Artificial intelligence chatbots have emerged as the most significant health technology hazard for 2026, according to a new report from ECRI, an independent, nonpartisan patient safety organization.
The finding leads ECRI’s annual Top 10 Health Technology Hazards report, which highlights emerging risks tied to healthcare technologies that could jeopardize patient safety if left unaddressed. The organization warns that while AI chatbots can offer value in clinical and administrative settings, their misuse poses a growing threat as adoption accelerates across healthcare.
Unregulated Tools, Real-World Risk
Chatbots powered by large language models, including platforms such as ChatGPT, Claude, Copilot, Gemini, and Grok, generate human-like responses to user prompts by predicting word patterns from vast training datasets. Although these systems can sound authoritative and confident, ECRI emphasizes that they are not regulated as medical devices and are not validated for clinical decision-making.
Despite those limitations, use is expanding rapidly among clinicians, healthcare staff, and patients. ECRI cites recent analysis indicating that more than 40 million people worldwide turn to ChatGPT daily for health information.
According to ECRI, this growing reliance increases the risk that false or misleading information could influence patient care. Unlike clinicians, AI systems do not understand clinical context or exercise judgment. They are designed to provide an answer in all cases, even when no reliable answer exists.
“Medicine is a fundamentally human endeavor,” said Marcus Schabacker, MD, PhD, president and chief executive officer of ECRI. “While chatbots are powerful tools, the algorithms cannot replace the expertise, education, and experience of medical professionals.”
Documented Errors and Patient Safety Concerns
ECRI reports that chatbots have generated incorrect diagnoses, recommended unnecessary testing, promoted substandard medical products, and produced fabricated medical information while presenting responses as authoritative.
In one test scenario, an AI chatbot incorrectly advised that it would be acceptable to place an electrosurgical return electrode over a patient’s shoulder blade. Following such guidance could expose patients to a serious risk of burns, ECRI said.
Patient safety experts note that the risks associated with chatbot misuse may intensify as access to care becomes more constrained. Rising healthcare costs and hospital or clinic closures could drive more patients to rely on AI tools as a substitute for professional medical advice.
ECRI will further examine these concerns during a live webcast scheduled for January 28, focused on the hidden dangers of AI chatbots in healthcare.
Equity and Bias Implications
Beyond clinical accuracy, ECRI warns that AI chatbots may also worsen existing health disparities. Because these systems reflect the data on which they are trained, embedded biases can influence how information is interpreted and presented.
“AI models reflect the knowledge and beliefs on which they are trained, biases and all,” Schabacker said. “If healthcare stakeholders are not careful, AI could further entrench the disparities that many have worked for decades to eliminate from health systems.”
Guidance for Safer Use
ECRI’s report emphasizes that chatbot risks can be reduced through education, governance, and oversight. Patients and clinicians are encouraged to understand the limitations of AI tools and to verify chatbot-generated information with trusted, knowledgeable sources.
For healthcare organizations, ECRI recommends establishing formal AI governance committees, providing training for clinicians and staff, and routinely auditing AI system performance to identify errors, bias, or unintended consequences.
Other Health Technology Hazards for 2026
In addition to AI chatbot misuse, ECRI identified nine other priority risks for the coming year:
Unpreparedness for a sudden loss of access to electronic systems and patient data, often referred to as a digital darkness event
Substandard and falsified medical products
Failures in recall communication for home diabetes management technologies
Misconnections of syringes or tubing to patient lines, particularly amid slow adoption of ENFit and NRFit connectors
Underuse of medication safety technologies in perioperative settings Inadequate device cleaning instructions
Cybersecurity risks associated with legacy medical devices
Health technology implementations that lead to unsafe clinical workflows
Poor water quality during instrument sterilization
Now in its 18th year, ECRI’s Top 10 Health Technology Hazards report draws on incident investigations, reporting databases, and independent medical device testing. Since its introduction in 2008, the report has been used by hospitals, health systems, ambulatory surgery centers, and manufacturers to identify and mitigate emerging technology-related risks.
Over the past decade, patients have steadily shifted from word-of-mouth referrals to digital search when making healthcare decisions. Today, that evolution is accelerating even faster as artificial intelligence (AI) tools, not traditional search engines, emerge as the new front door to finding care.
