How Can I Test Whether AI Assistants Correctly Describe Telehealth Prescribing and State Availability?

Why do assistants get telehealth details wrong?

• Rules differ by state. Clinicians generally need to be licensed in the state where the patient is located, so availability depends on where your clinicians hold licences.
• Controlled substance rules are temporary and changing. In the US, pandemic-era flexibilities that allow some controlled substances to be prescribed by telemedicine without a prior in-person visit have been extended several times. The fourth extension runs until 31 December 2026. Assistants may describe old rules, or assume every provider prescribes everything allowed.
• Services vary by state and over time, as companies launch and withdraw services.
• Marketing copy is vague. "Get treatment online" does not say which conditions, which medications, or where.
• Compounded medications cause confusion. Compounded drugs are not FDA-approved, but reviews, ads and articles sometimes blur the difference.

What should the fact sheet contain?

Before testing, write down the correct answer for every question, per state:

Which questions should you test?

For each important state:

• "Can I get [medication] prescribed online in [state]?"
• "Is [company] available in [state]?"
• "Does [company] prescribe [controlled substance or class]?"
• "Do I need a video visit with [company] to get a prescription?"
• "Is [company]'s [medication] FDA-approved?"
• "Does [company] take insurance in [state]?"
• "How much does [company] cost for [service]?"

Run each several times in ChatGPT, Gemini, Perplexity, Copilot and Google's AI features, with location set to the state where possible.

How do you score the answers?

For each answer, mark each fact as correct, wrong, missing or overstated. Watch especially for:

• Availability overstated, such as "available nationwide".
• Prescribing overstated, such as naming controlled substances you do not prescribe.
• FDA status wrong, such as calling a compounded medication FDA-approved.
• Clinical steps missing, such as implying a prescription is guaranteed or available without assessment.
• Out-of-date prices or insurance status.

Report accuracy per state and per topic. A single overall score hides the state where assistants consistently get it wrong. How device regulatory status gets misdescribed is covered in whether AI can distinguish FDA-cleared from FDA-approved devices.

What should you publish?

• A state availability page, in text, listing each service and the states where it is available, and where it is not.
• A prescribing policy page stating what your clinicians can prescribe, what they do not prescribe, and that prescriptions depend on a clinical assessment.
• Medication pages that state clearly whether each medication is FDA-approved or compounded, with plain explanations.
• Clinician licensing information, such as where your clinicians are licensed, or a statement that care is provided by clinicians licensed in the patient's state.
• Insurance and pricing pages by service and state.
• A questions section that answers the test questions above in plain words.

Keep the same facts in app store listings, ads, partner pages and press materials, since assistants read those too. How online pharmacies check licensing descriptions is covered in how an online pharmacy can check whether AI describes its licensing accurately.

How often should you test?

Monthly for your main states, and immediately after any change in state availability, prescribing policy, medications, pricing or the regulatory rules that apply to you. Controlled substance rules in particular have changed at the end of several recent years. Recheck after each change takes effect, so assistants are not left describing an outdated position.

What might an illustrative case look like?

Consider an illustrative telehealth company offering weight management in 30 states, with brand-name and compounded medications, and no controlled substance prescribing. Testing showed assistants saying it was "available in all 50 states" and that it "offers FDA-approved semaglutide". The compounded version it sold in some states was not FDA-approved. Some answers also said it could prescribe ADHD medication. The cited sources included an affiliate review and an old press release. The company published a state availability page, medication pages stating FDA status plainly, and a prescribing policy saying it does not prescribe controlled substances. It asked the affiliate to correct its review. Two months later, state and medication answers were accurate in most runs.

What do people ask about telehealth and AI answers?

Should we list states where we are not available?

Yes. Stating where you are not available helps assistants avoid recommending you to patients you cannot treat.

Are affiliate reviews a common source of errors?

Often. They may exaggerate availability or medication claims. Ask for corrections and monitor them.

Can we prevent assistants recommending us for services we do not offer?

State clearly what you do not offer. Specific exclusions are among the most useful sentences for assistants.