G&H In what ways is artificial intelligence transforming the care of patients with disorders of gut-brain interaction?
BS In this era of data analytics and artificial intelligence (AI), there is a lot of focus on collection of data elements, like imaging results, genomics, and advanced laboratory testing, to drive precision medicine. But for most clinicians, especially those who care for patients with disorders of gut-brain interaction (DGBIs), the most important data by far are patients’ words—the words elicited at the very first point when they come to a health system to tell their biopsychosocial story. If doctors fail to elicit the right words in the right order, interpret those words incorrectly, or misinterpret what a patient is saying, then they will not know what the next step should be, what the right diagnosis is, or what physical examination, laboratory tests, or imaging studies to perform. Precision medicine is only realized at the end of a long sequence of correct decisions. If the very first step is improperly conducted, then all the AI in the world will not make a difference.
At Cedars-Sinai, we are using AI to collect better histories from patients with irritable bowel syndrome (IBS) and those with other forms of gastrointestinal (GI) disease. This AI system listens to patients talk about their symptoms ahead of time in a HIPAA-compliant manner, collects a detailed comprehensive history, and summarizes the results into a note that it adds to the medical chart; it can even identify an initial assessment and plan because it is able to review the entire chart. The doctor ultimately is the one who looks a patient in the eyes and in their own words brings the wisdom of the human healer with the computational ability of the AI system in a process I call BI, or blended intelligence. This is where the incredible wisdom of humans, especially experienced clinicians, is combined with the pattern-seeking abilities of computational systems, and neither tries to be the other. The two together are stronger and better at connecting with people, hearing their stories, interpreting them correctly, and then moving forward with the right approach. We are also testing smart glasses that project information from the chart into the wearer’s line of sight, enabling the clinician to maintain eye contact with the patient. The glasses display instantaneous heads-up information, including possible questions to ask, alarm features, and a differential diagnosis (Figure). This combines human and machine to enhance the process and outcomes of care.
G&H How is the application of AI in microbiome analysis helpful in IBS and other neurogastroenterological disorders?
BS I think this is to be determined. This raises an opportunity to consider the deeper question of, what is intelligence, and how would our view of it relate to microbiome analysis. Intelligence, as I see it, is the ability to find patterns within chaos—patterns that are meaningful, predictive, and if identified and acted upon can improve the lives of human beings. If intelligence is the ability to survive and thrive on this planet by identifying emergent patterns that are meaningful and predictive, then thinking about the trillions of microorganisms living inside the human body, there are undoubtedly meaningful patterns inherent in the microbiome. Being able to identify and act upon these patterns could put us in a much better position to diagnose disease and treat it appropriately with diet, mind-body, and medical interventions. Identifying these patterns using machine learning and other forms of AI to analyze large samples of output—for example, by finding meaningful patterns in stool samples—could help us predict in a meaningful way whether we can change somebody’s life for the better. That is what we are trying to accomplish and have not achieved. This is my philosophical thought about how to get there, but we are not there yet with the microbiome.
G&H What has research shown so far on the use of AI models to predict IBS development?
BS There is some interesting research particularly on computer vision systems. Research suggests that IBS might be diagnosable using a colonoscopy AI model with a high reported degree of accuracy. This is mind-blowing because GI doctors are taught that the colon is normal in people with IBS. In fact, this is why colonoscopy is not recommended in younger people with IBS because it often does not reveal anything (no higher rate of polyps or cancer, no inflammatory bowel disease) or confirm whether a patient meets the Rome IV criteria. The features of the colonic surface that the AI system uses to distinguish IBS vs controls are unclear, but it is likely finding meaningful patterns that are predictive. This research raises the question of whether there is diagnostic bias—we find only what we know to look for—or whether these changes in the colon are not detectable by a human. Multiple features—for example, the branching patterns of the capillaries, their geometry and ramifications, the subtle differences in the color of the colon surface—could potentially be important in IBS, but we rarely talk about those sorts of patterns. If they were important, then they could potentially teach us a lot about IBS pathophysiology. Who knows, there may come a point when IBS can be diagnosed in a clinic during a rectal examination with a tiny camera on a fingertip using a computer vision system. What may seem like an impossibility—in this era of exponential growth of technology where there is an exponent on the exponent—could become a reality. All of us have to be hypervigilant about the pace of change. Obviously, more evidence and reproducibility are needed, and clinicians need to understand, as always, the science behind these discoveries. There are no shortcuts in science. But the key will be to remain tremendously open-minded about the creative ways these technologies can be leveraged to make diagnoses and improve human health in ways never before imagined.
