prescription at 49 online sellers. Forty-five of them — 91.8% — issued one. Thirty-three, or 67.3%, required no spoken contact with a clinician. Two compounded prescriptions were issued in five minutes or less. The final results were published in JAMA on July 6, 2026, and they describe what those sellers did with that one profile.

The page's central original calculation is a comparison, not a new clinical score. All 49 sites collected height and weight, while 18 — 36.7% — requested any patient-reported blood pressure, glucose, cholesterol, or triglyceride value. That is a 63.3-percentage-point difference in questionnaire coverage, calculated by The RX Index from the published counts.

What makes this subject hard is that several different evidence systems answer several different questions. State statutes and board rules can be binding. FDA-approved prescribing information describes contraindications, warnings, and condition-specific monitoring for approved products. The Federation of State Medical Boards publishes a model policy that has force only where an authority adopts or applies it. Professional guidelines describe recommended clinical practice but are not statutes. None of those sources can be substituted for another.

This page keeps those layers separate on purpose. We built it as an evidence crosswalk: what sellers were observed doing, what websites publicly advertised, what current laws, labels, model policy, and clinical guidance actually establish, and — clearly marked — what The RX Index calculated from the published counts. Every percentage carries its denominator. Nothing is pooled that should not be.

49
Online GLP-1 sellers studied (Aug–Dec 2025)
91.8%
Issued a prescription (45 of 49)
67.3%
Required no spoken clinician contact (33 of 49)
86.7%
Prescriptions issued were compounded (39 of 45)
63.3 pp
Gap: height/weight coverage vs. clinical-values coverage
$217
Median all-in monthly price (2025 fieldwork; not current)

Evidence-crosswalk files: CSV · JSON — 82 records, four evidence layers, denominators and interpretation guardrails preserved in every row.

Headline findings from the 49-site operational study

Table 1. Headline findings — 49 online GLP-1 sellers, Aug–Dec 2025
Finding Result Denominator
Prescription issued45 of 49 — 91.8%All sites approached
No spoken clinician communication required33 of 49 — 67.3%All sites approached
Patient-reported clinical values requested18 of 49 — 36.7%All sites approached
Automatically charged and shipped after approval34 of 45 — 75.6%Sites that prescribed
Prescription issued was compounded39 of 45 — 86.7%Sites that prescribed

Source: Chetty AK, Chen AS, Ross JS, Ramachandran R. Online Prescribing of GLP-1 Receptor Agonists. JAMA. Published online July 6, 2026. doi:10.1001/jama.2026.9131. Fieldwork: August–December 2025; websites serving Connecticut; one eligible simulated patient.

The original RX Index calculation: All 49 sites collected height and weight, which are used to calculate body mass index. Eighteen sites asked for at least one value within the study's combined category of patient-reported blood pressure, glucose, cholesterol, or triglycerides. The difference between 100.0% and 36.7% is 63.3 percentage points.

That is a descriptive difference between two questionnaire categories. It is not a validated quality score. It does not prove that all four clinical values were requested by any one site, that every patient needed the same tests, or that collection of either category was sufficient for a complete evaluation.

What this data shows — and what it doesn't

It shows how 49 online sellers handled one eligible simulated applicant: what they asked, whether they required spoken contact, how quickly they issued prescriptions, what type of medication they prescribed, and whether they charged and shipped without another confirmation step.

It does not show national prevalence. The operational study used one simulated patient, one patient profile, one destination state, and a purposively selected sample. It did not measure follow-up care, adherence, adverse events, weight change, laboratory outcomes, or any other patient outcome. It did not determine that any individual prescription was lawful, unlawful, appropriate, inappropriate, safe, or unsafe.

And "no spoken interaction" is not the same as "no clinician review." HHS describes asynchronous telehealth as a modality in which patients and providers share information at different times. Whether a particular asynchronous encounter is lawful and clinically adequate depends on the patient's location, the professional's authority to practice there, the facts of the encounter, and the applicable standard of care.

We are being precise about that because the distinction changes the meaning of the study.

What did the 2026 JAMA study find about GLP-1 telehealth prescribing practices?

Answer capsule: A researcher using one eligible simulated-patient profile approached 49 online GLP-1 sellers between August and December 2025. Forty-five issued a prescription and 34 mailed medication; the median time to prescription was one day or less, and two compounded prescriptions were issued in five minutes or less. Thirty-three of the 49 sites required no spoken communication with a clinician.

How the study was built

The research team constructed one simulated profile: a 27-year-old man with a body mass index of 35 and reported hypertension, high cholesterol, and sleep apnea. This was not a borderline BMI profile. It was one clearly eligible profile chosen to test the process rather than to test how sellers handled a contraindication, a disputed indication, or a complicated medical history.

The researcher took that same profile through intake, payment, prescribing, and fulfillment at 49 websites serving Connecticut. Five sites offered branded GLP-1s only, 17 offered compounded GLP-1s only, and 27 offered both. The work was published as a research letter in JAMA on July 6, 2026 by Ashwin K. Chetty, Alissa S. Chen, Joseph S. Ross, and Reshma Ramachandran.

The design matters. A secret-shopper study measures process, not outcome. It can show what an applicant was asked to do and what happened next. It cannot show whether a prescription later helped or harmed a patient.

