Freed AI Scribe: Workflow, Privacy, Pricing, and Evidence (2026)

Freed AI scribe turns a clinical conversation into a draft note for clinician review. Its current documentation covers capture, editing, templates and transfer to an EHR. The buying decision is whether the plan, recording setup and final handoff fit your practice—not whether an impressive draft can be produced once. Freed getting-started documentation.

Our assessment: start with Starter for low-volume documentation, Core for unlimited notes, Premier when browser-EHR push is a requirement, and Groups for multi-clinician administration. The decision table below explains where each recommendation stops. A feature list does not establish specialty accuracy, and this article does not rank Freed against Vero on measured clinical performance.

Publisher disclosure: Vero sells competing AI scribe software. This is a public-evidence assessment by Sam Ellis, not an independent clinician’s hands-on review. Sources checked September 18, 2026. No live Freed encounter, product screenshot session or comparative benchmark was performed for this article, and it has not received separate clinical review. See our editorial and corrections policy.

Freed AI scribe pricing

These direct-subscription prices were checked against Freed’s plan and billing help article on September 18, 2026. Dollar amounts are USD; confirm taxes and your checkout total. An annual equivalent is not a monthly installment price.

On smaller screens, swipe or scroll each table horizontally to see all columns. Keyboard users can focus the table region and use the arrow keys.

Freed advertises a seven-day Premier trial without payment information. Its help article also describes 12 months of free Core for eligible trainees and up to 50% off an eligible monthly Premier subscription purchased directly through Freed; annual and app-marketplace subscriptions are excluded from that discount. These are eligibility-based offers, not a permanent free plan for every clinician. Current trial and eligibility details.

The public pricing page places template building and Clinical Evidence under Core, and lists EHR push, patient context, dictation, saved phrases and coding features under Premier. Group features include administration, note sharing and SSO. One source discrepancy matters: the billing help article also lists Clinical Evidence in Starter. Ask Freed to confirm the entitlement in your account before purchasing for that feature.

App-store billing is different. Freed says Apple and Google marketplace subscriptions have a flat monthly price rather than Starter/Core/Premier tiers and may cost more. A direct subscription still works in the mobile app and across supported devices using the same account; buying through a store is not required for mobile access. Compare the checkout price and included capabilities, not just the plan name. Freed’s marketplace and device guidance, checked September 18, 2026.

Choose a tier using your actual workload

These are Vero’s buyer-fit inferences from Freed’s published plan features, checked September 18, 2026—not product-test rankings.

Core paid for 12 months totals $948 at the listed monthly rate. Premier totals $1,428 on monthly billing, versus $1,248 prepaid annually: a $180 difference before taxes. Annual commitment makes more sense after testing than before it. Do not describe a monthly subscription annualized for budgeting as an annual contract price.

Separate subscription cost from labor. For an illustrative 160 encounters per month, an extra 90 seconds of review per encounter equals four hours. At a clinic-assigned $100 hourly value, that is $400 of time—not necessarily recoverable revenue. This example has no measured Freed inputs. It shows why correction, transfer and failed-session recovery belong in purchasing decisions.

Cancellation, renewal and exporting your notes

Freed’s subscription-management instructions, checked September 18, 2026, describe individual-plan cancellation as stopping renewal after the current paid cycle. Store purchases are managed through Apple or Google; Groups users should contact their account manager. The direct trial does not automatically become a paid subscription.

The August 13 platform terms, sections 9 and 18, checked September 18, contain provisions to reconcile before buying:

• Notice: section 18 specifies at least 30 days’ nonrenewal notice, while section 9 describes cancellation at the billing-cycle end.
• Prepayment: section 9 says prepaid subscription fees are not refunded.
• Export: section 18 provides reasonable assistance for data still held by Freed, on written request within 30 days after termination, in a mutually agreed format. After that window, storage obligations end and deletion follows, with backups handled under retention policies.

Ask Freed to confirm in writing which notice deadline applies to your plan and purchase channel, the final charge and access dates, and the export format, scope, timing and any cost. Do not assume the help article overrides the contract or that the export window restores already-deleted data. Verify a usable export before ending access; subscription cancellation, account deletion and termination should not be treated as interchangeable events.

The documented Freed workflow

This walkthrough follows documentation checked September 18, 2026; it is not an observed usability test. Freed’s getting-started guide describes a redesigned interface and notes that iOS and extension users may see an older layout.

1. Identify the visit. Enter the appropriate encounter information and establish the permitted capture boundary.
2. Capture the conversation. Freed documents “Capture conversation” and “End visit” controls.
3. Review the draft. Check the note, edit text and choose or change the template. The documented transcript route is the visit’s More options menu, then Transcript.
4. Transfer reviewed content. Copy the whole note or sections, or evaluate the Chrome-extension handoff.

