Nabla Review 2026: The Only AI Scribe That Won a Randomized Trial

Every AI medical scribe vendor will show you a case study. Almost none of them will show you a randomized trial where their product was tested head-to-head against a named competitor by a health system that had no stake in the outcome.

Nabla will. That single fact is the most interesting thing about this company in 2026, and it deserves more scrutiny than the marketing headline it has become.

This is an honest review: what the trial actually found, what it did not, how Nabla has repositioned itself over the past year, and why the version of Nabla you can buy today is not the one most of the internet still describes.

The short version

Nabla is a Paris- and New York-based ambient AI company whose assistant listens to the clinical encounter and drafts the note, with dictation and coding support layered on top. As of mid-2026 the company reports 85,000+ clinicians across 130+ healthcare organizations, processing roughly 20 million encounters a year.

Three things separate it from the pack:

1. It won a randomized trial on the primary endpoint — against Microsoft DAX — published in NEJM AI.
2. It has moved decisively upmarket. Nabla was once the easy self-serve pick for a solo clinician. In 2026 it sells through demos and contracts.
3. It is betting on "agentic," not just ambient — EHR commands, order drafting, and coding, funded by a $70M Series C in June 2025 that brought total funding to $120M.

If you are a health system or a mid-sized group on Epic or athenahealth and you want evidence rather than testimonials, Nabla belongs on your shortlist. If you are a solo clinician who wants to sign up tonight and start charting, read the pricing section carefully — the answer changed.

The trial: what it actually showed

In December 2025, NEJM AI published "Ambient AI Scribes in Clinical Practice: A Randomized Trial." It is worth walking through properly, because it is the strongest piece of independent evidence any scribe vendor currently points to — and because the caveats matter as much as the result.

Design. A three-arm pragmatic randomized trial at UCLA Health, running November 4, 2024 to January 3, 2025, with 238 outpatient physicians across 14 specialties. Arms: Nabla, Microsoft DAX, and usual care. Randomization was balanced on baseline documentation time, burnout score, and clinic days per week. Roughly 48,000 outpatient visits were analyzed.

Primary outcome — time in note:

• Nabla: 9.5% reduction vs. control (95% CI −17.2% to −1.8%; p = .02)
• DAX: 1.7% reduction vs. control (95% CI −9.4% to +5.9%; p = .66) — not statistically significant
• Nabla vs. DAX difference: 7.8% (95% CI −15.5% to −0.1%; p = .05)

Secondary outcomes (both scribe arms pooled against control): Mini-Z 2.0 burnout score improved by 2.76 points (p < .001), task load fell 35.8 points on a 0–400 scale (p = .01), and work exhaustion fell 0.27 on a 0–4 scale (p = .01). An exploratory analysis found a dose-response effect — physicians who used the scribe on a higher share of encounters benefited more.

Now the honest caveats

Nabla's summary of this trial is accurate. It is also, inevitably, the most flattering true version. Four things to hold in mind:

• Single institution, two months. One academic health system, one Epic build, one training approach, over an eight-week window. That is not the same as evidence it will outperform DAX in your community clinic.
• 9.5% is real but modest. If you spend 90 minutes a day in notes, that is roughly nine minutes back. Meaningful across a year and a large medical group; not the "hours a day" the category's marketing implies. Nabla's own site separately claims 55% of clinicians save an hour or more daily — that is a self-reported vendor survey, a very different class of number from the trial's objective Epic metrics, and the two should not be quoted in the same breath.
• The burnout wins were pooled. The improvements in burnout, task load, and exhaustion were measured for both scribes together versus control. They are not a Nabla-over-DAX result, and the authors flagged them as secondary endpoints needing confirmation in larger multicenter trials.
• The measurement has a blind spot. The study used Epic Signal metrics, which do not capture time spent editing a note outside the native platform. A scribe whose drafts get heavily reworked elsewhere could look better than it is.

None of that makes the finding weak. A statistically significant win on an objective, pre-specified primary endpoint in a randomized trial is rare in this category, and Nabla earned it. It simply is not a guarantee, and any vendor who tells you a single-site RCT is a guarantee is overselling. Our broader guide to what the research says about AI scribe accuracy puts this in context alongside the hallucination literature.

What Nabla actually is now

The product has three named pieces:

• Ambient AI Assistant — listens to the encounter and drafts a structured note in seconds, across in-person and telehealth visits, with multilingual support that is genuinely broad (Nabla has long marketed 35+ languages, including Haitian Creole, several Arabic dialects, and a wide range of South and East Asian languages).
• Dictation — clinical-grade speech recognition embedded in the EHR, which is what pulled Nabla into direct competition with legacy Nuance/Dragon deployments rather than just other scribes.
• Coding — ICD-10, HCC, and MCC suggestions grounded in the documentation and AMA guidelines, offered as a toggle. Note the gap: CPT coding is not currently a Nabla feature, and a proactive coding agent still sat on the roadmap as of mid-2026. If revenue capture is your primary goal rather than documentation relief, look at dedicated AI medical coding tools instead of assuming the scribe covers it.

The strategic direction, stated plainly at the Series C, is agentic: a context-aware assistant that initiates EHR commands and drafts orders rather than only producing text, plus role-specific agents starting with hospital nurses. That is the same land-grab Suki has been making for years and that Abridge is funding at a larger scale — see our Suki review and Abridge review for how the three theses differ. Treat roadmap items as roadmap items when you negotiate.