Full operational results

Table 2. Full operational results — 49 online GLP-1 sellers, Aug–Dec 2025
Measure Count Percentage Denominator
Offered compounded GLP-1s only1734.7%All 49 sites
Offered branded GLP-1s only510.2%All 49 sites
Offered both branded and compounded GLP-1s2755.1%All 49 sites
Used an intake questionnaire49100.0%All 49 sites
Collected height and weight49100.0%All 49 sites
Asked about medical conditions4898.0%All 49 sites
Asked about medications and allergies4693.9%All 49 sites
Asked about weight-loss goals3979.6%All 49 sites
Asked about prior nonpharmacologic attempts3265.3%All 49 sites
Asked about eating-disorder history2755.1%All 49 sites
Asked about diet and physical activity2653.1%All 49 sites
Asked questions supporting compounded-formulation personalization2244.9%All 49 sites
Requested patient-reported blood pressure, glucose, cholesterol, or triglycerides1836.7%All 49 sites
Asked whether the applicant had a primary care clinician918.4%All 49 sites
Required a video visit1326.5%All 49 sites
Required a telephone call36.1%All 49 sites
Required no spoken clinician communication3367.3%All 49 sites
Issued a prescription4591.8%All 49 sites
Mailed medication3469.4%All 49 sites
Issued prescription was compounded3986.7%45 sites that prescribed
Issued prescription was branded613.3%45 sites that prescribed
Automatically charged and shipped after approval3475.6%45 sites that prescribed
Prescribed despite incomplete requested photo evidence920.0%45 sites that prescribed
Issued a prescription within five minutes or less24.4%45 sites that prescribed
Mailed medication was compounded3294.1%34 sites that mailed medication
Mailed medication was branded25.9%34 sites that mailed medication

Source: Chetty AK et al. JAMA. July 6, 2026. doi:10.1001/jama.2026.9131. Percentages retain the study's denominators: 49 sites approached, 45 sites that prescribed, or 34 sites that mailed medication.

Timing, supply, price, and cross-platform prescribing

Table 3. Timing, supply, price, and cross-platform prescribing
Measure Study result
Median time from request to prescription1 day or less; IQR: 1 day or less to 2 days
Prescriptions issued within five minutes or less2, both compounded
Median medication supply mailed1 month
Median all-in monthly price$217.33; IQR: $159.36–$289.24
Clinicians who prescribed to the same simulated patient through two or more websites3 clinicians
Sites that did not issue a prescription4 of 49 — 8.2%

Source: Chetty AK et al. JAMA. July 6, 2026. doi:10.1001/jama.2026.9131. Price includes medication, appointment, and membership charges during 2025 fieldwork.

The $217.33 figure is a study-period price, not a current market-price claim. Compounded and branded pricing, shortage conditions, and access pathways have changed since the fieldwork. The GLP-1 Price and Access Tracker carries current figures and dates them separately.

Among the four sites that did not complete a prescription, two stopped the process at a request for bloodwork. The study also reported separate interruptions involving duplicate-prescription information and a photo or weight mismatch. Those are process findings from one simulated profile, not estimates of how often safeguards succeed across the market.

Why the final publication controls

Secondary recaps of this study have used slightly different wording and, in some cases, different figures. Where any recap conflicts with the final journal publication, the final publication is the source of record. The dataset preserves that rule at the row level: final-publication values control, denominators stay attached, and unresolved claims do not enter the published table.

What questions did online GLP-1 sellers ask before prescribing?

Answer capsule: Every one of the 49 sites used an intake questionnaire, and nearly all asked about medical conditions (98.0%) and medications or allergies (93.9%). Coverage was less consistent for eating-disorder history (55.1%), diet and physical activity (53.1%), and patient-reported blood pressure, glucose, cholesterol, or triglycerides (36.7%). All 49 collected height and weight.

Read the questionnaire-coverage ladder top to bottom and a more exact pattern falls out. Basic anthropometrics and broad medical-history categories were nearly universal. Several more specific domains were collected by only about half of sites or fewer.

Table 4. Questionnaire coverage ladder — 49 sites
Question category Sites collecting Percentage
Height and weight49 of 49100.0%
Medical conditions48 of 4998.0%
Medications and allergies46 of 4993.9%
Weight-loss goal39 of 4979.6%
Prior nonpharmacologic weight-loss attempts32 of 4965.3%
Eating-disorder history27 of 4955.1%
Diet and physical activity26 of 4953.1%
Compounded-formulation personalization22 of 4944.9%
Patient-reported blood pressure, glucose, cholesterol, or triglycerides18 of 4936.7%
Whether the applicant had a primary care clinician9 of 4918.4%

Source: Chetty AK et al. JAMA. July 6, 2026. doi:10.1001/jama.2026.9131. All percentages use the 49-site denominator.

Three lower-coverage categories are worth naming precisely.

Eating-disorder history — 55.1%. Just over half of the sites asked. The study measured whether the question appeared; it did not test how the answer would change a prescribing decision or whether a separate assessment occurred later.

Diet and physical activity — 53.1%. Slightly over half asked. Current professional guidance places nutrition, physical activity, and behavioral care inside comprehensive obesity treatment. Asking about those subjects is not the same as delivering that care, and the study did not measure whether any site provided or coordinated it.

Patient-reported clinical values — 36.7%. Fewer than four in ten sites asked for at least one value within the combined category of blood pressure, glucose, cholesterol, or triglycerides. These were patient-reported values, not necessarily laboratory results. A remembered blood-pressure reading and a newly ordered laboratory panel are different evidence, and this study did not merge them.

The RX Index derived measures

These are arithmetic on the published counts. Each formula is shown so the result can be reproduced without accepting an editorial score.

Table 5. The RX Index derived measures — computed from published counts
Derived measure Formula Result
Height/weight coverage minus patient-reported clinical-values coverage100.0% − 36.7%63.3 percentage points
Medical-condition coverage minus eating-disorder-history coverage98.0% − 55.1%42.9 percentage points
Medication/allergy coverage minus diet/physical-activity coverage93.9% − 53.1%40.8 percentage points
Sites requiring any spoken clinician interaction(13 video + 3 phone) ÷ 4932.7%

Source: The RX Index calculations from counts published in Chetty AK et al., JAMA, July 6, 2026. These are descriptive coverage contrasts, not validated clinical-quality scores and not evidence that every applicant required an identical evaluation.

How much clinician contact did online GLP-1 prescribing sites require?

Answer capsule: Thirteen of the 49 sites required a video visit and three required a telephone call — meaning 16 sites, or 32.7%, required some spoken interaction with a clinician. Thirty-three sites, 67.3%, required none. A site requiring no spoken interaction may still have had a clinician review the intake asynchronously; the study measured the mode of contact, not whether review occurred.

