The Recording is On, Are the Patients Still Telling the Truth

Percy Bhathena is VP of Product & AI at DeliverHealth, overseeing InstaNote and InstaCode, the company's ambient documentation and AI medical coding platforms.

Ambient AI documentation is built on a simple and compelling premise: record the clinical encounter, reduce the documentation burden, give time back to the clinician, and improve the quality of care.

That premise is real. The efficiency gains are documented. The reduction in physician burnout is measurable. And when ambient documentation is built to feed cleanly into downstream workflows, coding, prior authorization, referral management—the compounding value across the revenue cycle is significant.

But a peer-reviewed study published in JAMA Digital Health has revealed a finding that the ambient AI industry has largely kept quiet about. And it deserves a direct, honest response.

When patients know they are being recorded, they are significantly more likely to self-censor on sensitive health topics. The study surveyed 103 patients. The findings are specific: 35% reported they would censor themselves when discussing mental health concerns, 40.8% when discussing sexual health, and 51.5% when discussing illicit activity. Separately, 77.7% of patients wanted to know whether discussing illegal activity in a recorded encounter carried legal risks. And when patients received full details about AI involvement, data storage, and corporate access, rather than only basic information, consent to recording dropped from 81.6% to 55.3%.

Those numbers should give every ambient AI company pause, including ours.

What the Data Actually Tells Us 

The study is not a verdict against ambient documentation. It is a data point about human behavior, specifically, about how the presence of a recording device changes what people choose to say in a clinical setting.  

Patients are not being irrational. The concern about permanent records, about who has access to a verbatim recording, about the difference between a physician’s clinical note and their own words captured in full, these are reasonable responses to a technology that is genuinely new to the clinical encounter. Patients have had decades to grow accustomed to the idea that physicians document their visits. The ambient recording of their exact voice, in their exact words, is a different experience. And for many patients, it feels meaningfully less private. 

The clinical responses to this shift are serious. The categories of information patients are withholding, sexual health history, substance use, behavioral risks, and are precisely the ones that inform diagnosis, treatment planning, medication decisions, and appropriate clinical documentation. A note that is missing this information is not just incomplete. It is less useful to the clinician, less accurate for coding purposes, and potentially less safe for the patient.

Downstream Effect On Documentation and Revenue Cycle 

When the clinical note is incomplete, the effects extend beyond the encounter. Coding accuracy depends on the quality and specificity of the documentation. InstaCode, DeliverHealth’s AI medical coding solution, works in concert with InstaNote to translate clinical encounters into accurate codes and faster reimbursement, but only when the underlying documentation reflects the full clinical picture. Incomplete disclosure creates an upstream gap that compounds downstream. 

This is Not a Technology Problem 

The reflexive response to this finding might be to reconsider the technology. That would be the wrong conclusion here. 

The technology records accurately. The note reflects what was said. The issue is not with what the system captures; it is with what the patient decides not to say. No software update addresses that gap because it is not in the software itself.  

This isn’t a technology issue. It’s a trust issue, and that trust issue must be resolved by the clinician, because they’re the only other person in the room.
- Percy Bhathena

That framing is the most important because it shifts the response to where it belongs, the clinical relationship. Patients need to understand what is being recorded, why it is being recorded, where it goes, how long it is retained, and who has access to it. In the absence of clear answers to those questions, they default to caution. That is a rational response, and it is one that clinical education and communication practices can address.

The Moment Before the Recording Starts 

The intervention that matters most happens before the recording begins. 

Clinicians who have been practicing for fifteen or twenty years have developed a natural clinical rapport that accounts for sensitive topics. They know when to pause, when to ask directly, and when to follow an indirect signal. What they have not previously needed to do is explicitly explain their documentation process to patients before the encounter begins. Ambient recording changes that. 

The introduction of the recording device requires a brief, clear, human-centered explanation, and the framing of that explanation makes a meaningful difference in patients' responses. Two versions of the same conversation produce different results. 

Two Framings. One Outcome. 

“I’m going to record our conversation today. It goes into your permanent medical record.” vs. “I use a tool that records our visit so I can focus on you instead of the keyboard. It creates the same notes I’d always put in your chart, nothing more, nothing less.” Both are accurate. One builds trust. The other activates concern about permanence, leading patients to self-sensor. 

