AI and Surgical Robots in 2026: How Artificial Intelligence Is Changing Surgery

The Robot Is in the Operating Room.
How Much of the Surgery Can AI Really Do?

What surgical robots can already do, where artificial intelligence actually helps, and how close medicine is to autonomous operations

AI and Surgical Robots in 2026: How Artificial Intelligence Is Changing Surgery

Walk into a modern robotic operating room and the scene already looks like science fiction: several mechanical arms hover over the patient while the surgeon sits at a console a few meters away, working through tiny incisions. The robot can filter tremor, scale hand movements and maneuver instruments in spaces that are awkward for the human wrist. None of that requires the robot to think. The real change now is that artificial intelligence is beginning to enter the same room.

AI adds something robotics alone does not have: interpretation. Computer vision can try to recognize anatomy in the live video feed; predictive models can estimate risk; software can analyze technique; experimental systems can even carry out selected surgical steps. Surgery is therefore moving from machines that simply reproduce a surgeon’s movements toward systems that can also perceive, advise and, in limited settings, act.

That is where the hype needs to stop and the definitions need to become precise. Most surgical robots in hospitals today do not decide where to cut, do not choose a surgical strategy and do not continue operating if the surgeon steps away. The U.S. Food and Drug Administration describes common robotically assisted surgical systems as devices that require direct human control. In other words, robotic surgery and autonomous AI surgery are not the same technology.

Robotic surgery is already a clinical reality. Artificial intelligence is now adding a new layer of guidance, perception and decision support.

First, a crucial distinction: a surgical robot is not necessarily an AI

The word “robot” makes many people imagine an autonomous machine. In the operating room, that is usually misleading. A typical soft-tissue robotic system is better understood as an extremely sophisticated extension of the surgeon’s hands. The surgeon sits at a console, watches a magnified three-dimensional image and moves controllers. The system translates those hand movements into smaller, steadier movements of instruments inside the patient.

Several major advantages come from robotics alone, without any artificial intelligence: motion scaling, tremor reduction, wristed instruments that can move in tight spaces, stable 3D visualization, and better ergonomics for the surgeon. Newer platforms are also adding richer sensory feedback. For example, da Vinci 5 force-feedback instruments cleared in the United States can transmit push-and-pull forces at the instrument tip back to the surgeon’s hands.

ROBOTICS ≠ AI If a surgeon moves a controller and a robotic arm copies the movement, that is robotic assistance. If software analyzes the surgical video, identifies a blood vessel, predicts the next surgical phase or autonomously performs a defined task, that is where AI and automation begin to enter the picture.

A surgical robot extends the surgeon’s hands. AI goes further by helping interpret anatomy, identify risks and provide real-time guidance.

So what does AI actually add to surgery?

A useful way to understand surgical AI is to picture a ladder rather than a single technology. At the bottom are tools that analyze data or highlight structures for the surgeon. Higher up are systems that automate a defined task. At the top is the idea most people imagine when they hear “robot surgeon”: a machine that can complete an entire soft-tissue operation without continuous human control. That last step remains experimental.

1. Computer vision: giving the robot “eyes that understand”

A normal surgical camera produces video. A computer-vision system tries to interpret that video. It can be trained to recognize instruments, organs, blood vessels, nerves, tumors, tissue planes, bleeding, or the current phase of an operation. In simple terms, the camera stops being only a camera and begins to behave more like a navigation system.

Imagine a laparoscopic gallbladder operation. The surgeon sees a complex field of soft tissue where the bile duct, cystic duct and nearby vessels must be distinguished correctly. An AI system can be trained to outline structures or warn when the dissection moves toward a dangerous area. This is similar in spirit to lane detection in a car—except the “road” is deformable tissue, the camera moves constantly, blood or smoke can obscure the view, and a mistake can injure a patient.