The clinical checkpoint is broader than spelling. Compare what the patient reported, what another speaker reported, what the clinician observed and what remains uncertain. A polished examination section is not trustworthy merely because it sounds conventional. If the source contains no examination finding, establish where that information came from.

Inspect the final record after transfer too. A correct draft can become an incomplete chart entry if a field truncates text, a paragraph lands in the wrong section, or a second push duplicates content. Decide who authenticates the note and how an uncompleted handoff is surfaced before routine use.

What EHR push does—and what still needs testing

Freed describes EHR push as a Chrome-extension route for browser-based EHRs, naming athenahealth, Practice Fusion, Tebra and SimplePractice among examples. That documents a transfer mechanism, not proof that your configuration supports every field or a bidirectional clinical interface.

Run a permitted test in the exact destination: select the correct patient and encounter, transfer a synthetic note, inspect section mapping, save, reopen, and check for duplication. Ask how extensions are deployed on managed workstations and what happens after an EHR interface update. A desktop or virtualized EHR should not be assumed compatible because another installation with the same brand works. Our EHR integration guide separates text transfer from clinical workflow integration.

Telehealth, mobile and multilingual encounters

Freed documents a Zoom Marketplace app that joins hosted meetings as a clinical notetaker and processes the encounter after the call. Its instructions cover account connection and permissions, including holding the notetaker in a waiting room while obtaining consent. That is a specific Zoom workflow, not proof of equivalent behavior on every video platform.

Use two test participants on separate devices. Each should speak a unique phrase, then check that both phrases reached the transcript and were attributed correctly. Repeat with the headphones and microphone normally used by the clinician. A successful local-microphone test cannot establish remote-channel capture.

Freed’s pricing FAQ lists multilingual capture with English-note output. Evaluate the actual language pair with a qualified reviewer; test negation, speaker changes and translated clinical terms. Do not infer equal accuracy across languages, accents or specialties from a language list. Rehearse mobile interruptions rather than assuming web results predict phone behavior.

Privacy and contract checks

Privacy is a configuration and contracting question, not a badge comparison. Freed’s security page, checked September 18, states U.S. storage, encryption and SOC 2 Type 2 status. It describes temporary recording storage followed by deletion, and configurable note deletion. Request the relevant assurance documentation and match its scope to the service being purchased.

Read that explanation alongside the August 13, 2026 platform terms, especially sections 7, 8 and 15 and Schedule A. These are checkpoints for organizational review, not a legal determination. Freed platform terms and BAA.

Resolve the difference between public deletion summaries and configurable retention in writing. The terms also restrict competitive analysis during trials; a competitor-led benchmark requires appropriate authorization.

Freed’s August 11, 2026 privacy policy separately addresses account/contact information, usage tracking and support-chat records. Do not put patient material into a sales or support conversation merely because the clinical application has a BAA. Ask for the approved support channel and minimum information needed to investigate an issue.

For U.S. covered entities, HHS cloud guidance makes the BAA, risk analysis and safeguards relevant to cloud use. A security statement does not complete the clinic’s assessment. Identify access owners, offboarding procedures, incident escalation and the documentation fallback for outages.

For Canada, the Office of the Privacy Commissioner’s jurisdiction overview distinguishes federal, provincial and health-sector requirements. U.S. storage alone does not establish either permission or prohibition for every Canadian practice. Review the applicable province, custodian duties, cross-border arrangements and contracts rather than treating PIPEDA as a product certification.

What the evidence can and cannot show

Official documentation establishes what a vendor describes and sells. A customer story identifies an experience worth investigating. A controlled study evaluates an intervention under a defined design. None automatically supplies the missing evidence from another category.

For example, Afshar and colleagues’ 2025 randomized trial enrolled 66 practitioners over 24 weeks in ambulatory clinics in two U.S. states. The full report identifies Abridge as the tested system, not Freed. Work exhaustion/interpersonal disengagement improved by 0.44 points; professional fulfillment did not improve significantly under the study’s analysis. Documentation time decreased by 0.36 hours per day (about 22 minutes; 95% CI, 0.17–0.55 hours saved). The after-hours-work finding was sensitive to extreme observations and lost statistical significance when the top 3% were excluded. These findings support evaluating ambient documentation, not claiming equivalent Freed results or superiority over Vero.

We have not measured Freed’s error rate, generation latency or correction time. That is an evidence boundary, not evidence of poor performance. A credible product result needs the tested workflow, sample, raw outputs, scoring rules and reviewer qualifications.

