Where we're going
Trust by design: how every new format stays accurate
A new format is only useful if people can trust it. These are the rules every HealthLiteracyCopilot concept follows, from locked facts to labelled simulations.
The facts stay locked
Reviewers lock the facts once. Every format is checked against them, and a format that drops or changes a fact does not pass.
- Each output lists the locked facts it covers
- Additions come only from approved fact modules
- A diff against the core goes to the reviewer
- A wrong denominator, a dropped "not" or a new clinical claim blocks release
People review the real output
Clinical, compliance and language reviewers approve the audio, video or interactive piece itself, and what ships is exactly what they approved.
- Each output gets its own review
- Spanish and other languages get qualified human review before release
- Approvals travel with the package
The right check for each format
Each format is assessed with an instrument built for it, not a readability score stretched to fit.
- PEMAT-P for print and PEMAT-A/V for audio and video
- A named, adapted rubric for interactive content
- The CDC Clear Communication Index for key messages
- WCAG 2.2 AA, with manual evidence alongside automated checks
Release, staleness and withdrawal
When a locked fact is replaced, every package built from it would be marked stale and held back until it is reviewed again.
- A clinical withdrawal would end hosted access at once
- Downloaded copies cannot be recalled, so each release would keep a distribution record
- Recipients of a withdrawn copy would be told to replace it
Simulated is labelled simulated
Simulated readers screen drafts for likely confusion. Their findings are labelled simulated and never presented as patient results.
- Simulated participants, not patients
- Personas are design aids; language, age and ethnicity do not determine literacy or preferred format
- Claims about real people need consented research with real people
Accessible in every format
Captions, transcripts, audio description, reduced motion and a complete text fallback are part of the output, not an add-on.
- Never color alone
- A full text and print version of every output
- Automated checks support people; they do not certify conformance
Voices and likeness with consent
Synthetic voices are disclosed. Nobody's voice or likeness is used without consent, and peer stories are labelled as composites.
- Disclosed synthetic voices
- No unconsented likeness
- Composite stories, never presented as real members
No member data needed
HealthLiteracyCopilot reviews templates and content. Personal values would be filled in by the organization's own systems, never sent to HealthLiteracyCopilot.
- Templates, not records
- Offline exports send nothing back
- Messages sent from the organization's own platforms under its consent rules
- No health details beyond what a channel allows
What we will say, and what we won't
We will say
- Decision aids, relevant pictures and spoken content have research support as techniques
- Every output is checked against the facts your reviewers locked, and each one gets its own review
- Spanish and other languages get qualified human review before release
- Every output has a complete text and print fallback
- Simulated-reader findings, always labelled as simulated
We won't claim
- That our videos beat text, or match a clinician
- Gains in Stars, HEDIS measures, screening uptake or readmissions
- Simulated results presented as patient results
- Accessibility conformance from automated checks alone
- Approval for regulated promotion
What it will never do
- Diagnosis, treatment selection or dosing by software, or triage without rules a clinician approves
- Synthetic testimonials presented as real patients
- Anyone's likeness or voice used without consent
- AI signing in place of a qualified human interpreter for critical information
Questions
Are the people and numbers real?
No. Every person, plan, phone number and number in the demos is illustrative. Simulated readers are simulated participants, not patients.
Does any of this give medical advice?
No. The concepts explain what a care team has already decided. None of them diagnose, choose treatments or set doses, and any routing or zone follows rules a clinician approves.
Would we need to share member data?
No. HealthLiteracyCopilot would build and review the templates and content. Any personal values would be filled in by your own systems.
Bring your review rules
Talk to us about becoming a design partner: tell us how your clinical, compliance and language reviewers work today, and we will show how the vision fits.
HealthcareGPS is SOC 2 Type II audited and HIPAA compliant. How we protect your documents