The research base is large but not yet as clinically mature as headlines sometimes imply. A 2026 systematic review of 188 studies on surgical scene understanding found that 70.7% relied on small, single-center datasets, only 10.1% used external datasets, and very few studies demonstrated meaningful clinical translation. Another 2026 systematic review of 95 studies on AI guidance in robotic surgery similarly concluded that technical validation is advancing faster than proof of patient benefit.

2. Preoperative planning and risk prediction

Before the first incision, AI can combine imaging, laboratory values, medical history and previous surgical data to estimate risks or help plan an approach. In principle, a model might identify a patient with unusually high bleeding risk, predict the probability of conversion to open surgery, or reconstruct a three-dimensional map from CT or MRI scans.

A good prediction model is not automatically a useful clinical tool. The Perioperative ORACLE randomized clinical trial tested machine-learning risk estimates in more than 5,000 patients. Showing those estimates to clinicians did not significantly improve their prediction of 30-day mortality, and the improvement for acute kidney injury did not reach conventional statistical significance. The lesson is simple: an algorithm can be statistically impressive and still fail to change a real decision at the bedside.

3. Surgical skill assessment and AI coaching

A robotic system records far more than video. It can also capture how instruments move: speed, path length, smoothness, economy of motion and timing. Machine-learning systems can use these signals to distinguish novice from expert performance and potentially provide objective feedback.

This may prove more useful than it sounds. Surgical training still depends heavily on an expert watching another surgeon and judging performance. A future resident could finish a simulated procedure and receive a report showing unnecessary instrument movements, excessive tissue force, slow steps and moments of elevated risk. The mentor would still matter, but the feedback would become far more continuous and measurable.

4. Automation: from a single task to an autonomous procedure

This is the part that attracts the most attention. Surgical automation does not arrive in one jump. It develops in levels.

Level	What the robot does	Human role	Reality in 2026
0	No autonomy	Human performs everything	Conventional surgery
1	Robot assistance	Continuous direct control	Most soft-tissue robotic surgery
2	Task autonomy	Human selects a defined automated task	Clinical in selected systems/tasks
3	Conditional autonomy	Robot proposes/executes a patient-specific plan after approval	Clinical in selected rigid-anatomy applications
4	High autonomy	Robot plans and adapts with human supervision	Preclinical/experimental for soft tissue
5	Full autonomy	Robot can complete the whole procedure independently	Not routine clinical soft-tissue surgery

A 2024 review of FDA-cleared surgical robots through 2023 found that 86% were Level 1, 8% Level 2 and 6% Level 3; none were Level 4 or 5. New platforms have appeared since that review, but the broader picture has not flipped: clinical surgical robotics is still dominated by systems in which a human remains in control.

Surgical autonomy is not an on-off switch. It develops gradually—from simple robotic assistance to systems that may eventually perform complex tasks under human supervision.

The milestones that changed the autonomy debate

STAR: autonomous suturing in living tissue

In 2022, researchers at Johns Hopkins reported a major milestone with the Smart Tissue Autonomous Robot, or STAR. The system performed laparoscopic intestinal anastomosis—the task of sewing two ends of intestine together—in pigs. It tracked deformable tissue, generated suture plans and executed much of the procedure autonomously, while a human operator selected or approved plans and monitored the process.

This was important because soft tissue is fundamentally harder than bone. Bone stays where the preoperative scan says it is. An intestine moves, stretches, deforms and shifts with breathing. STAR showed that a robot could repeatedly measure that changing environment and adapt.

SRT-H: a robot that learned a long surgical sequence

The next dramatic step came in 2025. The Surgical Robot Transformer-Hierarchy (SRT-H) used language-conditioned imitation learning—a form of learning from expert demonstrations—to perform a complex, multi-step segment of gallbladder surgery. The system used a high-level planner to decide what task should happen next and a lower-level controller to turn that instruction into instrument movements.

