HugoScore hugoscore.org

Full review

Claimable CAIHL draft report

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

Agency checklist · October 30, 2026

Mixed 7 of 12 yes

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

Assessed function: Claimable patient appeal workflow (patient-facing AI-drafted insurance appeals, including pharma/provider-sponsored referrals); enterprise analytics for pharma and health systems not assessed

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
utilizing our artificial intelligence driven software solution Source

AI is disclosed in the Terms of Service (Service section) and throughout the site. A separate in-product consent screen was not observed (login required).

2. Operator and model maker namedNonot-documented
Not independently replicated
Quotation not recorded; evidence repair pending. Source

Claimable, Inc. is named as operator. No public source names the maker of the language model. The AI blog says only 'custom-built technology', 'large language models' and NVIDIA Inception membership.

3. Patient chooses to use itYesdocumented
Not independently replicated
account holders themselves and individuals (e.g. child or elderly parent) whom account holders choose to assist in filing of appeals Source

Patient creates an account and completes intake. Referrals from pharma, pharmacies or providers exist (for-pharma FAQ: 'invited through direct referral or self-enrollment'), but the patient still starts and approves the appeal.

4. Patient sets the taskYesdocumented
Not independently replicated
Patients remain the decision-makers, and AI simply structures their story with supporting evidence. Source

Statement is from the partner-facing page, not a binding document. Sponsored programs are disclosed 'during consent' per the same page. Whether sponsored flows steer patients toward a sponsor's drug was not tested.

5. Patient can decline or stopYesdocumented
Not independently replicated
account holders themselves and individuals (e.g. child or elderly parent) whom account holders choose to assist in filing of appeals Source

Self-started rule applied. Also: 'nothing is sent without their approval' (for-pharma FAQ).

6. Patient sees the AI outputYesdocumented
Not independently replicated
you review the final document before anything is submitted Source

Patient sees the AI-drafted appeal before submission.

7. Patient can correct or challengeYesdocumented
Not independently replicated
Patients verify accuracy, make edits, and authorize submission Source

Edit and flag route documented in partner FAQ and for-providers page ('can make any changes before it's sent'); the AI blog says patients can 'mark it for further examination'.

8. Patient can get a copyNonot-documented
Not independently replicated
Quotation not recorded; evidence repair pending. Source

Privacy policy lets users 'review and update certain Claimable account information'. No export or download of the appeal or data is documented. The Data Request Form is mentioned only for marketing opt-out.

9. Patient can deleteYesdocumented
Not independently replicated
You can request that we delete your account by contacting us with your account deletion request or deleting your account through the Services. Source

Terms say Claim Data 'may be deleted and not be retrievable' on account deletion, but the same Terms grant Claimable a perpetual, irrevocable license to Claim Data. What survives deletion is unclear.

10. Training or sale does not override patient choiceNodocumented
Not independently replicated
when aggregated and anonymized, any other purpose in any form, format, or media that we choose Source

Binding Terms (section 6) and Privacy ('research and development', sharing de-identified data) allow reuse. The AI blog says the AI 'is trained to recognize patterns in successful appeals'. No opt-out is documented. The safe-sharing blog says data is not sold or shared, which conflicts with the binding documents. Answered by the binding documents.

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

Vendor-hosted only. Terms grant a revocable, non-transferable right to use the Service. No self-hosting or portability.

12. No advertising or data businessNodocumented
Not independently replicated
We may use this aggregated, de-identified or otherwise anonymized data and share it with third parties for our lawful business purposes Source

Privacy also allows sharing with 'licensors or partners' for 'their analysis or marketing purposes subject to your consent', and CCPA 'sale' of device data. The for-pharma page markets 'aggregate analytics' of insurer denial patterns to manufacturers. Terms allow fees to be waived when sponsored by pharma or providers.

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

Claimable: CAIHL reassessment

September 8, 2026. Published AI-assisted draft.

Claimable’s patient appeal workflow, including sponsored referrals, documents patient editing and approval before delivery. Enterprise analytics have a separate scope. Sponsorship and contradictory sharing language require scrutiny, with implementation unverified.

Scope and agency posture

Claimable’s patient appeal workflow, including sponsored referrals, documents patient editing and approval before delivery. Enterprise analytics have a separate scope. Sponsorship and contradictory sharing language require scrutiny, with implementation unverified.

Posture: Potentially agency-expanding, with sponsorship-alignment caveat.

Axis: 72/100, retained provisionally. No numerical recalibration was performed.

Sources and documented findings

  • The provider page says patients can change the letter, optionally share it with their clinician, and approve it before delivery. Provider workflow, steps 3–4, freshly retrieved September 8. Vendor description, not observed behavior. General export and reuse rights were not established.
  • The pharma page describes patient approval, sponsorship disclosure at consent, and access revocation. Pharma FAQ, recovered September 8 evidence. Actual consent prominence and sponsor influence on available strategies remain untested.
  • Partner appeal credits are spent on patient submission and returned after 30 days without action. Partner service agreement. This agreement does not describe every enterprise deployment.
  • Privacy wording both denies promotional sharing and permits consent-based partner analysis or marketing. Privacy policy, sharing section. The passages do not establish an actual unauthorized disclosure.

Patient authority

Inference from the documented workflow: the documented approval gate supports patient direction within the appeal task. Provider referral and sponsorship do not erase that gate. Authority over a different treatment goal or enterprise strategy is not established.

Critical capacity

Editorial assessment: drafting and delivery can make contesting a denial practically possible. Evidence-backed argument is a documented product claim, not verified citation quality or proof of broader critical understanding.

Informed control

Editorial assessment: the public controls are meaningful, but contradictory sharing language and unobserved consent limit confidence in how patients evaluate the service's conditions. Commercial interests may coincide with a chosen appeal without encompassing all patient interests.

Assessment and limits

Claimable’s patient appeal workflow, including sponsored referrals, documents patient editing and approval before delivery. Enterprise analytics have a separate scope. Sponsorship and contradictory sharing language require scrutiny, with implementation unverified.

Confidence: Medium for published documentation, low for implementation.

Remaining uncertainty: Consent presentation, independent output-use rights, control outside the supported appeal goal, and actual partner data flows.

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.