Table 6. Clinician contact requirements — 49 sites
Interaction requirement Count Percentage
Video required13 of 4926.5%
Telephone required3 of 496.1%
Any spoken interaction required — The RX Index calculation16 of 4932.7%
No spoken interaction required33 of 4967.3%

Source: Chetty AK et al. JAMA. July 6, 2026. doi:10.1001/jama.2026.9131. The 32.7% combined figure is The RX Index's sum of the published video and telephone counts.

What "no spoken interaction" means — and what it doesn't

This is the finding most likely to be misquoted, so the wording matters.

It means: at 33 sites, the simulated patient reached a prescription without a required voice or video conversation with a clinician.

It does not mean: no clinician reviewed the intake. HHS recognizes asynchronous telehealth, including patient intake and later provider review. Connecticut's telehealth statute also defines asynchronous communication within its telehealth framework. Recognition of the modality does not decide whether a particular encounter complied with state law or met the standard of care.

It also does not mean the reverse: a live call does not establish that an evaluation was complete. Modality is one process feature, not a quality verdict in either direction.

The study reported other process observations that should stay separate from the contact count: three clinicians prescribed to the same simulated patient through two or more websites; nine sites prescribed after incomplete requested photo evidence; and two compounded prescriptions were issued within five minutes. Those observations raise different questions about verification, cross-platform information, and review time. The study did not adjudicate the adequacy of any individual encounter.

What do laws, labels, model policy, and clinical guidance establish?

Answer capsule: There is no single national checklist called "the GLP-1 telehealth standard." State law can be binding; FDA-approved labels describe product-specific contraindications, warnings, and monitoring; the Federation of State Medical Boards publishes a nonbinding model policy; and the Obesity Association publishes professional clinical guidance. The operational study did not collect enough information to decide compliance with most of those sources.

This is the section most summaries of the JAMA study skip. A process finding and a compliance finding are not the same thing. To understand the difference, the authority and the measurement have to remain visible at the same time.

Layer 1: Binding law and professional authority State statutes, licensing rules, prescribing law, and the profession's standard of care where the patient is located.
Layer 2: FSMB model policy Nonbinding model unless adopted by a state board. Addresses static questionnaires, informed consent, documentation, and continuity.
Layer 3: Professional obesity-treatment guidance 2026 Obesity Association guidance: comprehensive evaluation, laboratory testing, and follow-up. Not a statute.
Layer 4: FDA-approved prescribing information Product-specific contraindications, warnings, dosing, and monitoring for Wegovy and Zepbound. Not a universal intake protocol.

Layer one: binding law and professional authority

The practice of medicine generally occurs where the patient is located. HHS lists the main cross-state pathways: a full license, a temporary-practice law, reciprocity, a compact, a telehealth registration, or another state exception. A website's availability in a state does not by itself establish that every clinician involved was authorized to practice there.

State law may also govern informed consent, patient-provider relationship formation, recordkeeping, prescribing, and the standard of care. Those rules are not uniform, and many turn on facts the operational study did not report.

Layer two: the FSMB model policy

The Federation of State Medical Boards adopted The Appropriate Use of Telemedicine Technologies in the Practice of Medicine in April 2022. It is a model policy, not a nationwide statute or regulation. Its force depends on whether and how a state board adopts, cites, or applies it.

The model policy says a documented medical evaluation and relevant clinical history must support diagnosis and treatment. It says diagnosis, prescribing, or treatment based solely on a static online questionnaire, or on an intake that fails to obtain information necessary to meet the applicable standard of care, is not acceptable. It contrasts that with an adaptive questionnaire that can ask follow-up questions or obtain more history.

That distinction matters, but the operational study did not classify any questionnaire as static or adaptive. It did not report whether a clinician could send follow-up questions, request outside records, arrange testing, or refer the patient elsewhere. The fact that all 49 sites used questionnaires therefore cannot be converted into an FSMB compliance rate.

The model policy also addresses clinician identification, informed consent, documentation, privacy, and continuity. Those are meaningful domains. They were not measured in the 49-site study.

Source: Federation of State Medical Boards, The Appropriate Use of Telemedicine Technologies in the Practice of Medicine, adopted April 2022.

Layer three: professional obesity-treatment guidance

The Obesity Association, a division of the American Diabetes Association, published Pharmacologic Treatment of Obesity in Adults: Standards of Care in Overweight and Obesity on January 13, 2026. It is professional clinical guidance, not a statute.

The guidance calls for a comprehensive evaluation before prescribing an obesity medication, including a relevant history, physical examination, and laboratory testing for obesity-related diseases and complications that may inform medication selection. It places medication inside a broader care plan that includes nutrition, physical activity, and behavioral therapy. It recommends early follow-up at least monthly during the first three months of treatment, and it does not recommend use of compounded products that are not FDA approved.

The operational study can be placed beside those recommendations, but it cannot be scored against them. Asking about diet is not the same as providing behavioral treatment. Requesting a patient-reported value is not the same as performing a physical examination or obtaining laboratory testing. Prescribing a compounded product identifies the product category; it does not, by itself, establish the circumstances of compounding or whether a clinician followed or rejected a professional recommendation.

Source: American Diabetes Association Professional Practice Committee for Obesity, Pharmacologic Treatment of Obesity in Adults: Standards of Care in Overweight and Obesity, published January 13, 2026; doi:10.2337/doci25-0008.

Layer four: FDA-approved prescribing information

The current Wegovy prescribing information and Zepbound prescribing information contain product-specific contraindications, warnings, dosing instructions, and monitoring directions.

Both labels list personal or family history of medullary thyroid carcinoma and multiple endocrine neoplasia syndrome type 2 as contraindications. Both state that the products are not recommended in patients with severe gastroparesis. The labels include condition-specific monitoring, including blood-glucose monitoring in patients with diabetes and renal monitoring when adverse reactions could cause volume depletion. Wegovy's current label says to discontinue at least two months before a planned pregnancy. Zepbound's current label contains oral-hormonal-contraception instructions for four weeks after initiation and after each dose escalation.

