Patient care navigation and health copilot AI
Anistratenco Health Assistant
A patient-initiated health assistant with cited research and local-context controls described by the vendor. Potentially agency-expanding, but provider-specific processing terms and untested correction behavior limit confidence. Unscored.
Hugo Campos authorized publication of this AI-assisted draft. No comprehensive human verification or hands-on product testing.
Summary judgment · Unscored in current comparisons
Current consumer assistant; no hands-on behavior or future feature credit.
Potentially agency-expanding
Chosen questions, inspectable sources and record control could support patient authorship. This is a conditional qualitative judgment, not demonstrated alignment or a numerical ranking.
Patient agency
How this tool changes agency
May help people interpret their own information, inspect sources and prepare questions they choose.
External processing, metered access and unverified correction of inferred context can limit informed control.
Patient agency assessment
Who sets and changes the goal?
Consumer users choose questions and context; persistence of corrections or refusal to follow vendor framing was not tested.
Can patients tell AI is involved?
The public offering explicitly describes health AI reasoning.
Can patients meaningfully choose?
Voluntary consumer access and deletion/export routes are described; complete exit and provider-copy deletion were not verified.
Can patients correct or challenge the output?
Data rights and user input ownership do not establish reliable correction of generated interpretation or memory.
Does it help patients understand or act?
Cited research and chosen health questions may support reflection and visit preparation; advocacy and citation fidelity were not tested.
Can patients evaluate the conditions of use?
Storage and processing are distinguished, but active provider/model identities and their effective terms were not established.
Text findings
What happens to patient data?
Conditional provider protections
The policy describes encrypted local context and cloud ciphertext. Chosen analysis can require plaintext processing. No-storage requests depend on provider support; provider retention/training terms can apply. No own-model training of raw health conversations/files is stated. [S2, S4]
Who is left out or burdened?
Paid and metered
Monthly plans show Plus $19 and Pro $49 with compute allowances. Digital access, affordability, accessibility and subgroup performance remain unevaluated. [S3, S5]
Are the clinical boundaries clear?
Documented; behavior untested
Terms limit the current service to consumer informational assistance, excluding diagnosis, prescribing, emergency response and continuous monitoring. Citations can be wrong. [S3]
Who defined what good looks like?
Not established
The inspected sources do not establish patient partnership in defining evaluation outcomes or independent validation of the current assistant.
Review provenance
Criteria
CAIHL-derived HugoScore framework and current qualitative review method
Reviewer
AI-assisted public-source review prepared in OpenAI Codex
AI / model
OpenAI Codex / GPT-6 (exact variant not recorded)
Human review
Hugo Campos authorized review and publication. No comprehensive human verification is claimed.
Review date
2026-09-28
Limitations
Public-source documentary review only. No account creation, health-data upload, interaction tests, security audit, clinical evaluation or vendor interview. Host-specific and provider-specific behavior is unverified.
Review method
CAIHL public-source documentary review; current functionality separated from roadmap and adjacent offerings. No authenticated testing.
AI-assisted draft · Medium for published terms, low for practical agency and behavior (AI-assisted draft)