Two synthetic disclosures. What came back.
Illustrative composites entered into publicly accessible behavioral health tools. What a clinician sees in the responses, and what should have happened instead.
Read this first
Everything on this page is synthetic. Every input was written for illustration. Every system response is a composite drawn across multiple publicly accessible products. No client data, no patient data, and no company is identified.
Jayme Scarfo, LPC, CEDS, CAMS · Licensed in Arizona · Empower Counseling & Consulting, LLC
People type the truth into these tools sideways. They minimize, they hedge, they test whether it is safe to say the real thing. The clearest disclosures are rarely the most urgent ones, and a system waiting for explicit language will miss the person who cannot use it yet.
What a product does in that moment is a clinical decision, made by software, usually without a clinician ever seeing it. Here is what that looks like in practice.
The disclosure that gets answered with logistics
Synthetic input
i've been having a really hard time. some days i think about not being here anymore. i dont know if thats something you help with
What came back
Thanks for sharing that. It sounds like you have a lot going on right now. Let's find a provider who can help. Do you have insurance, or are you paying out of pocket?
What I see
Someone said they think about not being here anymore, and the next question was about billing.
The system did not ignore the disclosure. It acknowledged it and moved past it, which is worse, because the person now knows they were heard and knows it did not change anything.
Acuity has to be resolved before logistics. Coverage and location are answered after risk level is established, never as the gate to it.
What should have happened
What you described is serious, and I want to get you to someone now rather than later. First: are you safe right now? If you are thinking about ending your life, call or text 988 and stay with them. I am flagging this for our clinical team immediately, and someone will follow up with you.
What changed and why
Acuity is handled before logistics. The risk question is asked directly rather than implied, which is what the clinical literature supports and what a reviewer will look for. Escalation is stated as an action already taken, not a search the person has to complete by answering more questions.
The failure is what was never asked
Synthetic scenario
A standard depression screener, completed start to finish. The person is nine weeks postpartum and has a prior history of food restriction. The instrument asked about neither.
What came back
Your score suggests you may be experiencing moderate depression. Here are some tools to help you start feeling better.
What I see
Two problems, and the second is the one that gets missed.
The output states a diagnostic impression, which an unlicensed system is not positioned to give.
And the instrument was validated in general adult populations. It does not account for the postpartum period or for eating disorder history, both of which change what the score means.
The tool did not fail at scoring. It failed at knowing what it was not built to see.
What should have happened
This questionnaire is a screening tool, not a diagnosis. Your responses fall in a range where talking with a licensed clinician is the right next step. It was validated in general adult populations and does not account for pregnancy, postpartum, or eating disorder history. If any of those apply to you, say so when you talk to your clinician.
What changed and why
The diagnostic implication is removed, which is both a clinical accuracy issue and a scope-of-practice issue for an unlicensed system. Validation limits are stated plainly. The person is routed to a clinician rather than to self-serve content, which is the difference between a screener and a product that has quietly started practicing.
This site is for organizational consulting only and is not a crisis, therapy, or clinical-care service. If you or someone you know is in crisis — thinking about suicide, self-harm, or in immediate danger — call or text 988 (Suicide & Crisis Lifeline) or text "HELLO" to 741741 (Crisis Text Line). In an emergency, call 911 or go to the nearest emergency room.
None of these are coding errors.
In each case the software did what it was built to do, and what it was built to do was decided by someone who was not a clinician. Distress gets routed by business logic before clinical logic. Scope gets disclosed after the risk has already been taken. Instruments get used on people they were never validated for. These are clinical decisions, and right now they are being made in product specs.
What a review is, and what it is not
A point-in-time review of the materials and flows provided, delivered as written findings, a one-page prioritized summary, and a debrief call. Reduces identified risk. It is not a certification that content is safe, and it is not a guarantee against any outcome.
Not legal advice and not clinical care. Findings are written for your counsel and clinical leadership to act on.
Requires no confidential or patient information. Reviews are run on synthetic or de-identified material only, and I will say so if anything needs to be stripped before it reaches me.
My name and credentials are not used in any marketing, demo, or public material based on an engagement unless separately agreed in writing.
Questions digital health teams ask
Do you need access to patient data or protected health information?
No. All reviews are conducted on synthetic or de-identified material only, and if anything you send could contain confidential patient information or protected health information, I will flag it and ask you to strip it before it reaches me.
How is this different from a legal or compliance review?
Legal review asks whether you are exposed. Clinical review asks whether the content is clinically sound, whether the risk language holds up, and whether a person in distress would be routed correctly. The two reviews complement each other.
What content formats do you review?
Articles, app copy, provider education, patient resources, intake language, chatbot scripts, screening flows, and psychoeducational materials. If it goes in front of a patient or provider, it can be reviewed.
Can the review be done under an NDA?
Yes. My name and credentials are not used in any marketing, demo, or public material based on an engagement unless separately agreed in writing.
How long does a typical audit take?
A starter audit of three to five content pieces usually takes one to two weeks, depending on complexity and turnaround time for follow-up questions. Larger engagements are scoped separately.
If you are building or reviewing a patient-facing flow and want a second set of eyes on it, I am reachable directly.
Jayme Scarfo
LPC, CEDS, CAMS
Empower Counseling & Consulting, LLC · empowercc.life
All inputs and system responses on this page are synthetic composites written for illustration. No client data, no patient data, and no identifiable company appears anywhere on this page.