A label is not a universal intake questionnaire. The current labels reviewed do not specify one identical baseline laboratory panel for every patient. They also do not erase the clinician's obligation to evaluate the individual patient under applicable law and professional standards.

The RX Index evidence-to-measurement matrix

This table is the page's original standards crosswalk. It does not turn adjacent facts into compliance findings. It shows exactly where the study provides a direct measure, an incomplete proxy, or no measure at all.

Table 7. Evidence-to-measurement matrix — version 1.0
Source statement or domain Evidence type Closest operational-study measure What the study can establish
Clinician must be authorized to practice where requiredState law and licensureNot measuredNo conclusion about clinician licensure or registration
First-interaction notice and consent duties under Connecticut lawBinding state lawNot measuredNo conclusion about notice, consent, or documentation
Ask about consent to disclose records to the patient's primary care provider under Connecticut lawBinding state lawAsked whether applicant had a primary care clinician: 9 of 49The question count does not establish the statutory consent or record-transfer step
Treatment based solely on a static questionnaire is not acceptable under the FSMB model policyNonbinding model policy unless adoptedAll 49 used a questionnaireThe study did not classify questionnaires as static or adaptive and cannot establish compliance
Adaptive questionnaires must allow follow-up questions or further history under the FSMB model policyNonbinding model policy unless adoptedNot measuredNo conclusion about follow-up capability
Clinician identity, credentials, consent, and documentationModel policy and state-law domainsNot measuredNo conclusion
Comprehensive pretreatment history, physical examination, and relevant laboratory evaluationProfessional clinical guidance18 of 49 requested patient-reported clinical valuesThe study measured a questionnaire category, not a physical examination or laboratory evaluation
Nutrition, physical activity, and behavioral therapy within comprehensive careProfessional clinical guidance26 of 49 asked about diet and physical activityAsking a question does not establish that treatment or referral was provided
Early follow-up after medication initiationProfessional clinical guidanceNot measuredNo conclusion about follow-up care
Product-specific contraindications and condition-specific monitoringFDA-approved labelsBroad medical-condition and medication questions were measuredThe published categories are not detailed enough to assess label-level screening
Compounded products not recommended by the 2026 professional guidanceProfessional clinical guidance39 of 45 issued prescriptions were compoundedThe study establishes medication type, not guideline compliance or the facts governing compounding
Freedom to fill elsewhere in the FSMB model policy's physician-owned-pharmacy and sale-of-goods discussionNonbinding model-policy domain unless adopted34 of 45 were automatically charged and shippedAutomatic fulfillment does not reveal whether an outside pharmacy option existed

Sources: The RX Index Evidence-to-Measurement Matrix, version 1.0; FSMB 2022 model policy; Connecticut General Statutes § 19a-906; 2026 Obesity Association guidance; current Wegovy and Zepbound prescribing information; operational measures from Chetty AK et al., JAMA, 2026. "Not measured" is not a finding of compliance or noncompliance.

Is a video visit legally required for a GLP-1 telehealth prescription?

Answer capsule: No single nationwide rule identified in the federal sources reviewed requires live video before every noncontrolled semaglutide or tirzepatide prescription. The current Wegovy and Zepbound listings identify their DEA schedule as "None," while the DEA telemedicine flexibilities extended through December 31, 2026 concern Schedule II–V controlled medications. State law, licensure, patient location, relationship formation, consent, and the applicable standard of care still control.

The controlled-substance distinction

The federal telemedicine debate often centers on the Ryan Haight Act and DEA rules because those authorities govern controlled-substance prescribing. The DEA's fourth temporary extension runs through December 31, 2026 and addresses remote prescribing of Schedule II–V controlled medications.

The current DailyMed listings for Wegovy and Zepbound state DEA Schedule: None. That means the temporary controlled-medication flexibilities are not the rule that determines whether those products can be prescribed through telehealth.

That does not leave a vacuum. State professional-practice law, licensing rules, prescribing law, informed-consent requirements, recordkeeping duties, and the standard of care remain in force. The right question is not merely "Was there video?" It is "What did the applicable law require for this clinician, this patient location, this medication, and this encounter?"

Five-state primary-source snapshot for noncontrolled prescribing

This is a targeted snapshot, not a fifty-state survey. Each row was checked against the cited state source on August 1, 2026. The categories are deliberately narrow because the statutes do not use one shared vocabulary.

Table 8. Five-state primary-source snapshot — noncontrolled prescribing, verified August 1, 2026
State What the cited primary source says What it means for a questionnaire-only claim
California An appropriate prior examination need not be synchronous and may be achieved through telehealth, including a self-screening tool or questionnaire, provided the licensee complies with the appropriate standard of care. California does not make a questionnaire automatically sufficient; it permits the modality subject to the standard of care.
Connecticut The statute expressly includes asynchronous communication within telehealth, requires authorized practice and the profession's standard of care, and imposes first-interaction notice, consent, and primary-care-record-disclosure questions. Absence of video is not the whole legal analysis; the study did not measure several Connecticut-specific duties.
Arkansas A professional relationship is not established only through an internet questionnaire, email, patient-generated history, text, fax, or a combination of those. A separate statutory route allows an Arkansas-licensed professional with access to a maintained health record to use appropriate technology, including telephone, and, if clinically appropriate, prescribe a noncontrolled drug. Arkansas bars the broad claim that an internet questionnaire alone always establishes the relationship, but current law is not a universal live-audio-and-video mandate for noncontrolled prescribing.
Alabama A physician-patient relationship may form without a prior in-person examination. The law requires the same duty of care as in person, identity and location verification, clinician disclosure, consent, and other duties. More than four telehealth services in 12 months for the same unresolved condition generally triggers an in-person visit or referral within 12 months, subject to statutory exceptions. Alabama permits relationship formation without a prior in-person examination but imposes encounter and longitudinal-care duties not measured by the study.
Arizona Except for specified exceptions and Schedule II drugs, a board generally may not require an in-person examination before prescribing; a physical or mental-status examination may occur during telehealth. Informed consent and state practice rules still apply. Arizona rejects a blanket in-person-exam rule for non-Schedule II prescribing, but it does not make every form-only prescription adequate.