Traction and integrations

The buying signal here is real and increasingly enterprise-shaped:

• M Health Fairview selected Nabla's combined Ambient AI Assistant and Dictation for systemwide deployment in February 2026 — 10+ hospitals and 60+ clinics on Epic, following a multi-phase pilot with early rollout across emergency medicine, cardiology, pediatrics, surgery, and inpatient care.
• athenahealth Marketplace: 30+ provider groups adopted Nabla in the first year after its September 2024 listing. One group, knownwell, reported a 24% reduction in per-visit documentation time and a 64% increase in clinician adoption over six months.
• EHR coverage spans Epic, athenahealth, Oracle Health, NextGen, Greenway Health, and others, with API access for teams building custom workflows.

On security, Nabla holds SOC 2 Type II and ISO 27001, encrypts at rest with AES-256 and in transit over TLS, and lets organizations choose their data region at setup. Its trust portal is public. That is a solid posture — but read the next section before assuming it applies to whatever plan you land on.

Pricing: the part that changed

Here is where most of the Nabla content online is now out of date.

Nabla does not publish pricing. The figure repeated across third-party trackers is roughly $119 per clinician per month, with some listing a higher tier around $239. Those numbers are consistent enough to be a useful planning anchor, and they put Nabla in the middle of the market — above Freed at the low end, well below enterprise DAX-class contracts. See our AI scribe pricing breakdown for the full landscape.

But two caveats do real work:

Nabla moved off public per-month tiers in 2026 and now sells via demo and contract. The self-serve, card-on-file experience that older reviews describe is not how you buy Nabla today. Budget for a sales cycle, a pilot, and negotiated terms.

The free tier is not a HIPAA plan. Nabla has long offered a free tier — commonly reported as up to 30 consultations per month — and it is still widely listed. Multiple sources indicate the free plan does not come with a signed BAA. Without a BAA, running real patient encounters through it is not a HIPAA-compliant arrangement, full stop. Use the free tier to evaluate note quality on your own voice or a simulated encounter; do not use it on live PHI. If that distinction is fuzzy, start with what HIPAA compliance actually means for AI tools, and remember that an ambient scribe is a recording device with its own consent obligations.

Where Nabla is weaker

A few consistent criticisms, with the appropriate grain of salt attached:

• Template and customization depth. Reviewers repeatedly flag limited customization relative to scribes built around user-editable templates, with deeper controls sitting in higher tiers. If your clinicians have strong, idiosyncratic note formats, test this specifically.
• Complex multi-problem visits. Several reviews report quality drop-off when a single encounter covers many active problems. This is a real weakness across the category, not unique to Nabla, but it is exactly the visit type where a scribe either earns its cost or does not — so build it into your pilot instead of testing on simple acute visits.
• No CPT coding today. Already noted, worth repeating if coding is part of your ROI model.
• The SMB motion is gone. Nabla is no longer optimizing for the solo clinician, and it shows in the buying experience.

A caveat on the caveats: a large share of the "Nabla review" pages ranking in search are published by competing scribe vendors. Their criticisms are not automatically wrong, but they are not disinterested either. Weigh them accordingly — and weigh ours by the same standard. Where we cite a number above, it comes from the peer-reviewed trial, a company press release, or the vendor's own site, and we have labeled which is which.

How Nabla compares

• vs. Abridge — Abridge is the deeper enterprise play with heavier health-system infrastructure and revenue-cycle ambition. Nabla is lighter to deploy and has the better independent efficacy evidence.
• vs. Suki — Suki's differentiator is the hands-free voice command layer. If your clinicians will actually talk to the EHR, Suki leans harder into it; if you mainly want a strong note, Nabla is the more focused buy.
• vs. Heidi and Freed — both are far easier and cheaper to adopt solo, with transparent pricing and same-day setup. Our Heidi vs. Freed comparison covers that end of the market, and our roundup of free and low-cost scribes is the better starting point for a one-clinician practice.
• vs. your EHR's built-in scribe — if you are on Epic, the native option is now a genuine alternative and the calculus is not obvious. We worked through it in Epic's AI scribe vs. standalone tools.

If you are evaluating Nabla, do this

1. Ask for the trial, not the summary. Read the NEJM AI paper yourself and ask Nabla how your setting differs from an academic Epic build at UCLA.
2. Pilot on hard visits. Multi-problem chronic care, not straightforward acute complaints.
3. Confirm the BAA in writing before a single real encounter is recorded, and confirm which plan it attaches to.
4. Get audio and transcript retention in the contract — how long, where, and whether your data trains anything.
5. Separate roadmap from product. Agentic orders and proactive coding are announced directions. Price the product you can use in month one.
6. Measure the same thing UCLA did. Pull baseline time-in-note from your EHR before the pilot, or you will be arguing about vibes at renewal.

The bottom line

Nabla is the ambient scribe with the best independent evidence behind it — a real randomized trial, an objective primary endpoint, and a statistically significant win over the category's biggest incumbent. In a market where almost every claim is a vendor case study, that is worth a great deal.

It is also a company that has visibly chosen its lane. The Nabla of 2026 is an enterprise and mid-market product with a sales cycle attached, a strong multilingual note engine, a dictation story aimed squarely at legacy Nuance deployments, and an agentic roadmap it has not finished shipping. If that is the buyer you are, it is one of the two or three names that should be on your list. If you are a solo clinician looking for a card-swipe scribe, the market has better-fitting options — and Nabla is no longer really pitching to you.

Explore Nabla in our directory, compare AI medical scribes side by side, see what fits small medical clinics or family medicine, or start with our guide to choosing an AI medical scribe.

---

This article is for general information only and is not legal, medical, or purchasing advice. Nabla does not publish pricing; the figures here come from third-party trackers and will change. Trial results are drawn from the peer-reviewed NEJM AI publication; adoption and time-savings figures are drawn from company press releases and vendor materials as of mid-2026 and are labeled as such above. Confirm current pricing, plan inclusions, integrations, data-retention terms, and a signed BAA directly with the vendor before making any decision.