AI Medical Scribe Pricing: What Vendors Publish and What They Do Not

For an individual clinician buying from a website, the answer in August 2026 sits between nothing and about 150 US dollars per month. Heidi publishes a free tier. Freed lists individual plans at 39, 79 and 119 US dollars per month depending on note volume and features, with the top tier at 104 per month when billed annually. Sunoh.ai lists a starting rate of 149 US dollars per user per month, shown against a listed rate of 199.

For an organisation, there is no published answer at all. Abridge, Suki, Ambience Healthcare and Microsoft all quote. That is not evasion so much as a different commercial reality: enterprise deals bundle integration, security review support, training and account management, and the resulting number depends on all four. Expect a range, expect it to be negotiated per contract, and expect volume to move it.

The gap between those two worlds is the most useful thing on this page. If you are one clinician, the whole cost is close to the published number. If you are a health system, the licence may be less than half of what the programme costs you in year one. Our main scribe comparison works through which side of that line you are on before price enters the conversation.

Two, usefully. Freed publishes a full tier list with monthly and annual rates. Sunoh.ai publishes a per user monthly rate. Heidi publishes its tier structure and a free entry point but no amounts. Nabla offers a free trial and published tiers, and we could not retrieve dollar amounts from a public pricing page when we checked in August 2026. Suki, Abridge, Ambience Healthcare, DeepScribe and Microsoft Dragon Copilot publish nothing.

Publishing correlates with distribution rather than with quality. A vendor selling self serve to individual clinicians has to publish, because the buyer will not take a sales call to learn a price. A vendor selling to health systems does not, because the price is one term in a negotiation that also covers retention, integration and support. Neither approach tells you anything about the note quality.

What it does tell you is how the purchase will feel. Published pricing means you can decide this afternoon. Quoted pricing means a discovery call, a security questionnaire, a pilot and a procurement cycle. For a two clinician practice that difference is the deciding factor, which is why our private practice page starts there rather than with features.

Four, and they behave very differently as you grow.

• Per clinician per month. The dominant model. Simple to forecast, and unforgiving where usage is uneven, because a clinician who uses it twice a week costs the same as one who uses it all day.
• Tiered by note volume. Freed's Starter tier caps notes; the tier above it does not. Attractive for part time clinicians and a trap if volume grows mid contract.
• Per encounter or per note. Aligns cost to use, which finance teams like. It also makes the monthly bill unpredictable, which the same finance teams do not like.
• Enterprise contract. A negotiated annual figure, usually with a seat band, often with integration and services folded in. The unit cost is deliberately hard to extract, which is worth insisting on anyway.

When comparing across models, convert everything to cost per clinician per month at your actual volume, then again at one and a half times that volume. The second number is the one that matters, because the plan that looks cheapest at today's volume is frequently the most expensive one at next year's, and switching costs are real once templates and habits have formed.

Four, and together they routinely exceed the licence in year one for anything larger than a small practice.

Integration. Getting notes into the chart as discrete data rather than pasted text is work, and it is usually billed by your EHR vendor or absorbed by your own analysts rather than the scribe vendor. This is the cost that most often surprises organisations that bought a self serve product and then tried to scale it. Our Epic integration page sets out what to ask before you assume it is included.

Training and the adoption curve. Clinicians are slower for the first two to three weeks, not faster. That is a real cost and it appears in your clinic throughput, not on an invoice. Budget for it explicitly or it will look like the product failed.

Note review time. Every ambient note is a draft. The minutes a clinician spends reviewing and correcting are the single largest ongoing cost of the category, and no vendor prices it because no vendor bears it. If review takes four minutes a note across twenty notes a day, that is more than an hour of clinician time daily, which dwarfs any licence fee.

Security review and contracting. For an organisation, this is weeks of internal effort per vendor. It is a strong argument for shortlisting two vendors rather than five, and for sequencing the review early, which is part of what our vendor selection service does.