Primary sources: Cal. Bus. & Prof. Code § 2242(a); Conn. Gen. Stat. § 19a-906; Arkansas Act 829 of 2021, codified in Ark. Code §§ 17-80-402 and 17-80-403; Ala. Code §§ 34-24-703 and 34-24-704, enacted by Act 2022-302; Ariz. Rev. Stat. § 36-3602.

Connecticut, specifically

The operational study approached websites serving Connecticut, so Connecticut law is the most directly relevant state-law context for the simulated encounters.

Connecticut General Statutes § 19a-906 expressly contemplates asynchronous telehealth. It requires a telehealth provider to be appropriately licensed, certified, or registered and to conform to the profession's standard of care. At the first telehealth interaction, the provider must inform the patient about treatment methods and the limitations of the telehealth platform, obtain consent, and document the notice and consent. The provider must also ask whether the patient consents to disclosure of telehealth records to the patient's primary care provider and, if the patient consents, provide those records in a timely manner.

The JAMA study did not measure those duties. It did not report whether each clinician was authorized in Connecticut, what notice was shown, whether consent was documented, or whether records were offered to a primary care clinician. The absence of those measurements is not evidence that the duties were satisfied or violated. It is a map of what the study cannot answer.

This is an educational summary, not a fifty-state legal survey and not legal advice. A state-specific conclusion requires checking the current statute, regulations, licensing-board guidance, profession, patient location, and encounter facts.

Is bloodwork always required before a GLP-1 prescription?

Answer capsule: No universal baseline laboratory panel is specified for every patient in the current Wegovy and Zepbound prescribing information reviewed. Those labels contain patient- and condition-specific monitoring instructions instead. Separately, the 2026 Obesity Association professional guidance calls for relevant laboratory testing as part of a comprehensive pretreatment evaluation.

Those statements answer different questions, and they are not in conflict.

The product label is not a universal intake protocol. FDA-approved prescribing information identifies contraindications, warnings, dosing, and monitoring for specified circumstances. The current Wegovy and Zepbound labels reviewed do not name one identical baseline laboratory panel for every person before treatment.

Professional guidance can recommend a broader evaluation. The 2026 Obesity Association guidance calls for a relevant history, physical examination, and laboratory testing for obesity-related diseases and complications that may affect medication selection. That is a clinical recommendation. It is not a nationwide statute, and it does not mean every patient receives the same panel.

The accurate sentence is: No universal baseline panel was identified in the current labels reviewed. The inaccurate sentence is: Bloodwork is never required before a GLP-1 prescription. A patient's conditions, medications, symptoms, pregnancy status, diabetes status, kidney risk, and other facts can change what evaluation or testing is appropriate.

Two published findings add context without answering the universal-rule question. Two of the four sites that did not complete a prescription stopped the process at a request for bloodwork. In the separate 79-site advertising study, 8 websites — 10.1% — stated that labwork was required.

Keep these terms distinct:

  • Patient-reported clinical values — the applicant enters a remembered or known value.
  • Laboratory testing — a specimen is collected and analyzed.
  • Photograph or scale evidence — an image is submitted in response to a platform request.
  • Identity verification — the service confirms who the applicant is.
  • Duplicate-prescription checking — a service tries to identify overlapping prescribing across channels.

The operational study measured pieces of these domains separately. There is no defensible single "medical verification" percentage.

Sources: current Wegovy prescribing information; current Zepbound prescribing information; 2026 Obesity Association guidance; operational and advertising studies cited above.

What happened after an online GLP-1 prescription was approved?

Answer capsule: Thirty-four of the 45 sites that issued a prescription — 75.6% — automatically charged the payment method and shipped medication after approval without another patient confirmation step. Thirty-nine of the 45 prescriptions issued were compounded, and 34 of all 49 sites ultimately mailed medication.

The fulfillment stage contains a denominator trap because the number 34 appears twice.

Table 9. Fulfillment measures — note two separate 34-site figures with different denominators
Fulfillment measure Count Percentage Denominator
Automatically charged and shipped after approval3475.6%45 sites that prescribed
Prescription issued was compounded3986.7%45 sites that prescribed
Prescribed despite incomplete requested photo evidence920.0%45 sites that prescribed
Mailed medication3469.4%All 49 sites approached
Mailed medication was compounded3294.1%34 sites that mailed medication
Mailed medication was branded25.9%34 sites that mailed medication

Source: Chetty AK et al. JAMA. July 6, 2026. doi:10.1001/jama.2026.9131. The two 34s are different measures with different denominators.

Quoting "34 sites" without the denominator is not enough. Thirty-four of 45 automatically charged and shipped after approval. Thirty-four of 49 ultimately mailed medication. Applying the wrong denominator changes the percentage and the meaning.

The photo finding also needs exact wording. Nine sites issued a prescription after the simulated patient supplied only an upper-body photograph when a full-body or scale photograph had been requested. The study documents that the requested evidence was incomplete. It does not establish why the photo was requested, who reviewed it, or whether another verification step occurred.

What did compounded GLP-1 websites publicly say they required?

Answer capsule: A separate study of 79 websites selling compounded GLP-1s, published in JAMA Health Forum on January 17, 2025, found that 98.7% stated some clinician involvement was required, 39.2% stated a phone or video visit was required, and 10.1% stated labwork was required. These are advertised practices captured between July and September 2024 — statements websites made about themselves, not observations from prescription attempts.

This is a different evidence layer: different sample, different fieldwork window, different method, and different denominators. It belongs beside the operational study, not inside it.