In the Science Robotics experiment, SRT-H completed a 17-task surgical sequence across eight ex vivo porcine gallbladders with a 100% success rate and recovered from some suboptimal states without human intervention. That is a striking proof of concept—but it is still a small preclinical experiment on tissue outside a living patient. The jump from that controlled setting to a bleeding, breathing, anatomically variable human being is enormous. SRT-H shows that high-level autonomy is becoming technically plausible, not that autonomous gallbladder surgery is ready for routine care.

Computer vision can turn a surgical camera into more than a camera—helping identify anatomy, highlight critical structures and suggest safer surgical paths.

Why autonomous surgery is much harder than self-driving a car

The comparison with autonomous vehicles is tempting because both technologies combine cameras, sensors, planning and control. But surgery has several properties that make autonomy unusually difficult.

·         Human anatomy varies. Textbook anatomy is an average, not a guarantee.

·         Soft tissue deforms. Pulling on one structure can move several others.

·         The visual scene can change in seconds because of bleeding, smoke, fluids or tissue motion.

·         The robot may need to infer force even when the camera image looks normal.

·         Rare complications matter disproportionately. A system can perform well in 99 routine cases and still fail catastrophically in the one unusual case.

·         There is no safe equivalent of “pull over to the side of the road” inside the abdomen.

This is why many researchers expect autonomy to arrive first as small, bounded islands of automation: camera positioning, suturing, stapling, drilling, bone preparation, needle placement, identification of anatomy, or execution of a surgeon-approved plan. The entire operation is a much harder target.

Does robotic surgery actually produce better outcomes?

The evidence is less dramatic than the marketing. Robot-assisted surgery can be compared with open surgery or with conventional laparoscopy, and those are very different comparisons. Against open surgery, robotics may help more patients receive a minimally invasive operation, with potential benefits such as less blood loss, pain or time in hospital for selected procedures. Against modern laparoscopy, the advantage is often smaller and highly procedure-dependent.

A 2024 BMJ Open overview examined 165 systematic reviews across multiple specialties. Most outcomes for robotic surgery were positive or neutral, with the evidence generally more favorable when robotics was compared with open surgery. Compared with laparoscopy, many outcomes were similar; robotic procedures also tended to take longer, and the underlying evidence was often low quality.

The key distinction is between technical precision and better patient outcomes. A robot can move an instrument more steadily than a human hand, yet the final result also depends on which patients are selected, the surgeon’s experience, anesthesia, infection prevention, postoperative care, tumor biology and many other factors. A steadier instrument is useful; it is not, by itself, proof of a better operation.

The same caution applies even more strongly to AI. Much of the current literature demonstrates that an algorithm can identify anatomy or predict an event—not that adding the algorithm reduces mortality or complications. The 2026 reviews of surgical computer vision repeatedly call for prospective, multi-center studies that measure real patient outcomes rather than only algorithmic accuracy.

What surgical AI looks like in a real operating room in 2026

The most accurate description of surgical AI in 2026 is not “robots are becoming surgeons.” It is that the operating room is becoming measurable in a way it never was before. Cameras, robotic instruments and software can record what is happening, quantify it and increasingly interpret it in real time.

·         Before surgery: AI can analyze imaging, help segment anatomy, estimate risk and support planning.

·         During surgery: computer vision can identify instruments and anatomy, track procedural phases and potentially warn about dangerous regions.

·         At the console: force sensing, motion analytics and improved imaging can give the surgeon more information than vision alone.

·         After surgery: video and kinematic data can be used for training, quality review and research.

·         Across many cases: aggregated data may reveal which techniques, movements or decisions correlate with better outcomes.

The hardware is changing too. In 2026, the FDA authorized Johnson & Johnson’s table-integrated OTTAVA robotic system for several upper-abdominal general-surgery procedures. Intuitive also expanded its da Vinci 5 platform, including force-feedback instruments. These systems show how quickly robotic hardware is advancing, but neither should be mistaken for an autonomous surgeon: a human remains responsible for the operation.

Remote surgery: distance can disappear before the surgeon does

Telesurgery is another concept that is often mixed with autonomous surgery. They are very different. In telesurgery, the surgeon may be hundreds or thousands of kilometers away, but a human is still operating the robot.