Instead of spending time talking to friends or browsing through pages of Google results, patients now often simply ask ChatGPT, Google’s AI Overviews, and other consumer AI assistants a simple question: “Who is the best doctor near me?” These tools don’t return lists anymore. They return answers. And which doctors are recommended depends on signals most practices still don’t understand or fully control.
The result? A patient visibility vortex is emerging, where AI will decide which providers appear, which disappear, and which rise above their competitors in 2026 and beyond.
AI is Rewriting the Patient Search Journey
According to rater8’s 2025 Patient Preferences Survey, 31% of patients already use AI tools to research providers. Even more striking: 52% trust AI results as much as or more than traditional search. This shift is accelerating.
AI models now ingest large volumes of public information: practice websites, review sites, news articles, directory listings, Reddit posts, and social media posts. They synthesize all of it into a single recommendation.
As John Bulmer, Public Information Officer at Capital Cardiology Associates, observed in a recent rater8 panel webinar: “Your website is no longer the first place prospective patients meet you — it may be the second or third. And now, with AI pulling information from sources you may not even realize, your broader online presence has never mattered more.”
The old rules of patient search habits no longer apply. Online visibility isn’t earned once and done; it must be constantly maintained because AI evaluates recency, consistency, and credibility across every corner of the internet. Practices that can’t keep pace risk becoming digitally invisible, even if they provide exceptional care in real life.
Inconsistent Information is Killing Your Online Visibility
When AI tools scan the web, they look for clarity. If a practice’s online presence is fragmented or difficult to parse (e.g., different hours listed across directories, mismatched provider bios, or outdated service information), AI hesitates to recommend that practice.
This is where many practices fall behind. Their information may be technically available, but it isn’t standardized. Imagine telling someone to visit your practice, but on one map the building is open, on another it’s closed, and on a third the doctor they’re trying to see doesn’t even work there anymore. That inconsistency erodes trust instantly.
Practices that maintain consistent provider names and credentials, matching hours and phone numbers across major directories, schema-optimized provider pages, and regularly updated content give AI confidence. That confidence translates directly into recommendations.
As healthcare consumerism moves into an AI-first model, structured data will become the new digital bedside manner that signals accuracy, reliability, and professionalism before the patient ever walks through the door.
The Power of the Patient Voice in the Age of AI
Of all the signals AI consumes, verified patient feedback has emerged as one of the most powerful trust indicators.
Unlike testimonials or website copy, verified reviews provide rich, unfiltered, keyword-dense sentiment about the patient experience. AI systems favor this content because it’s recent, specific to the provider, generated by real patients, and difficult to manipulate. This explains why many practices with strong clinical reputations still underperform in AI-driven search. They lack the volume and recency of patient-generated content that AI models prioritize.
Verified reviews, particularly those captured through structured, patient-initiated systems, give AI the credibility it needs to confidently recommend a provider. These reviews also reduce the influence of outdated or unrepresentative feedback, helping practices build a more balanced and accurate online reputation.
Preparing for the Visibility Vortex of 2026
As AI assistants become the default method of care navigation, practices need to think less about SEO tactics and more about visibility ecosystems. That includes:
Maintaining structured, consistent practice and provider data across all major directories
Building a steady cadence of verified patient reviews across Google, Healthgrades, WebMD, and specialty sites
Creating clear, AI-readable content that answers common patient questions
Monitoring how AI interprets their brand across platforms
The patient search process is changing faster than most organizations realize. But with the right strategy, healthcare providers can position themselves at the center of this visibility vortex: earning trust, improving transparency, and making sure their best physicians are the ones AI recommends next.
By Emily Bonham, senior vice president of product management, AGS Health.
In healthcare revenue cycle management (RCM), we’ve long relied on automation systems that process rules-based workflows with limited or no need for complex logic and nuanced judgement. Robotic Process Automation (RPA) has been highly effective at automating repetitive, high-volume tasks such as claim status checks and data entry.
However, its limitations are increasingly apparent. Today’s revenue cycle challenges demand more than just speed and efficiency; they require adaptability, context, and intelligent decision-making.
That’s where agentic AI comes in.