Start from time, not from money, because the money follows the time and the time is the number you can actually measure. Baseline three figures for two weeks before any tool arrives: minutes spent documenting per encounter, the share of notes closed the same day, and time spent on documentation outside clinic hours.

Then decide what a recovered hour is worth in your setting, and be honest about it. If the clinician goes home earlier, the value is retention, which is real but hard to book. If the clinic adds encounters, the value is revenue and can be calculated. If nothing changes except that the notes are better, the value is documentation quality and coding accuracy. All three are legitimate; they are not interchangeable, and a business case that quietly swaps between them will not survive a finance review.

The AI scribe ROI calculator takes those inputs and produces a break even figure you can hold up against any quote. Do this before the demos. Walking into a demo with no number in your head is how the number ends up being the vendor's, and the vendor's number always assumes the best case adoption curve.

Three things worth stating. We could not retrieve dollar amounts for Heidi's paid tiers or for Nabla's plans from their public pricing pages in August 2026, so both are recorded as tiers published without amounts rather than as a number we found elsewhere. We could not verify any enterprise pricing, because none is published and the figures circulating in third party comparisons are not sourced.

And we could not verify how durable the published prices are. Both Freed and Sunoh.ai display their headline rates as discounts against higher listed prices. That is a normal commercial practice and it means the number you budget against today may not be the number in six months. Where a price matters to your case, get it in a written quote with a term attached rather than relying on a screenshot of a pricing page.

We deliberately did not repeat prices from vendor written comparison blogs, which dominate search results for this query. A page published by one scribe vendor comparing itself to another is marketing, and its numbers are selected accordingly.

Six levers move an enterprise scribe price, and only one of them is the seat count.

1. Term length. Multi year commitments move price more than volume does at mid scale.
2. Seat bands with a true up. Better than a flat headcount, because clinician adoption is never uniform and you should not pay for seats nobody opened.
3. Integration scope. Ask what is included and what is billed separately. This is where quotes diverge most.
4. Retention and model training terms. These are negotiable at the enterprise end and are worth real money in risk terms. Our HIPAA posture comparison sets out the clauses to ask for.
5. Pilot credits. Many vendors will credit pilot cost against a first year contract if you ask before the pilot, and not if you ask after.
6. Exit terms. Data export format and cost. Cheap to negotiate at signature and expensive later.

The single most effective move is to name your ceiling early and hold it. Vendors price against perceived willingness to pay, and an organisation that has done its own modelling is a different negotiation from one that has not.

Not at drafting notes. The gap in note quality between a 79 dollar per month product and a negotiated enterprise contract is far narrower than the price difference suggests, and for a solo clinician it may be invisible. What the enterprise price buys is everything around the note: integration depth, security review support, negotiated contract terms, training resources and a named account team.

If you do not need those things, you are not getting less for less. If you do need them and try to buy at the low end, you will pay for them anyway through your own staff time, and you will pay more because your team is doing it for the first time. That is the actual trade, and it is a question about your organisation rather than about the products.

The corollary matters too. A large organisation that buys a self serve product department by department ends up with several contracts, no integration, no consistent retention terms and no inventory of where patient data went. That is a governance problem created by a procurement shortcut, and it is more expensive to unwind than the licence saving ever was. The HIPAA and AI compliance page covers what that inventory needs to look like.

Measure your own documentation time for two weeks. It costs nothing, it takes ten seconds a day per clinician, and it is the only number in this entire process that is genuinely yours. Most organisations skip it and then find themselves arguing about whether the deployment worked with no baseline to argue from.

Then set a ceiling using the ROI calculator, shortlist two products from the side of the market you actually belong to, and run them in the same specialty over the same period. If you are a solo clinician, start with the free tier and the published prices, since the entire evaluation can be done this month for very little money. If you are an organisation, the sequencing question is bigger than the pricing question, and that is what a deployment roadmap is for.

The organisations that regret this purchase are almost never the ones that paid too much. They are the ones that bought before they knew what they were measuring, and then had no way to tell whether the money had done anything.