Recent demonstrations show that long-distance robotic surgery is technically feasible. A published series described five human procedures performed with the remote surgeon in Shanghai and patients in Kuwait, roughly 7,000 kilometers away. Average round-trip network latency was about 167 milliseconds, and an experienced robotic surgeon remained physically present with the patient as a safety backup.

That local backup is not a footnote—it is part of the safety model. Remote surgery still depends on reliable networks, cybersecurity, emergency conversion plans and people at the patient’s side who can respond if the connection or robot fails. Even so, the concept is powerful: expertise could travel farther than the surgeon’s body.

Where AI could genuinely make surgery safer

The biggest safety gains may come from helping the surgeon notice more, not from removing the surgeon. An operation demands constant attention to anatomy, instruments, tissue tension, bleeding, vital signs and the sequence of the procedure. AI is unusually well suited to watching several of those streams at once.

·         Anatomy recognition: highlight structures that must be protected.

·         Bleeding detection: notice subtle changes in the image or vital signs before they become obvious.

·         Context awareness: recognize what phase of the operation is underway and prepare the relevant information.

·         Force awareness: combine visual and force data to identify excessive tissue stress.

·         Decision support: compare the current situation with large numbers of prior cases.

·         Quality control: identify skipped steps, deviations or unusual instrument trajectories.

·         Training: provide objective feedback on technical performance.

A useful analogy is aviation. Modern aircraft did not become safer because computers made pilots irrelevant; they became safer partly because automated systems continuously monitor engines, navigation, altitude, weather and flight controls. Surgery may develop in a similar direction: the surgeon keeps authority while software watches the operation in parallel and calls attention to what deserves a second look.

This broader “AI copilot” model is also changing clinical work outside the operating room; see our related guide AI Assistants for Doctors: Can Algorithms Replace a Physician?.

The risks are not theoretical

Automation bias

If an AI overlay says a structure is safe, will a tired surgeon trust it too quickly? Humans tend to over-trust automated systems once those systems are usually correct. In surgery, the dangerous error may be the rare confident mistake.

Dataset bias and unfamiliar anatomy

Many surgical AI models are trained on relatively small datasets from one hospital, one camera system or one procedure. That may not generalize to different patient populations, different equipment or unusual anatomy. The 2026 surgical-scene review found external validation in only about one in ten studies.

Cybersecurity and system failure

A conventional surgical instrument can break. A connected robotic system can also fail because of software, sensors, networks or cybersecurity. The more autonomy a system has, the more important fail-safe behavior becomes: how does it stop, hand control back, or reach a safe state?

Responsibility

If a surgeon follows an AI recommendation and a patient is harmed, who is responsible—the surgeon, hospital, software developer, robot manufacturer, or all of them? At low levels of autonomy the answer remains relatively clear because the surgeon controls the device. Higher autonomy makes liability and informed consent more complicated.

Cost and unequal access

Robotic platforms are expensive to buy, maintain and staff. The economic case is strongest when a high-volume center uses the platform efficiently and when robotics enables a minimally invasive operation that might otherwise require open surgery. If AI and robotics become the standard of care only in wealthy systems, the technology could widen rather than reduce disparities.

The most likely future is supervised autonomy: AI performs carefully defined tasks while the surgeon sets the goals, monitors progress and can take control at any moment.

What happens next?

The near term: AI becomes a second pair of eyes

In the next few years, the changes most likely to reach everyday practice are not fully autonomous operations. They are better anatomy recognition, automated phase documentation, skill analytics, smarter image guidance, force-aware warnings and AI tools that surface relevant information before a difficult step. These features are easier to test because they support a surgeon rather than replace one.

The early 2030s: more tasks, not necessarily whole operations

By the early 2030s, the interesting shift may be from assistance to delegation. A robot could be asked to suture a defined segment, position a camera, prepare bone, place a clip, maintain traction or carry out a surgeon-approved portion of a dissection. The surgeon would still set the strategy, but might no longer need to control every millimeter of motion.