G&H How is AI being used to support diagnosis of functional dyspepsia and IBS?
BS I personally have not done research on this topic, but there are many ways AI could be used to investigate dyspepsia. Similar to the previous example, an AI model could evaluate surface features of the stomach or duodenum or determine whether features of biopsy samples and histopathology can reproducibly distinguish people with functional dyspepsia vs controls. These are possible patterns that perhaps no one would have considered looking at, but an unbiased machine self-learning system might be able to find subtle clues in the histopathology or macroscopic appearance of stomach and duodenum. This AI model could be applied to every part of the GI tract; whether looking at the esophagus, stomach, small bowel, or colon, there are undoubtedly patterns that computer vision systems could potentially identify and be used to support diagnosis of many conditions, not just dyspepsia.
G&H What are the key roles of AI systems in GI symptom management?
BS As I mentioned, if we think about symptoms as words the patient says and these patient-reported outcomes as data, then there is a tremendous opportunity in gastroenterology and beyond for AI systems to help us interpret patients’ words. At Cedars-Sinai, we use AI not just as a scribe, but as what I call a superscribe. This superscribe is not only listening to the conversation but also has full access to the patient’s medical chart. As it is listening to the conversation, it can pull data from the chart and has a heads-up display where it can help us translate the symptoms instantaneously into a differential diagnosis in real time in the clinic. It also can make a list of recommended questions that the clinician should ask the patient that sometimes might be missed. This is especially important for trainees who are learning how to take a history, and this is a lost art now. The purpose is not to de-skill clinicians to the point where we let the AI do all the work, but to learn from the AI system and understand how it came to recommend these specific questions about a patient’s symptoms. The AI scribe works on a tablet and phone, or any kind of computerized device, and provides results in real time, including how it is interpreting the symptoms and what it is recommending for management based upon a full review of the chart, not just from the patient-clinician conversation. A number of research studies are being conducted on this system. This superscribe is being used not just in gastroenterology but in many other fields across the Cedars-Sinai health system to amplify this BI approach of combining the experience of human doctors with the computational abilities of AI.
G&H How effective are AI-driven interventions (dietary and lifestyle) compared with traditional approaches for patients with DGBIs?
BS More data are needed to examine how AI-enabled management of DGBIs compares with traditional approaches. There is reason to believe that AI-enabled care will do better in many different ways. I previously mentioned taking a better history, reviewing the chart, and finding patterns in the chart that humans might miss when they are very busy and do not have access to every bit of data in the forefront of their mind, whereas AI does. The National Institutes of Health (NIH) should ideally fund this kind of research in the GI space. We can rigorously evaluate whether, when, and how to use AI in the care of DGBI patients, or GI patients in general. My colleagues and I recently completed our first randomized controlled clinical trial on an 8-week AI-enabled virtual reality (VR) therapy for IBS patients we are developing called SynerGI, which is being funded by the NIH. The results, which are quite positive, are being presented at Digestive Disease Week 2026. The trial compares use of immersive VR with sham VR in patients with IBS to determine improvements in quality of life. The VR program provides a form of cognitive behavioral therapy that works on the gut-brain axis to help manage IBS symptoms and stress. Another project we are developing, which we are very excited about, is an IBS simulator using VR supported by various forms of AI. This utilizes a VR headset and an abdominal belt that contracts and simulates safely the sensations of abdominal pressure and cramping while in the body of a person with IBS going through their day. Simulated activities include driving in a car and feeling cramping and needing to pull over, feeling GI symptoms during a business meeting and having difficulty concentrating, or being out to dinner and struggling with bloating and discomfort while everyone else is eating and having a good time. This VR AI simulator should be ready in the fall, and we plan to test it with doctors, especially young doctors, to see if we can improve their empathy for patients with IBS. There is often a lack of recognition, if not a stigma, about the biopsychosocial illness experience that people with IBS have. If just 10 minutes in a simulator can alter one’s negative perception of this illness, then that is an important, and perhaps unexpected, use of technologies and AI in training and educating clinicians.