Publicly stated practices: the 79-site advertising study

Table 10. Advertised practices — 79 compounded GLP-1 websites, July–Sept 2024
Advertised characteristic Count Percentage
Any clinician involvement stated78 of 7998.7%
Questionnaire reviewed by a clinician stated50 of 7963.3%
Messaging with a clinician stated3 of 793.8%
Telephone or video visit stated31 of 7939.2%
Virtual contact stated but method unspecified20 of 7925.3%
In-person visit stated5 of 796.3%
Labwork stated as required8 of 7910.1%
Mentioned compounded status at least once68 of 7986.1%
Consistently identified the medication as compounded13 of 7916.5%
Provided a definition of compounding39 of 7949.4%
Referred to compounded medication as generic7 of 798.9%
Stated compounded medication was not FDA approved34 of 7943.0%
Stated or implied compounded medication was FDA approved29 of 7936.7%
Claimed accreditation, certification, or equivalent for a partner compounding pharmacy58 of 7973.4%
Linked to an FDA-approved product label18 of 7922.8%
Mentioned adverse effects68 of 7986.1%
Mentioned contraindications50 of 7963.3%
Mentioned warnings and precautions49 of 7962.0%
Cited clinical-trial evidence to support safety claims13 of 7916.5%
Cited clinical-trial evidence to support efficacy claims65 of 7982.3%
Made an efficacy claim outside FDA-approved branded labels32 of 7940.5%
Discussed effects of discontinuation29 of 7936.7%

Source: Chetty AK, Chillakanti M, Ramachandran R, Ross JS, Chen AS. Online Advertising of Compounded Glucagon-Like Peptide-1 Receptor Agonists. JAMA Health Forum. 2025;6(1):e245018. doi:10.1001/jamahealthforum.2024.5018. Fieldwork: July–September 2024; websites identified through Google Shopping sponsored results. Open access under CC BY.

Product mix and advertised first-month prices

During the fieldwork window, the researchers identified 98 unique websites selling any GLP-1 receptor agonist. Seventy-nine sold compounded GLP-1s or prescriptions for them and entered the analysis. All 79 offered compounded semaglutide, 57 offered compounded tirzepatide, and 3 offered compounded liraglutide. Twenty-three also offered a branded GLP-1, and 29 advertised a compounded GLP-1 with a supplement.

Table 11. Product mix — 79 compounded GLP-1 websites, 2024 advertising study
Product advertised Websites Percentage of 79-site sample
Compounded semaglutide79100.0%
Compounded tirzepatide5772.2%
Compounded liraglutide33.8%
Branded GLP-12329.1%
Compounded GLP-1 with a supplement2936.7%

Source: Chetty AK et al. JAMA Health Forum. 2025;6(1):e245018.

Table 12. Median advertised first-month prices — 2024 fieldwork (not current prices)
Compounded product Median advertised first-month price Interquartile range
Semaglutide$231$189–$294
Tirzepatide$330$275–$399
Liraglutide$248$223–$274

Source: Chetty AK et al. JAMA Health Forum. 2025;6(1):e245018. Prices included discounts and reflect July–September 2024 advertising, not current market prices. Only three sites offered compounded liraglutide.

The disclosure-consistency gap

Sixty-eight of 79 websites mentioned at least once that the medication was compounded. Only 13 identified it as compounded consistently across the website. The RX Index calculates a 69.6-percentage-point difference between at-least-once disclosure and consistent disclosure: 86.1% − 16.5%.

Another way to state the same denominator honestly is that 66 of 79 websites — 83.5% — did not consistently identify the medication as compounded. That is not a legal determination about any current website. It is a descriptive result from the 2024 advertising sample.

Fifty-two websites — 65.8% — displayed a certification mark, and 50 of those displayed LegitScript certification. The study separately found that 29 websites stated or implied their compounded medication was FDA approved. A certification display and an accurate FDA-status disclosure are different fields; one cannot be used as a substitute for the other.

Can the 49-site and 79-site studies be compared directly?

Answer capsule: No. The two studies used different samples, fieldwork periods, methods, and definitions, and the operational sample was drawn from websites the research team had previously identified. Their sample sizes cannot be added, and their percentages cannot be treated as a before-and-after trend.

This deserves its own section because the tempting comparisons are also the wrong ones.

Do not publish any of the following:

  • "128 sites were studied." The samples overlap by an unknown amount. Adding 49 and 79 does not describe a unique population.
  • "Oversight declined between 2024 and 2025." Neither study was designed as a longitudinal measurement, and the later sample was not an independent second census.
  • "Websites failed to deliver what they advertised." The studies did not match and test the same named sites as paired observations.
  • "67.3% had no clinician review." The measure was spoken communication, not review.
  • "The study proves these prescriptions were unsafe." Neither study measured patient outcomes.
  • "The study proves these prescriptions were unlawful." Neither study adjudicated legality.

What is valid is a within-study comparison using the same sample and denominator. The 63.3-point questionnaire-coverage difference comes from two measures collected across the same 49 sites. The 69.6-point disclosure-consistency difference comes from two measures collected across the same 79 sites. Neither is presented as a trend between years.

Why does compounded medication status matter in this data?

Answer capsule: Thirty-nine of the 45 prescriptions issued in the operational study were compounded. FDA does not approve compounded drugs or review them for safety, effectiveness, or quality before marketing, and the agency has warned telehealth companies against describing compounded drugs as approved generics, as the same as approved products, or as clinically proven to produce the same result.

Compounded status changes what the numbers mean for three separate reasons.

Regulatory status. A compounded drug is not FDA approved. That is categorically different from an FDA-approved product and from an FDA-approved generic. In the advertising study, 34 of 79 sites stated that the compounded medication was not FDA approved, while 29 stated or implied that it was FDA approved.

Disclosure consistency. Sixty-eight sites mentioned compounding at least once, but only 13 did so consistently. The descriptive gap is 69.6 percentage points, and the share not identifying the product consistently is 83.5%.