Distinguish raw-draft quality from acceptable final-record effort. One error may be obvious and quickly corrected; plausible wording may require substantial investigation. Report both endpoints, retaining failed sessions in the denominator. Otherwise, apparently strong performance might describe only the easiest completed visits.

A reproducible synthetic evaluation

FREED-EVAL v1.0 — proposed protocol, not executed. Intended users are authorized clinic evaluators. Check account terms before testing and obtain permission for competitor-led testing or public release of screenshots and output where required. Use fictional inputs, not patient recordings.

Perform two repetitions of each script below: one quiet-room condition and one repeatable background-noise condition, for 12 planned sessions. This small workflow rehearsal is not a representative accuracy study. Keep delivery unchanged between repetitions. Use the same nonclinical background sound at a recorded volume and distance; if the sound level is not measured, label it unmeasured.

Record the date, plan, app/build identifier if available, device, operating system, browser version, microphone, speaker distance, room conditions, network, template and destination EHR. Record unavailable version information as unavailable. Freeze reference scripts before capture so expected answers cannot be rewritten to match the generated output.

Score the draft before correcting it

Preserve the initial draft and transcript. Mark expected facts as retained, omitted, altered or unsupported. Classify attribution, negation, numbers/timing, uncertainty and plan errors separately. A qualified clinical reviewer should define severity and escalation rules before seeing results. Use a second reviewer for disputed cases; a single numeric score should not resolve clinical significance.

Calculate transcript word error rate against the exact reference speech, using a recorded normalization rule and counting substitutions, deletions and insertions. The AI medical transcription guide explains the calculation. A summarized note is not a verbatim transcript: do not report its word differences as transcription WER.

Measure active setup, review, correction, transfer and recovery separately from processing wait. Record failures, intentional stops and unacceptable final records. A short successful rehearsal should lead to a monitored pilot, not a claim of universal safety.

Example completed evaluation: F-01

Example only: no product was run; no time was measured.

Invented candidate: “The patient fell last week.”

Finding: incorrect speaker attribution and unsupported patient fall history. Raw-draft result: fail.

Corrected meaning: the patient denies falling; the companion reports their own fall last week.

Disposition: retain the candidate and correction; a clinical reviewer assigns severity. This constructed error is not an observed Freed defect.

For an authorized published test, capture dated screenshots of template selection, recording state, raw output and final transfer. Annotate where expected facts are retained or lost. No product screenshots are supplied here because none were captured in a hands-on test; a mock interface would not fill that evidence gap.

Freed or Vero: which workflow fits?

These are buyer-fit inferences from documentation checked September 18, not measured comparative verdicts. Vero is the publisher and has a commercial interest in this comparison.

Shortlist Freed using the situation-specific table above: limited volume, unlimited notes, browser-EHR push or team administration. Its documented subscription ladder makes those options distinguishable. Defer purchase if the required tier, transfer compatibility or exit terms remain unresolved.

Consider Vero when your workflow starts with several source types. Vero’s note-creation documentation describes recording, typed context and uploaded documents, images or audio in one encounter, followed by template selection, review and copy/export. Mixed-source documentation is a concrete reason to evaluate Vero—not proof that Freed cannot meet a similar requirement. Ask both vendors to demonstrate the same input set.

Vero also documents that two patient encounters cannot be recorded simultaneously and that active recording must stop before switching encounters. Include that limitation in the assessment. Compare the reviewed final record and handoff effort, not the number of icons on a features page. Explore the Vero scribe workflow and use the digital-scribe pilot scorecard for consistent evaluation.

Do not choose either product while a mandatory requirement remains unresolved: a prohibited record type, required data location, missing enterprise control, unsupported EHR route or unacceptable test defect. Record the requirement, evidence owner and decision deadline. “Not verified” should remain an open procurement item, not silently become “supported” or “unsupported.”

A practical purchase decision

Before paying, assemble four pieces of evidence: the exact quote, approved data-handling configuration, completed synthetic rehearsal and end-to-end handoff in the intended environment. Then run a monitored pilot with a clinical owner and documented fallback. The ambient capture guide provides detailed speaker-boundary and recovery scripts.

Choose the product that produces an acceptable reviewed record with manageable total effort under your actual constraints. Recheck after material template, model, app or EHR changes. Understand export, renewal and offboarding before the first routine encounter, not after becoming dependent on the service.

Written by Sam Ellis, a Vero contributor covering AI-assisted documentation, patient-care workflows and healthcare privacy. Product and contract sources were checked September 18, 2026; separate clinical review remains outstanding.