HugoScore hugoscore.org

Full review

Advoca CAIHL draft report

Evidence-linked HugoScore draft report for a health AI tool that affects patients.

Agency checklist · October 30, 2026

Leans expands 9 of 12 yes

Document-based draft. Independent confirmation is incomplete. Hugo has not personally verified this service.

Assessed function: Patient/carer-selected Advoca app for recording, preparation, health questions and research participation. Vendor-hosted, with a distinct research-commercialization channel.

Astra review conflicted: OpenAI model in role record.

A no can mean missing documentation or institution-dependent control; it does not by itself establish harmful behavior. These categories are not clinical safety or product-quality ratings.

  • Public documents do not establish actual product behavior.
  • The independent review screened rules and selected evidence; it did not replicate all twelve answers for this tool.
See all twelve answers and their evidence
QuestionAnswerEvidence / reviewSource and reasoning
1. Told it is AIYesdocumented
Not independently replicated
AI-generated summaries of your appointments in simple language, created from the verified transcripts. Source

Privacy Policy and the Consumer Health Data policy both describe LLM-generated summaries and an LLM chatbot.

2. Operator and model maker namedYesdocumented
Not independently replicated
We use Amazon Web Services, Google Cloud Platform, and OpenAI to process audio or transcripts and generate summaries. Source

Operator Advoca Health Ltd (Companies House 16650138). Same section says these providers also perform the chatbot LLM inference under zero data retention.

3. Patient chooses to use itYesdocumented
Not independently replicated
Your explicit, informed consent obtained through the App before you can use these features. Source

Section 3(a), US legal basis. Patient or carer installs the app and consents in a consent flow.

4. Patient sets the taskYesdocumented
Not independently replicated
Plan your notes beforehand with appointment planning, record the conversation on the day, and get a clear, sectioned summary afterwards Source

Patient chooses recording, journal, chat topics and plans. Advoca also runs in-house analysis to identify research opportunities (3(g)) and markets a research reward, which is a secondary operator goal.

5. Patient can decline or stopYesdocumented
Not independently replicated
Your explicit, informed consent obtained through the App before you can use these features. Source

Self-started: Q3 is yes and no care depends on using Advoca. Source is the Q3 source. Account can be deleted at any time and research community is optional.

6. Patient sees the AI outputYesdocumented
Not independently replicated
AI-generated summaries of your appointments in simple language, created from the verified transcripts. Source

Transcripts, summaries and chat answers are shown to the user. Audio is not stored on servers.

7. Patient can correct or challengeNonot-documented
Corrected after independent review; accepted by editor
Your Data Rights (UK and EU Users) Source

The output-rectification right is under UK/EU rights, with a separate California provision; a universal route is not established.

8. Patient can get a copyYesdocumented
Not independently replicated
You can request a copy of all data associated with your account Source

Section 7 says machine-readable format via the application. Home page: 'Export or delete it anytime'.

9. Patient can deleteYesdocumented
Not independently replicated
You can delete all data associated with your account (appointment data, profile health data, personal data, and account data). Source

Specific appointment data can also be deleted in the app, which revokes shared access. Inactive accounts are removed after 2 years.

10. Training or sale does not override patient choiceYesdocumented
Not independently replicated
Joining Advoca’s research community is optional for all users. You can use Advoca without joining. Source

AI training on anonymised data and partner sharing occur only with separate opt-in consent, withdrawable at any time. Policy says Advoca does not sell personal data. Separate caveat: product-improvement analytics, including health journal and chat content, run through PostHog for all users with no opt-out.

11. Patient can run it or take it elsewhereNonot-documented
Not independently replicated
Quotation not recorded; evidence repair pending. Source

Vendor-hosted. Copy of data on request only. No self-hosting or documented move-elsewhere route.

12. No advertising or data businessNodocumented
Not independently replicated
your data will be fully anonymised and may be used for research, development, and commercial purposes. Source

Opt-in community consent allows anonymised data to be shared with trusted partners for commercial purposes. The Life Sciences page offers 'Real-world evidence from our patient community' and the home page promises a share of rewards on study matches. Advertising attribution data goes to Airbridge and Meta (not health data).

Initial coding: Claude Sonnet helpers, compiled by Claude Opus 5.5 (Anthropic). Selective review and corrections: GPT-6 Astra (OpenAI). Editor: Hugo Campos. Acceptance of corrections is not product verification.

Checklist rules · Independent review and limitations · Later evidence follow-up

Earlier report

Advoca: CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

Advoca lets patients and care partners record visits, ask questions, and share information. Its Washington consumer health policy ties free access to research consent, with paid access after withdrawal. The policy and homepage give conflicting accounts of audio processing.

Scope and agency posture

Patient/carer-selected Advoca app for recording, preparation, health questions, and research participation. Vendor-hosted consumer assistance, with a distinct research-commercialization channel.

Posture: Potentially agency-expanding, with research-access and processing qualifications.

Axis: 84/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

  • The current homepage describes source-linked answers, a checkable transcript/recording, and patient sharing. It also claims audio never reaches its servers. These are vendor statements.
  • The consumer health policy, sections 1 and 3, says free-tier use requires research consent, withdrawing it requires a paid subscription, and audio reaches cloud processors. Its stated scope is Washington consumers. The audio description conflicts with the homepage.
  • The terms, sections 7–9, recognize user-content ownership and separate research consent. Personal/private sharing is permitted, with restrictions on the service's own content.

Patient authority

Inference from the documented workflow: Patients choose questions, records, and recipients. Consent withdrawal is documented but has a financial consequence for free-tier users. That is a substantive condition on continued access, not simply optional research alongside identical free service.

Critical capacity

Editorial assessment: Source-linked answers and retained originals could help users check summaries and prepare challenges. Actual source matching and correction are untested.

Informed control

Editorial assessment: The detailed policy enables scrutiny, but its account of processing differs from a prominent privacy promise. Research consent and model-training use should be visible before reliance.

Assessment and limits

The baseline's broad free/optional-research characterization needs correction. Patient direction remains supported, while the consent-price linkage and conflicting audio claims justify a qualified posture rather than automatic high-confidence expansion.

Confidence: Medium for published conditions, low for implementation.

Remaining uncertainty: Whether the tier conditions apply identically outside Washington, actual audio routing, transcript correction, and post-anonymization withdrawal effects.

Published documentation is evidence of stated conditions, not proof of actual implementation. Unknowns did not receive automatic negative points. Funding, sponsorship, and public code do not determine agency by themselves.

Review provenance

  • Reviewer/model: OpenAI Codex / GPT-6.
  • Method: Focused public-source reassessment using CAIHL: patient authority, critical capacity, and informed control. Existing evidence plus one focused primary-source pass and at most one targeted follow-up. No live product testing. Numerical scores remain provisional editorial placements, not a new calculation.
  • Human review: Hugo Campos authorized publication of these AI-assisted draft reassessments on September 8, 2026. This does not claim comprehensive human verification of every finding.
  • Earlier review: 2026-06-08. Historical assessment. Earlier claims are not automatically reverified by this publication.