Agentic AI represents a next-generation approach to automation—one that mimics how humans think, make decisions, and interact with systems and people. Unlike RPA, which follows strict, predefined scripts, agentic AI models operate as autonomous agents. They’re context-aware, goal-oriented, and capable of reasoning across complex workflows. For revenue cycle teams under pressure from rising denials, staffing shortages, and shrinking margins, this kind of intelligence isn’t just nice to have—it’s becoming essential.
What Makes Agentic AI Different?
The simplest way to explain agentic AI is to compare it to a seasoned team member—one who not only knows how to complete a task but also when to escalate, adapt, or reprioritize based on changing circumstances. Agentic systems can:
Interpret and act on real-time data from multiple sources
Make decisions without human intervention
Learn from patterns and improve over time
Collaborate with human team members when needed
In practical terms, this means AI can now triage claims, initiate and complete payer calls, route work dynamically, or even autonomously document and code encounters—all with logic and consistency.
Why This Matters for RCM
Healthcare RCM is a perfect candidate for agentic automation because it sits at the intersection of structure and unpredictability. Processes are highly regulated, but real-world conditions vary constantly. Consider these examples:
Accounts receivable: Agentic AI can identify which claims require expert attention and which can be resolved through automation, ensuring staff spend their time where it’s most needed.
Insurance follow-ups: AI agents can navigate payer phone trees, wait on hold, retrieve claim information, and even update the EHR, without tying up human resources.
Denial management: Instead of flagging a denied claim for review, an agentic system can analyze the denial reason, check documentation, and suggest or initiate corrective actions.
These aren’t distant possibilities—they’re already being piloted and implemented in real-world environments.
The Human + Agentic AI Model
It’s important to note that agentic AI is not about replacing people—it’s about augmenting them. The most effective models combine human oversight with AI execution:
Human experts oversee automated workflows, handle edge cases, make nuanced judgment calls, or perform relationship-driven tasks.
AI agents handle high-volume, rule-governed, or low-dollar work with consistency and speed, while equipping staff members with insights and suggested actions.
This hybrid approach doesn’t just improve throughput; it also enhances job satisfaction for teams that no longer spend their days on tedious follow-ups or simple reconciliations.
Getting Started with Agentic AI
For organizations beginning to explore this space, here are a few guiding steps:
Consolidate and clean your data: Fragmented data across EHRs, billing systems, and vendor platforms limits AI effectiveness. Start by creating interoperable, governed data environments.
Identify high-ROI use cases: Look for repeatable processes with moderate complexity and clear financial upside, like denial prediction, prior authorization automation, or A/R follow-ups.
Experiment with short feedback loops: Choose pilots where you can quickly assess ROI and adjust based on results. Don’t aim for perfection—aim for momentum.
Build trust through transparency: Ensure your AI systems are auditable and explainable, especially when financial decisions are being made autonomously.
A Path to Sustainable Margins
Every healthcare leader is being asked to do more with less: deliver care, navigate compliance, and protect financial performance. Those who lead with tech-forward cultures by embracing intelligent automation and prioritizing data cleanliness in their revenue cycle operations are well-positioned to rise to the occasion. In contrast, those who resist innovation due to skepticism or overly protective and risk-averse policies risk falling behind—exposing their financial performance to volatility and long-term disruption.
Agentic AI offers a path forward, not as a magic bullet, but as a powerful tool for reclaiming time, improving accuracy, and aligning resources where they have the most impact.
It’s still early days for agentic AI in healthcare RCM, but the direction is clear. With the right balance of vision and pragmatism, revenue cycle leaders can unlock a new level of operational intelligence and move closer to sustainable, value-driven performance.
When it comes to providing comprehensive preventive care, more than annual checkups are required. A patient-facing approach is needed to support the establishment of a health-forward routine comprised of proper nutrition and exercise, mental and emotional health, and chronic condition management.
Numerous studies back this preventive care model, which produces wide-ranging benefits, from improved health outcomes to lower care utilization to lower healthcare costs.
Unfortunately, despite its known benefits, the US healthcare system has been slow in its uptake of preventive care for many patients; just 8% of Americans currently undergo routine preventive screenings and care. As a result, the US loses about $55 billion annually due to missed prevention opportunities. That’s about $0.30 for every dollar spent on healthcare services.
Reversing these trends is the driving force behind the growing interest in lifestyle or preventive medicine and the adoption of technology tools to support physicians and patients in this approach.