G&H Which AI tools have the potential to optimize workflow, and how might they be useful for managing patients with DGBIs?
BS Certainly, ambient scribes are important AI tools that are reducing provider burnout and burden in general by allowing doctors to spend more quality time with patients, have human-to-human conversations, and take good histories, rather than always focusing on documentation. These tools are an opportunity to address the constant barrage of bureaucratic administrative layers of documentation that have become deeper and thicker over the years and that separate us in time and space from our patients. The bureaucracy that surrounds the practice of medicine in America today is extraordinarily frustrating and challenging for many doctors; this administrative documentation, while fundamentally necessary, has become too much for many of us to bear. For doctors who often have to document at a level 5 complexity—such as GI doctors who are almost always operating at this complexity level because of the deep biopsychosocial decision-making that is needed with every DGBI patient—having the ability to capture that rich discussion between patient and clinician with ambient scribes is a major advance. However, that is not enough. As I said, having a superscribe, which we are using at Cedars-Sinai and that not only scribes but has a heads-up display and provides an additional layer of interpretation with access to the chart, is the type of workflow integration that I think will be increasingly adopted.
G&H What are the key challenges to integrating AI into daily practice and how can they be addressed?
BS One of the challenges is accuracy. It is important that these systems are correctly integrating data from the chart and finding patterns, and that they are not hallucinating. In the AI space, hallucinations occur when a large language model makes something up and presents it as if it is absolutely accurate, when in fact it is not. This is a problem; however, with new architectures, researchers are seeing fewer hallucinations. A major barrier for doctors is how to trust the AI system, especially if it is making decisions for us. With the technology improving as quickly as it is, I would warn against making premature conclusions about AI and have faith in the programmers who are very adept at correcting errors and increasing accuracy. But we still must stay wary, and it is ultimately our licenses on the line. We cannot cede our responsibilities to a computer. Another challenge is the need for highly supportive leadership. It is important to have open-minded leaders that are prepared to move forward because the health systems that optimize AI will rapidly accelerate beyond those that do not. Doctors do not need to adopt every AI tool, but they should not be machine-breaking Luddites either. The progress of technology is essentially unstoppable. Doctors should be able to work at the top of their licenses to deliver the best possible care using these technologies, and open-minded leadership is a major part of that.
G&H What are some of the priorities for future research?
BS More rigorous randomized controlled trials are needed across different DGBIs to understand how best to take histories, how best to use AI systems to diagnose disease, how to find patterns across datasets that are predictive of disease, and what predictive patterns are in the microbiome and in stool samples. Studies should evaluate use of AI and machine learning systems with histopathologic examinations and with imaging studies other than colonoscopy to see if there are unique patterns that humans have not yet identified and that are highly predictive of DGBIs. We need to continue to demystify the underlying causes of DGBIs. I suspect there are many patterns, and possibly quite physical patterns, even in the construction of the human body, in the suspension systems within the abdominal cavity, in structure and function of the brain, and in everything from the appearance of the mesentery to the appearance of crypts and villi. These are just a few of the many ways that we can examine DGBIs with AI systems.
Disclosures
Dr Spiegel has received research grants from Ardelyx, Guardant, Freenome, Ironwood, and Salix; he has also served on advisory panels for Ardelyx and Guardant.
Suggested Reading
Ajmera K, Patel O, Shah N. Artificial intelligence in gastroenterology: beyond diagnostics and toward lifestyle and dietary interventions for gastrointestinal disorders. Cureus. 2026;18(1):e100976.
Gross SA, Shaukat A, Afzali A, et al. Artificial intelligence for gastroenterology practice: a Modified Delphi Consensus. Am J Gastroenterol. 2026;121(4):1017-1024.
Suchak KK, Almario CV, Liran O, Chernoff R, Spiegel BR. The role of virtual reality in the management of irritable bowel syndrome. Curr Gastroenterol Rep. 2024;26(11):294-303.
Tabata K, Mihara H, Nanjo S, et al. Artificial intelligence model for analyzing colonic endoscopy images to detect changes associated with irritable bowel syndrome. PLOS Digit Health. 2023;2(2):e0000058.