The legal category is fact-specific. "Compounded," "counterfeit," "unapproved," "fraudulent," and "unlawfully compounded" do not mean the same thing. FDA explains that federal law contains different pathways and conditions for compounding by state-licensed pharmacies, physicians, and registered outsourcing facilities. Whether a particular product was compounded lawfully depends on the compounder, the drug, the conditions met, and the applicable law.

FDA's current telehealth guidance says companies should not:

  • describe a compounded drug as a generic version of, or the same as, an FDA-approved drug;
  • claim that a compounded drug was FDA approved or evaluated for safety and effectiveness;
  • claim that it is clinically proven to produce the same result as an FDA-approved drug;
  • claim that a pharmacy or outsourcing facility is "FDA approved" or "FDA licensed"; or
  • use branding that falsely implies the telehealth company is the compounder.

Source: FDA, "FDA to Telehealth Companies: What to Know When Promoting Compounded Drugs".

On March 3, 2026, FDA announced 30 warning letters to telehealth companies concerning false or misleading claims about compounded GLP-1 products. Those letters concern promotion, not the adequacy of the clinical encounters measured in the JAMA operational study. The Compounded GLP-1 Enforcement Tracker maintains the separate enforcement record.

A warning letter states the agency's position and requested corrective action. It is not a court judgment, and the status of a particular matter can change after the letter is issued.

How did The RX Index produce this evidence crosswalk?

Answer capsule: We extracted final published values from two peer-reviewed studies, kept their samples and definitions strictly separate, and cross-referenced current agency materials, drug labels, telehealth guidance, professional guidance, and selected state statutes. Every percentage carries its denominator, every original calculation shows its formula, and claims that could not be verified were withheld from the article or marked nonpublishable in the claim ledger.

Source hierarchy

  1. Final peer-reviewed original research for study methods and results.
  2. Current statutes, regulations, and issuing-government materials for legal and regulatory claims.
  3. Current FDA-approved prescribing information for label claims.
  4. The issuing professional organization's publication for model policy and clinical guidance.
  5. Official institutional summaries and high-quality secondary reporting only to corroborate or clarify access-restricted material, never to override a final publication or primary legal source.

Where a recap and the final journal article disagreed, the journal article controlled. Where a legal summary and the primary statute differed, the statute controlled. Where a claim could not be resolved, it did not enter the final article as fact.

Evidence layers

Every record in the 82-row dataset is assigned to one of four layers:

  1. Observed operational practices — 31 records. What happened during prescription attempts in the 49-site study.
  2. Publicly stated advertising practices — 33 records. What websites said about themselves in the 79-site study.
  3. The RX Index derived measures — 5 records. Reproducible arithmetic using published counts.
  4. Standards and requirements crosswalk — 13 records. Current regulatory, label, licensure, and clinical statements with interpretation guardrails.

The five-state legal snapshot in this article is a separate editorial table. It is not represented as a completed fifty-state dataset and is not counted among the 82 records.

Denominator preservation

The operational study reports results against several denominators:

  • 49 sites approached;
  • 45 sites that issued a prescription; and
  • 34 sites that mailed medication.

We preserved the applicable denominator in every row. This matters because the same numerator can appear in different measures. Thirty-four sites automatically charged and shipped among the 45 that prescribed; 34 sites mailed medication among all 49 approached. Those are not interchangeable statistics.

Derived-measure rules

A The RX Index derived measure is published only when:

  • both inputs come from the same study and compatible denominator;
  • the arithmetic can be shown in one line;
  • the result is labeled as The RX Index's calculation;
  • the underlying source remains visible; and
  • the interpretation guardrail appears with it.

We did not calculate a combined 128-site total, a 2024-to-2025 trend, a compliance score, a safety score, or a national prevalence estimate because the evidence does not support those outputs.

What we verified firsthand

For this release, we checked: the 2026 JAMA article title, author list, publication date, DOI, and final reported values; the open-access JAMA Health Forum full text, tables, sample construction, product counts, advertising measures, prices, and certification-mark count; the April 2022 FSMB model policy, including its static-questionnaire language and definitions; current official state sources for California, Connecticut, Arkansas, Alabama, and Arizona; current HHS telehealth modality and cross-state licensure guidance; the current DailyMed listings for Wegovy and Zepbound; current FDA compounding, GLP-1 safety, telehealth-promotion, and warning-letter materials; and the 2026 Obesity Association professional guidance.

Current-verification claims carry an August 1, 2026 verification date. Static study findings retain their original fieldwork and publication dates.

Dataset fields and file integrity

Each dataset row contains a stable metric ID, evidence layer, domain, measure, numerator, denominator, percentage or value text, denominator definition, fieldwork dates, publication date, geography, source type, source title, source URL, DOI where available, derivation formula where applicable, last-verified date, interpretation guardrail, and notes.

The release is available in CSV and JSON. The claim ledger records the publication decision for consequential claims. SHA-256 hashes allow a saved file to be checked against the published version.

Corrections and versioning

Data version 1.0 is preserved as a dated release. A corrected or expanded release will receive a new version number and a dated change note rather than silently replacing the record. The site's corrections policy describes how material factual errors are handled.

Limitations: what this data does and does not show

We would rather state these plainly than bury them in a footnote.

One patient profile. The operational study used one simulated patient who met the study's eligibility criteria. It shows what happened to that profile at those sites. It cannot show how the same sellers would handle a patient with a contraindication, a marginal indication, a different state location, or a complicated history.

One destination state. The sites served Connecticut. This was not a national probability sample.

Purposive sampling. The operational sample was drawn from websites the research team had previously identified. The advertising sample came from Google Shopping sponsored results during a defined fieldwork window. Neither is a random census of the U.S. market.

Process, not outcomes. The studies did not measure follow-up care, adherence, adverse events, weight change, laboratory outcomes, or any other patient outcome.

Modality is not review. No spoken contact does not prove no clinician review. A live call does not prove adequate review. The studies did not validate the quality of any individual clinical evaluation.