This is a trainable skill. It is also something that ambient AI vendors need to take seriously as we consider consent flows, onboarding materials, and the guidance we provide to clinical teams deploying our tools. The technology does not cause the disclosure problem on its own. But how it is introduced determines whether that problem gets better or worse. 

What Responsible Deployment Looks Like 

The ambient AI companies that take this finding seriously will do several things differently from those that do not. 

First: honest conversations with clinical partners. Deploying an ambient documentation tool without acknowledging that it may alter patient disclosure patterns is a disservice to the clinicians who rely on it. They deserve the full picture, including the parts that create new workflow requirements for them. 

It’s incumbent upon us to have very honest conversations with clinicians. If it’s real data, we should understand the impact of technology, regardless of whether it’s good or bad.
-Percy Bhathena

Second: watching the longitudinal data. The JAMA Digital study is a significant signal, but it represents a single study at an early stage of ambient AI adoption. The industry needs more data across specialties, patient populations, and deployment models before drawing definitive conclusions about what reliably mitigates the disclosure effect. Acting on one study with wholesale product changes would be premature. Ignoring it entirely would be a mistake. 

What the Industry Needs to Commit To 

The ambient AI documentation category is moving quickly. The efficiency gains are real. Burnout reduction is measurable. The downstream revenue cycle improvements, when documentation is structured to feed cleanly into coding and billing workflows, are among the most compelling ROI stories in healthcare AI today. 

Ambient Documentation and Revenue Cycle: The Connected Workflow 

InstaNote and InstaCode are designed to work as a connected system, from the spoken word in the clinical encounter to the paid claim. Learn more about the InstaNote + InstaCode bundle and how DeliverHealth's document-to-dollar workflow helps health systems reduce denials and accelerate reimbursement. 

But speed without accountability creates its own problems. The industry cannot deploy ambient documentation tools at scale while remaining silent about their behavioral effects on patients. The trust deficit that drives self-censorship disclosure is already present in the data. Left unaddressed, it will compound and eventually surface as an adoption problem, a compliance problem, or a care quality problem. 

What the industry needs is a shared commitment to transparency: about what is recorded and how it is used, about what ambient tools change in the clinical relationship, and about the ongoing research that tells us whether our practices are working. That commitment is not optional for companies that intend to be in this market for the long term. 

Patients who choose not to disclose are not making an irrational decision. They are responding to uncertainty. Reducing that uncertainty, through better clinician education, clearer consent conversations, and honest vendor communication, is not a competitive differentiator. It is a basic responsibility. 

The companies that earn lasting trust in this category will be the ones willing to have these conversations now, before the data forces the issue. 

About the Products In this Article 

InstaNote is DeliverHealth's ambient AI documentation solution. By recording and structuring the clinical encounter, InstaNote eliminates the documentation burden that contributes to physician burnout and off-hours charting, returning time to the clinician and presence to the patient encounter. InstaNote integrates deeply with leading EHRs, including Epic, Oracle Health, and athenahealth, so clinicians remain in the workflows they already use. Physicians also have the option to switch between ambient documentation and dictation, providing flexibility for patients who may be hesitant about the ambient recording. 

InstaCode is DeliverHealth's AI-assisted medical coding solution. Built to work in concert with InstaNote, InstaCode processes clinical documentation through an autonomous coding engine and surfaces coding suggestions and discrepancies for human review, reducing denial rates, accelerating reimbursement turnaround, and ensuring that the documentation captured in the encounter translates into accurate, defensible codes. 

Together, InstaNote and InstaCode form the foundation of DeliverHealth's mid-revenue cycle workflow, a connected system that supports the clinical encounter from the first word spoken to the final claim paid. Learn more at deliverhealth.com or contact our team to discuss how DeliverHealth can support your organization's ambient AI and revenue cycle strategy. 

About the author:

Percy Bhathena leads product and AI strategy at DeliverHealth, where he oversees the development of InstaNote and InstaCode — DeliverHealth's ambient documentation and AI medical coding platforms. With a background spanning enterprise technology and healthcare AI, Percy focuses on building tools that augment clinical workflows rather than replace the human judgment at their center. He speaks and writes on the intersection of AI, clinical practice, and responsible product development.

Sources:
Lawrence K, Kuram VS, Levine DL, et al. Informed Consent for Ambient Documentation Using Generative AI in Ambulatory Care. JAMA Netw Open. 2025;8(7):e2522400. Published 2025 Jul 1. doi:10.1001/jamanetworkopen.2025.22400