HugoScore hugoscore.org

Full review

Fight Health Insurance CAIHL draft report

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

Back to profile Markdown source

HugoScore CAIHL Draft Report: Fight Health Insurance

Status: Draft for human review Last reviewed: 2026-06-10 Review method: Public-source review of the official homepage, About Our AI page, privacy policy, terms of service, and the Fight Paperwork about page, plus SF Standard coverage confirmed via search; no product walkthrough, code audit, or vendor interview. Service: Fight Health Insurance Vendor: Fight Health Insurance, Inc. Category: Coverage appeal AI

Summary

Fight Health Insurance is a free AI appeal generator created by engineer Holden Karau after winning more than 90 percent of her own 40 insurance appeals. Patients scan a denial letter, optionally using on-device OCR so documents stay local, and the system drafts multiple appeal letters the patient edits and submits, with optional paid fax delivery. The site also offers plain-language denial explanation, policy explanation, AI chat, and a state-by-state help directory, and reports over 10,000 appeals generated. The company, Fight Health Insurance, Inc., was founded in 2023 and sustains the free consumer tool through Fight Paperwork, a paid professional version for practices, hospitals, and patient access teams.

From a CAIHL perspective, this is the most transparency-forward tool in the coverage appeal category. The code is mostly open source, a public About Our AI page names the base models, describes fine-tuning on synthetic examples built from public state appeal decisions, and warns plainly about hallucinated citations. The slogan, make your health insurance company cry too, states the alignment openly. The tensions are real but narrower than the category norm: the terms of service permit training models on submitted information and put the burden of stripping identifiers on the user, and the privacy policy allows business-partner sharing and ad-tech practices that may legally constitute a sale.

Evidence Reviewed

CAIHL Profile

  • Who does this AI serve? Patient-directed. Free for everyone, pay-what-you-want, funded by the separate provider-facing Fight Paperwork product, with no insurer or pharma funding disclosed.
  • Can patients tell AI is involved? Yes. Models, training data, energy use, and failure modes are documented publicly, which is rare in this category.
  • Can patients meaningfully choose? Yes. Voluntary and free, with an on-device OCR option, a choice between self-submission and paid fax, and a dedicated data deletion page.
  • Can patients correct or challenge what the AI produces? Yes, by design. The product generates drafts the patient is instructed to review and edit, and correction and deletion rights are documented.
  • Does it help patients understand or act? Yes. Denial explanation, policy explanation, state resources, and chat support understanding, and appeal generation supports action.

Agency Interpretation

Fight Health Insurance's clearest agency value is pairing free strategic action with genuine critical reflection. It does not just draft the appeal. It explains the denial, explains the policy, points to state regulators, and shows its own machinery, including model names, training sources, and limitations. Open-source code means outside evaluators can inspect what the AI actually does, which is the strongest contestability posture of any tool in this category. The absence of a marketed success-rate percentage is a point of restraint worth naming, since competitors lead with unaudited win rates.

The unresolved tension sits between the product's privacy-forward design and its legal paperwork. The terms state that submitted information may be used to train models and make it the user's responsibility to remove names, addresses, and insurance numbers first, while the About Our AI page says no real patient appeal letters are used in training. Both can be true, but the burden allocation is unfriendly to the very patients the tool serves. The privacy policy's business-partner sharing and interest-based advertising language also sits oddly beside the private-by-default pitch. A small team and a sustainability model resting on the paid professional product are practical fragilities for patients who come to depend on the free tool.

Key Unknowns

  • Whether user-submitted appeal content is actually used in model training today, given the tension between the terms of service and the About Our AI page.
  • Whether the business-partner and ad-tech sharing described in the privacy policy applies to health-related inputs or only to site analytics data.
  • Actual appeal success rates, which the service does not publish.
  • How much of the current production system is covered by the open-source repositories.
  • The relationship between the free consumer tool's roadmap and the paid Fight Paperwork product, and whether consumer features could move behind the professional paywall.
  • Language access, disability access, and low-literacy support.
  • Long-term sustainability of the free tier given the small team.

Publication Recommendation

Ready for human review as a draft profile. Confidence should stay at medium until the training-data tension between the terms and the AI page is clarified with the vendor, the scope of ad-tech sharing is confirmed, and the open-source repositories are spot-checked against the production service. The transparency posture is strong enough that several unknowns could be resolved by direct inspection rather than vendor disclosure.