No legal adjudication. Neither study determined whether a prescription complied with the law. The legal snapshot summarizes five primary-source examples; it is not a fifty-state legal survey. State statutes, regulations, and board guidance can change.

Different evidence authorities. A state statute, an FDA label, a model policy, and a professional guideline do not have the same legal effect. The article separates them because treating one as another would create a false compliance conclusion.

The two studies are not directly comparable. Different samples, periods, methods, and definitions. Their sample sizes cannot be added, and their percentages cannot be used as a trend.

The operational article is a research letter. Its compressed format limits the number of process details reported publicly. Absence of a measure in the article is not evidence that a site did or did not perform that step.

No universal label panel is not no evaluation. The current Wegovy and Zepbound labels reviewed do not specify one baseline laboratory panel for every patient. That does not mean laboratory testing is unnecessary for a particular person.

Patient-reported values are a composite category. The 18-site measure combines blood pressure, glucose, cholesterol, or triglycerides. It does not mean every one of those sites requested all four values.

Derived measures are descriptive. The 63.3-, 42.9-, 40.8-, and 69.6-percentage-point figures are arithmetic differences in coverage or disclosure. They are not validated quality instruments and have not been tested against outcomes.

Prices are historical. The $217.33 median all-in monthly figure comes from 2025 fieldwork. The advertising-study price medians come from July–September 2024. None is a current market quote.

Certification and FDA status are separate. A displayed certification mark does not establish that a product is FDA approved, that an advertisement is accurate, or that an encounter met the standard of care.

Compounding terminology matters. Compounded, unapproved, counterfeit, fraudulent, and unlawful are not interchangeable words.

This is educational content. It is not medical advice, legal advice, a provider assessment, or a substitute for a clinician who knows the patient's history or an attorney licensed in the relevant jurisdiction.

Frequently asked questions about GLP-1 telehealth prescribing practices

Can GLP-1 medications be prescribed through telehealth?

Yes. Noncontrolled semaglutide and tirzepatide products may be prescribed through telehealth when the clinician is authorized to practice for the patient's location and the encounter satisfies applicable law, relationship, consent, documentation, prescribing, and standard-of-care requirements. The exact rules vary by state and profession.

Does a GLP-1 prescription require a video visit?

There is no single nationwide live-video rule identified for every noncontrolled semaglutide or tirzepatide prescription. California expressly permits a prior examination through telehealth, including a self-screening tool or questionnaire, when the standard of care is met. Arkansas bars a relationship formed only through an internet questionnaire but also recognizes a health-record-and-appropriate-technology route for clinically appropriate noncontrolled prescribing. The applicable state law and encounter facts control.

Do you need bloodwork before starting a GLP-1?

The current Wegovy and Zepbound labels reviewed do not specify one universal baseline laboratory panel for every patient. The 2026 Obesity Association professional guidance calls for relevant laboratory testing as part of a comprehensive pretreatment evaluation. A clinician may order testing based on the patient's history, conditions, medications, symptoms, and the product's monitoring instructions.

How many online sellers issued a prescription in the 2026 JAMA study?

Forty-five of the 49 websites approached — 91.8% — issued a GLP-1-based prescription to the study's eligible simulated patient.

How quickly were prescriptions issued?

The median time from request to prescription was one day or less, with an interquartile range from one day or less to two days. Two compounded prescriptions were issued within five minutes or less.

How many sites required spoken contact with a clinician?

Sixteen of 49 — 32.7% — required either a video visit (13) or a telephone call (3). The remaining 33 sites — 67.3% — required no spoken interaction.

Does "no spoken contact" mean no clinician reviewed the prescription?

No. Asynchronous telehealth allows a clinician to review submitted information without a live conversation. The study measured whether spoken contact was required; it did not establish whether clinician review occurred or whether any review was adequate.

Are compounded GLP-1 medications FDA approved?

No. FDA does not approve compounded drugs or review them for safety, effectiveness, or quality before marketing. Compounded drugs are also not FDA-approved generic drugs. Whether a particular drug was compounded lawfully is a separate, fact-specific question.

Can the 49-site and 79-site studies be added together?

No. The samples overlap by an unknown amount and used different methods, time periods, and definitions. Adding them produces a number that does not describe a unique population.

Did the JAMA study prove that online GLP-1 prescribing is unsafe?

No. The study documented intake, prescribing, payment, and fulfillment processes for one eligible simulated patient. It did not measure follow-up care, adverse events, treatment outcomes, or the legality or clinical appropriateness of any individual prescription.

How to cite this page

Cite this page

The RX Index Editorial Team. "GLP-1 Telehealth Prescribing Practices: 2026 Data, Screening Questions, and Standards." The RX Index Research. Published August 1, 2026. Last updated August 1, 2026. https://therxindex.com/research/glp1-telehealth-prescribing-practices/

Cite the dataset

The RX Index Editorial Team. "The RX Index GLP-1 Telehealth Prescribing Practices Evidence Crosswalk." Version 1.0. The RX Index Research. Last verified August 1, 2026. https://therxindex.com/research/data/glp1-telehealth-prescribing-practices-v1.0.csv

Cite the evidence-to-measurement matrix

The RX Index Editorial Team. "The RX Index Evidence-to-Measurement Matrix for GLP-1 Telehealth Prescribing." Version 1.0. The RX Index Research. Last verified August 1, 2026. https://therxindex.com/research/glp1-telehealth-prescribing-practices/#evidence-to-measurement-matrix

Version history

2026-08-01 — v1.0 — Initial publication. Eighty-two-record evidence crosswalk covering 31 observed operational-practice records, 33 publicly stated advertising-practice records, 5 The RX Index derived measures, and 13 standards-and-requirements records. The article also includes the original Evidence-to-Measurement Matrix and a separate five-state primary-source legal snapshot. The legal snapshot is not presented as a completed fifty-state dataset.

Primary sources

Peer-reviewed research

Professional and clinical guidance

Statutes and enacted state sources

Federal agency and label materials