The longer horizon: the surgeon becomes a supervisor of surgical intelligence

A fully autonomous soft-tissue operation is a different problem. The robot would need to recognize when a case has stopped being routine, estimate its own uncertainty, respond to bleeding or unexpected anatomy and know when to stop rather than improvise. In medicine, knowing when not to continue may be as important as knowing what movement comes next.

That makes a hospital without surgeons a poor near-term forecast. A more plausible future is one in which surgeons supervise increasingly capable systems, step in for difficult judgments and complications, and remain responsible for the overall strategy while standardized technical tasks are delegated to machines.

Could AI replace the surgeon?

Not in the sense implied by the headline—and not soon. Surgery is more than cutting and suturing. Someone must decide whether an operation is appropriate, choose among competing approaches, interpret incomplete information, adapt when anatomy is unexpected, manage complications, speak with the patient and family, and accept responsibility for the outcome.

AI will probably automate those parts unevenly. Repetitive technical actions are comparatively tractable. Visual recognition and prediction are improving fast. Ambiguous judgment, responsibility and recovery from rare complications are much harder.

The more useful question is therefore not “Will AI replace surgeons?” but “How much of an operation will still require a surgeon to control every action directly?” In 2026, the answer is: most of it. But the boundary is beginning to move.

For another view of how AI changes medical decision-making rather than surgical mechanics, see Artificial Intelligence in Radiology: More Accurate Than Doctors?.

Conclusion: the smartest surgical robot may still be a partnership

Robotic surgery is no longer science fiction. AI-assisted surgery is real too—but autonomous surgery is still in the uncomfortable middle between laboratory achievement and routine clinical care. The technology can already recognize, measure, predict and automate selected tasks. It cannot yet reproduce the full adaptability of an experienced surgical team.

For now, the most defensible model is partnership. Robots contribute dexterity, stability, visualization and increasingly force sensing. AI can add perception, prediction, planning and objective feedback. Humans still carry the strategy, judgment, adaptability and responsibility.

If autonomous surgery eventually becomes routine, there may be no single day when a robot suddenly “replaces” a surgeon. The change is more likely to happen quietly: first a camera movement, then a suture, then a defined dissection step, then a complete standardized phase. At some point, the meaning of “the surgeon performed the operation” may itself need a new definition.

FAQ

Is robotic surgery the same as AI surgery?

No. Most robotic surgery today is directly controlled by a surgeon. AI can be added for image analysis, guidance, prediction, training or automation, but the robot itself does not automatically imply artificial intelligence.

Are there autonomous surgical robots in hospitals today?

There are clinical systems with limited task or conditional autonomy in selected fields, especially where anatomy is rigid and predictable. Fully autonomous soft-tissue surgery is not routine clinical practice. Current FDA guidance for common robot-assisted surgical systems still assumes direct human control.

Is robotic surgery safer than laparoscopic surgery?

It depends on the operation. Compared with open surgery, robotics can help expand access to minimally invasive procedures. Compared with conventional laparoscopy, many outcomes are similar, although robotic surgery may reduce conversion to open surgery or offer technical advantages in selected complex cases. It also often costs more and takes longer.

What was the SRT-H breakthrough?

In 2025, SRT-H used language-conditioned imitation learning to autonomously perform a 17-task segment of gallbladder surgery on eight ex vivo porcine specimens. The result was an important preclinical milestone, not a human clinical trial.

Can AI replace surgeons?

AI is more likely to automate specific surgical tasks before it replaces the full role of a surgeon. Technical actions are only one part of surgery; judgment, responsibility, communication and management of unexpected complications remain major human functions.

Can surgeons operate remotely?

Yes, experimental and early clinical telesurgery has demonstrated operations over thousands of kilometers. But remote surgery still requires reliable networks, cybersecurity, emergency plans and often an experienced local surgical team.