HugoScore hugoscore.org

Full review

Google Health Coach (Gemini) 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: Consumer Google Health Coach in the Google Health app, built with Gemini and requiring Google Health Premium. Separate from the Gemini apps.

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
Your AI-powered coach continually adjusts workout plans based on what's going on in your life Source

Store page also says 'built with Gemini'. The help center says the coach is 'built with Gemini'. I did not see an in-app screen or a documented consent step.

2. Operator and model maker namedYesdocumented
Not independently replicated
Your personal health coach, built with Gemini. Source

Google runs the Google Health service and makes Gemini, so one company is both operator and model maker. The Google blog says it 'grounded the coach in Gemini models'.

3. Patient chooses to use itYesdocumented
Not independently replicated
Once purchased, you can activate the Premium membership in the Google Health app. Source

The user buys Premium and enables the coach. The support page says the coach can be turned off in feature privacy controls.

4. Patient sets the taskYesdocumented
Not independently replicated
you can update details or shift your goals at any time Source

The user chats and sets goals. The coach also sends proactive Today-tab insights, which can be switched off ('No Coach insights'). A reviewer may judge proactive coaching as steering.

5. Patient can decline or stopYesdocumented
Not independently replicated
When you turn off the coach, your data stops being collected or processed for these features moving forward. Source

Self-started tool per the Q3 source, and no care depends on it. The rest of the Google Health app and device features continue.

6. Patient sees the AI outputYesdocumented
Not independently replicated
Instead, you can delete specific conversations in your Chat History. Source

Chat and voice conversations are kept in Ask Coach > History, where the user can see them.

7. Patient can correct or challengeNonot-documented
Not independently replicated
Quotation not recorded; evidence repair pending. Source

No documented way to correct or contest the coach's output or what it saved. The pages only describe feedback on a conversation, which triggers human review, and updating logs and goals.

8. Patient can get a copyYesdocumented
Not independently replicated
you can export your data in a commonly used file format Source

Export is through Google Health Data and Privacy settings. I did not confirm that coach chat transcripts are included in the export.

9. Patient can deleteYesdocumented
Not independently replicated
You can delete selected health and wellness data history through your Google Health Data and Privacy settings Source

Coach conversations can be deleted one by one or all at once (support 17055092). Leaving the service permanently deletes the data.

10. Training or sale does not override patient choiceYesdocumented
Not independently replicated
Your personal interactions with AI in the Google Health app are not used for training the underlying generative models Source

The same page says the opt-in research option does use coach interactions to train the AI model, and that choice is optional. Training is off by default.

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

Vendor-hosted only. Export exists but no run-it-yourself or move-the-function route is documented.

12. No advertising or data businessNodocumented
Not independently replicated
research and develop new health and wellness products and services, including Google’s enterprise products Source

Google says it will not use Fitbit users' health data for Google Ads. But the opt-in research program uses de-identified data for Google products, including enterprise products, and the rule counts that even with consent. The page also says Fitbit Enterprise programs exist with employers and insurers.

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

Google Health Coach (Gemini): CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

Google Health Coach describes adjustable goals, optional research participation, and controls for disabling the coach and deleting conversations. Enabling it grants broad account-data access. Disconnecting a feature and deleting information already used in conversations are separate actions.

Scope and agency posture

Consumer Google Health Coach in the Google Health app, powered by Gemini and requiring Premium. This is separate from the Gemini apps.

Posture: Mixed, patient-chosen but ecosystem-gated.

Axis: 66/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

  • Google's current product page says users can set goals and adjust plans through conversation. It describes fitness, sleep and health guidance, including integrated medical records, and says the service is not intended for medical use. https://healthapp.google/google-health-coach/
  • Google's help page says enabling Coach gives it access to all health and activity data in the account. Revoking a feature or deleting a data category does not automatically remove information from conversation history. Those conversations must be deleted separately. https://support.google.com/googlehealth/answer/17055092
  • The help page documents turning Coach off, removing Today insights, deleting individual or all conversations, and optional research consent. Human review is described for submitted feedback or research participation. Location permission can be withdrawn in device settings. These are documented controls, not independently tested behavior. https://support.google.com/googlehealth/answer/17055092

Patient authority

Inference from the documented workflow: Users choose goals and can adjust plans or disable the coach. Broad account access limits selective permission once enabled. No current evidence establishes a compulsory treatment or adherence path.

Critical capacity

Editorial assessment: Conversation and record context can help users prepare questions and interpret trends. Public pages do not establish consistent source citations for individual suggestions or correction of underlying medical records.

Informed control

Editorial assessment: The help page explains data categories and deletion limits unusually specifically. Feature revocation and conversation deletion are separate actions. Optional research and feedback review should not be described as default training or unrestricted human review.

Assessment and limits

The existing Mixed placement is supportable. Patient chosen guidance and goal adjustment coexist with broad data access and a proprietary ecosystem. Subscription cost alone is not an agency penalty.

Confidence: Moderate for documented controls and positioning. No account or coach interaction was tested..

Remaining uncertainty: Actual recommendation provenance, correction behavior, and the usability of separate deletion controls remain unverified.

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: 2026-06-10. Historical assessment. Earlier claims are not automatically reverified by this publication.