Emergence of Lifestyle Medicine
Lifestyle medicine uses evidence-based, whole-person, prescriptive behavioral, and therapeutic lifestyle interventions to prevent, treat, and manage chronic diseases such as cardiovascular diseases, type 2 diabetes, and obesity. By integrating the six pillars of lifestyle medicine—nutrition, physical activity, stress management, restorative sleep, social connection, and avoidance of risky substances—into patient care, lifestyle medicine helps patients improve their health and well-being.
A lifestyle or preventive approach to medicine can address up to 80% of chronic diseases and potentially reverse the decades-long rise in the prevalence of chronic conditions and associated costs. It can also improve both patient and provider satisfaction. This aligns with the Quintuple Aim of better health outcomes, lower cost, improved patient satisfaction, improved provider well-being, and advancement of health equity.
According to one report, business executives mentioned Artificial Intelligence (AI) more than 30,000 times in earnings calls at the end of 2023. AI, and debates around fears, capabilities and ethics have dominated discussions in both the board room and at the water cooler in most industries. I’ve experienced several major technology shifts and innovations throughout my career but the buzz around AI is groundbreaking.
In behavioral health, we’re talking about AI every day and uncovering how it can be a great complement to other technologies used in treatment centers and practices. Our provider clients have reported it boosts note-taking and documentation processes with improvements in accuracy and efficiency. Data produced by Eleos shows providers have reported that documentation time has been reduced up to 50%, allowing clinical teams to spend more focused time with patients.
Providers leveraging AI also said they have 90% of their notes submitted within 24 hours, reducing documentation backlog and avoiding denials due to late submissions. Another key benefit is care teams indicate they’re able to use AI insights to deliver evidence-based best practices, which is excellent for improving patient outcomes.
Testing the waters
Our colleagues at All Points North (APN), a multi-site, 77-bed Behavioral Health system based in Colorado, decided to move ahead with an AI solution. APN was already using Kipu’s EMR, so they chose to go with Eleos, which is integrated with the Kipu EMR. Eleos focuses on supporting documentation and note taking in therapeutic sessions through their AI solution, which was a key area APN was hoping to improve.
Andrea Boorse, senior manager of operations at APN, shared that their clients have two individual therapy sessions each week, rather than one—which means double the documentation. When they became aware of AI solutions that could listen in on sessions and help with that documentation, they decided to test the waters with Eleos’s Scribe solution, which automatically transforms raw conversations into progress note suggestions.
APN found that the tool started to understand and recognize therapists’ style and language, making the notes get more specific and tailored to each client. This has been a big help for APN since it now takes an average of 11 minutes to complete a note, compared to the industry standard of 15 minutes.
Embarking on AI implementation
With benefits like APN has experienced, I’ve seen a shared, cautious excitement across our industry that continues to suffer from provider and staffing burnout and attrition. By removing some of these administrative burdens, they hope to combat staffing issues and improve patient reach and care.
And while there is good reason to remain cautious, I think providers can confidently move towards AI solutions by focusing on three key areas.
Generative AI, until recently an uncharted frontier, is now encountering regulatory roadblocks. Fueled by minimal oversight, its meteoric rise is slowing as frameworks take shape. Businesses and users alike brace for the ripple effect, wondering how increased scrutiny will reshape this booming sector.
While AI automation could revolutionize efficiency and speed up processes in many customer-facing industries, healthcare demands a different approach. Here, “consumers” are patients, and their data is deeply personal: their health information. In this highly sensitive and regulated field, caution takes center stage.
The healthcare industry’s embrace of AI is inevitable, but the optimal areas for its impact are still being mapped out. As new regulations aim to curb this disruptive technology, a crucial balance must be struck: fostering smarter, more efficient AI tools while ensuring compliance and trust.
The Need for Regulation
Regulatory mechanisms and compliance procedures will play a crucial role in minimizing risk and optimizing AI applicability in the coming decade.
These regulations must be developed to effectively safeguard sensitive patient data and prevent unauthorized access, breaches, and misuse—necessary steps in gaining patient trust in these tools. Imagine the added friction of AI systems that misdiagnose patients, spew incorrect information, or suffer from regular data leaks. The legal and financial implications would be dire.
Optimized workflows simply cannot come at the cost of unaddressed risks. Regulated and responsible AI is the only way forward. And in order to achieve both, three foundational pillars must be met: explainability, control, and compliance.
