Original research · Dataset
Vendor Transparency Index
Eight disclosure checks applied identically to every vendor in the category. It measures what a vendor is willing to publish — a variable nobody else in this market tracks.
Short answer
Which clinical documentation vendors disclose the most publicly?
Deepgram leads at 94% of eight disclosure checks met; Ambience Healthcare sits lowest at 13%. The pattern is structural rather than ethical: vendors selling infrastructure to developers publish documentation because documentation is how they sell, while vendors selling enterprise applications disclose through sales processes and NDAs.
Cite as: Vendor Transparency Index, Compare Healthcare API, last reviewed 2026-09-01.
Key findings
- Disclosure and product category are almost perfectly correlated. Every vendor scoring above 70 sells infrastructure; every vendor scoring below 40 sells a clinician-facing application. The dividing line is not company size or funding — it is who the vendor thinks the buyer is.
- Benchmark methodology is the least-published item in the entire category. Not one vendor audited publishes a corpus, audio condition set and reference-transcript protocol complete enough for an integrator to reproduce a stated accuracy figure. Accuracy claims in this market are, without exception, unverifiable from public material.
- BAA availability is the most-published item, because it is table stakes for even entering a healthcare conversation. It therefore carries almost no discriminating information — a BAA tells you a vendor is in the market, not that it is a good fit.
- Rate limits and concurrency are published by the ASR vendors and withheld by the application vendors. For an integrator planning capacity this is a hard blocker: you cannot design for peak load against an undocumented limit.
- The gap between 'we are HIPAA compliant' on a marketing page and a published retention period plus a contractual training-on-PHI position is where most procurement delays in this category originate.
The index
Yes means a publicly reachable page states it. Partial means it is stated incompletely or conditionally. No means it is not published anywhere public.
| Vendor | Score | Public API reference | Self-serve credentials | Published price or model | Published rate limits | BAA availability stated publicly | Retention and deletion terms published | Training-on-PHI position published | Benchmark methodology published |
|---|---|---|---|---|---|---|---|---|---|
| Deepgram | 94% | yes | yes | yes | yes | yes | yes | yes | partial |
| AssemblyAI | 94% | yes | yes | yes | yes | yes | yes | yes | partial |
| AWS HealthScribe / Transcribe Medical | 88% | yes | yes | yes | yes | yes | yes | yes | no |
| Azure AI Speech | 88% | yes | yes | yes | yes | yes | yes | yes | no |
| Twofold | 69% | yes | yes | yes | partial | yes | partial | partial | no |
| Speechmatics | 69% | yes | yes | yes | partial | partial | partial | partial | partial |
| Nuance Dragon Copilot / DAX | 44% | partial | partial | no | partial | yes | partial | partial | no |
| Abridge | 25% | no | no | no | no | yes | no | partial | partial |
| Suki | 13% | no | no | no | no | yes | no | no | no |
| Ambience Healthcare | 13% | no | no | no | no | yes | no | no | no |
What each check asks
The questions are deliberately answerable from public pages alone.
| Public API reference | Can an engineer read the full API reference without talking to sales? |
|---|---|
| Self-serve credentials | Can a developer obtain a working key the same day, unaided? |
| Published price or model | Is a price, or at minimum an explicit pricing model, published? |
| Published rate limits | Are concurrency and rate limits documented before contracting? |
| BAA availability stated publicly | Does the vendor state publicly that a BAA is available? |
| Retention and deletion terms published | Are data retention periods and deletion controls documented publicly? |
| Training-on-PHI position published | Is the position on training models with customer data stated in public terms? |
| Benchmark methodology published | If accuracy is claimed, is the corpus and protocol published so it can be checked? |
Why this dataset exists
Nobody publishes it, and it predicts procurement pain better than product reviews do.
Every buyer's guide in this category evaluates features. None of them measure the variable that determines how long your evaluation takes and how much of your risk stays contractual: whether the vendor is willing to write things down in public. A vendor that publishes retention windows, subprocessors and a PHI-training position has already answered a third of your customer's security questionnaire before you ask.
The correlation with business model is close to perfect and worth stating plainly, because it lets you predict a vendor's behaviour before you engage. Infrastructure vendors publish; application vendors negotiate. Neither is dishonest, but they impose very different evaluation costs on a software team.
Evidence & sources
Regulatory and assurance references that define what these disclosure checks look for. Vendor-specific evidence is the vendor's own public pages, recorded per row on the audit date.
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: The contractual clauses a documentation vendor's BAA must contain.
U.S. Department of Health & Human Services · Regulation · Tier A — primary documentation
Supports: Administrative, physical and technical safeguards a vendor handling recorded encounter audio must implement.
AICPA · Standard · Tier A — primary documentation
Supports: What a SOC 2 Type II report does and does not attest to during a vendor security review.
- [4]HITRUST CSF
HITRUST Alliance · Standard · Tier A — primary documentation
Supports: The certification many health systems require of documentation vendors handling PHI at scale.
NIST speech recognition evaluation literature · Methodology · Tier B — published methodology
Supports: Why a headline WER figure without a stated corpus, audio condition and reference-transcript protocol is not comparable across vendors.
NIST · Standard · Tier A — primary documentation
Supports: A defensible structure for governing an AI documentation feature you ship to clinicians.
Source tiers are defined on the methodology page. Outbound links are unaffiliated and carry no commercial relationship.
Frequently asked questions
- What is the Vendor Transparency Index?
- An original dataset scoring every vendor in this category on eight public-disclosure checks: developer documentation, published pricing, retention windows, PHI-training position, benchmark methodology, subprocessor list, certification evidence and white-label terms. It measures what a vendor publishes, not how good its product is.
- Why does disclosure matter if the product is good?
- Because you cannot pass a health-system security review with claims you cannot document, and because undisclosed retention or training positions become your liability once the vendor processes your customers' PHI. Compliance checklist.
- Is a low disclosure score the same as a bad vendor?
- No. It means more of your evaluation has to be done under NDA and more of your risk is contractual rather than verifiable. Enterprise vendors score low here structurally, because they sell through sales processes rather than documentation.
- How often is the index updated?
- It is re-audited when vendor documentation changes materially. This edition was compiled on 2026-09-01 from publicly reachable pages only — nothing behind a login, NDA or sales call.