Imagine this: physicians spend more time with their patients than with their paperwork. Billing is quick and accurate, with minimal denials. Healthcare workers enjoy a positive work/life balance. Thanks to the rapid advancement of AI, this vision of healthcare is becoming increasingly possible.
Health system leaders are already investing toward this ideal state. From roundtable discussions at Healthcare Financial Management Association and Becker’s Healthcare to Zoom chats every week, I’ve connected with many C-suite executives at health systems about their expectations for AI. There is, across the board, a clear set of priorities for the next one to two years. The overarching vision is not just to integrate new technologies but to do so in a way that delivers tangible improvements in workforce experiences and satisfaction, revenues and costs, and patient care outcomes.
Here are a few resounding themes that I’ve heard.
Proving ROI
Proving ROI on AI investments is crucial: put plainly, you want to ensure you’re getting more than enough bang for your buck. Applications of AI need to map back clearly to measurable cost and revenue impacts. Health system CFOs expect predictable ROI and are screening new technologies closely.
Many AI tools on the administrative side can meet this proof of hard ROI. For example, organizations like ApolloMD have experienced significant improvements in coding efficiency and revenue capture by minimizing coding errors and denials through autonomous coding.
While vendors typically report impressive ROI from their technology, any vendor worth its salt will agree to a proof of concept allowing you to test and validate impact for your organization. For example, an easy way to build confidence in autonomous coding is to compare coding results between your team and the AI system before committing to go-live.
Increasing end-to-end strategies
Many AI tools have surfaced to address a single use case. However, health system leaders are more interested in comprehensive, integrated solutions across departments. Consider the case of ambient documentation and autonomous coding: ambient documentation works as a medical scribe using AI to document clinician-patient encounters, and then autonomous coding steps in as a medical coder to translate and assign the necessary codes for billing.
These types of end-to-end strategies are more compelling and impactful. Health system CEOs increasingly gravitate toward them to ease administrative burdens, speed up visit-related processes, and enhance patient outcomes. The market is supporting this expectation: Abridge, an ambient documentation platform, recently raised $150 million in funding, and Google Cloud added an autonomous medical coding solution to its marketplace earlier this year. Used in conjunction, these technologies offer more integrated – and more valuable – strategies for health systems.
Moving from inventing the first wheel ever to discovering the use of Artificial intelligence, we have come a long way. The world is changing for the better, and technological advancement has impacted numerous industries. And the healthcare industry is no exception.
The pandemic has highlighted several gaps in the accessibility of services to patients. Healthcare facilities have had to question old operating methods and adapt to better solutions for providing patients with better care. Moving into 2022, we can observe certain advancements in this sector. These are predictive of what improvements are likely to occur in the future. Listed below are some health-tech trends that are likely to impact the quality of care patients receive profoundly.
Better predictive analytics
The role of data is becoming prominent in improving healthcare services. Data helps identify trends in population health, thereby also helping to identify people at higher risk of developing specific medical issues. Such analysis includes gathering data from hospitals, specialists, primary care providers, and pharmacies. The information will help close gaps in providing patients with proper treatment on time. It will also help healthcare facilities manage a shortfall of resources during emergencies such as a pandemic.
Predictive analytics are likely to become more accurate and efficient in the future with more innovative data collection tools. It will help improve healthcare systems engineering, leading to better management and delivery of high-quality patient care.
Telehealth will become more common
In the past, access to healthcare depended on whether a patient could make it to a hospital or not. However, as communication and collaboration between different geographical locations increases, healthcare services will also expand. Telehealth is not a new idea, but it will gain popularity in the coming years. Doctors and nurse practitioners will be able to counsel patients over apps such as Zoom and other dedicated health portals.
Moreover, at-home testing kits will become more accessible, enabling patients to maintain privacy. According to the American Hospital Association, most healthcare services will be delivered at home or virtually by 2040. It will make healthcare much more accessible to people, especially those who live in remote areas.
Artificial intelligence chatbots have emerged as the most significant health technology hazard for 2026, according to a new report from ECRI, an independent, nonpartisan patient safety organization.
Moving from inventing the first wheel ever to discovering the use of Artificial intelligence, we have come a long way. The world is changing for the better, and technological advancement has impacted numerous industries. And the healthcare industry is no exception.