What Will Veterinary Medicine Look Like in 2035?

By Steve Mehler, DVM, DACVS

The future of veterinary medicine will not be defined by a single breakthrough. It will emerge when artificial intelligence, wearable sensors, continuous monitoring, precision medicine, targeted therapies, and enormous amounts of patient data begin working together.

What will veterinary medicine look like in 2035?

I do not think the biggest change will be a new drug, a new surgical procedure, or even artificial intelligence by itself. I think the real transformation will happen when several technologies that already exist begin working together.

Artificial intelligence is already analyzing radiographs. Dogs are already wearing activity monitors. We are sequencing tumors, monitoring patients remotely, and using monoclonal antibodies to treat disease. Individually, none of those developments is entirely revolutionary anymore.

But imagine what happens when all of those technologies start talking to one another.

That convergence could fundamentally change not only how we diagnose disease, but what it means to be a veterinarian. The veterinarian of 2035 may spend much less time asking, “What disease does this animal have?” and much more time asking, “What should I do with everything technology just told me?”

Technology will not make veterinarians irrelevant. It will change which parts of being a veterinarian are most valuable.

AI Will Move Beyond Reading X-Rays

When veterinarians hear “AI,” most of us currently think about isolated diagnostic tasks. An algorithm reads a radiograph, evaluates a cytology slide, analyzes an ECG, or examines a gait pattern.

Those applications are useful, but they may be the least interesting version of veterinary AI.

Consider a 10-year-old Labrador named Max who has visited the same hospital his entire life. Instead of asking an algorithm to interpret one set of radiographs, imagine giving it access to Max’s complete clinical history: 10 years of medical records, five years of laboratory results, every radiograph he has had, his medications, body weight, body-condition scores, activity data, resting heart rate, sleep patterns, and video showing how he moves at home.

The question is no longer simply, “Is there arthritis on this radiograph?” It becomes, “What is happening to Max?”

Perhaps his activity has fallen 12% over six months. His nighttime movement has increased, his walking speed has gradually decreased, and weight distribution between his pelvic limbs has changed. His resting heart rate has slowly increased. Several laboratory values are still within their reference intervals, but they are drifting away from his individual baseline.

A veterinarian looking at each data point separately may not recognize the pattern. An algorithm analyzing thousands of variables over time might.

That is where AI becomes genuinely interesting: not as a replacement for the radiologist or pathologist, but as a way to identify relationships within more information than a human can process simultaneously.

We May Diagnose Disease Before Animals Look Sick

Veterinary medicine is still largely reactive. Something changes, the caregiver notices, an appointment is made, diagnostics are performed, and eventually the disease is identified.

The weakness in that system is obvious: someone has to notice a problem first. Animals can be remarkably good at hiding disease, and gradual changes are especially easy to miss.

Osteoarthritis is a perfect example. Owners frequently tell me, “My dog isn’t painful,” but what they often mean is, “My dog isn’t crying.” Meanwhile, the dog stopped jumping into the car eight months ago, hesitates on the stairs, sleeps more, walks more slowly, and no longer follows the family upstairs.

Those changes are meaningful data. Historically, we simply have not had a practical way to measure them continuously.

That is changing. Wearables can monitor activity and sleep. Pressure-sensitive systems can objectively measure gait. Computer vision may allow movement analysis using ordinary video. Smart collars and other connected devices are beginning to collect physiologic information such as heart and respiratory rates.

When those measurements are combined longitudinally, an algorithm may detect a progressive gait abnormality months before Max becomes obviously lame.

The same principle could apply to a cat whose food intake, activity, body weight, water consumption, and litter-box behavior are all changing slightly. No single change necessarily signals disease, but together they may form a recognizable pattern.

Veterinary medicine could move from saying, “Your pet has disease,” to saying, “Your pet appears to be developing disease.” That is a profound shift.

The Physical Examination May Extend Into the Home

A veterinary examination lasts perhaps 10 or 20 minutes. There are more than half a million minutes in a year. For almost all of them, the veterinarian is blind.

That is why we depend so heavily on caregiver observations. We ask how an animal is walking, eating, drinking, sleeping, and behaving. Owners do their best, but human observation is subjective. Most of us cannot accurately remember how quickly a dog walked six months ago or detect that a cat is drinking 15% more water.

Technology could create what is functionally a continuous physical examination.

A wearable detects lower activity. A camera identifies a shorter stride. A smart feeder records decreased intake. A connected scale tracks weight. A litter box recognizes changes in elimination patterns. Software integrates the signals into the medical record.

The examination is no longer confined to the hospital. It continues between visits.

The reason for an appointment may change as well. Instead of an owner calling to say, “Something seems wrong,” the veterinary team may contact the owner and say, “We are seeing a pattern in your dog’s data that we think should be investigated.”

Objective Measurement Could Transform Pain Medicine

Pain is extraordinarily difficult to measure in animals. Our patients cannot tell us that a knee hurts 30% less today, so we rely on proxies: owner questionnaires, veterinary examinations, force plates, activity monitors, and gait analysis.

Every one of those tools has limitations. Questionnaires are subjective. Examinations are snapshots. Force plates are objective but usually confined to controlled environments. Activity monitors tell us how much an animal moves, but not necessarily why.

Now imagine combining several measurements continuously. Instead of asking only whether a dog “seems better,” we might determine that walking speed increased 14%, daily activity increased 18%, nighttime restlessness decreased, weight-bearing symmetry improved, and the dog is climbing the stairs more often.

We could ask better questions about treatment. Did the NSAID help? Did the monoclonal antibody, supplement, rehabilitation program, or surgery help? More importantly, how much did it help?

That would also affect veterinary research. Anyone who studies pain knows about the caregiver placebo effect: belief that a treatment should work can influence an owner’s perception of improvement. This is not dishonesty; it is human psychology. Objective longitudinal measurements could add another layer of evidence by showing whether the animal is actually functioning differently.

Precision Medicine Is Coming to Veterinary Medicine

Veterinary treatment is still usually selected primarily by diagnosis. A dog has lymphoma, a mast cell tumor, or osteoarthritis; therefore, we choose from the therapies commonly used for that condition.

Biologically, however, two animals with the same diagnosis may have very different diseases. Oncology has already made this clear. Tumors that look similar under a microscope can have different molecular characteristics, biological behavior, metastatic potential, and treatment responses.

Over the next decade, I think we will increasingly stop asking, “What drug treats this disease?” and start asking, “What treatment is most likely to work in this patient?”

That is precision medicine. It could incorporate molecular diagnostics, genomics, tumor profiling, biomarkers, pharmacogenomics, and information about an individual animal’s metabolism or immune system.

Oncology will probably lead this transition, but the concept could extend far beyond cancer. Imagine predicting which dog is most likely to respond to a particular osteoarthritis therapy, which patient has an elevated risk of an adverse drug reaction, or which animal is predisposed to a disease years before clinical signs appear.

We do not have all of those answers today, but that is the direction medicine is moving.

The Therapies Themselves May Become Smarter

As we become better at characterizing patients and diseases, therapies are becoming more biologically targeted.

Traditional pharmacology often focuses on blocking an enzyme, receptor, or inflammatory mediator. Newer approaches can be far more specific. Monoclonal antibodies are already increasingly important in veterinary medicine. Cancer immunotherapy is evolving. Gene therapies are being investigated, and regenerative medicine continues to develop.

One particularly fascinating technology is targeted protein degradation, including proteolysis-targeting chimeras, or PROTACs. Instead of merely binding to a problematic protein and blocking its activity, a PROTAC recruits the cell’s own protein-disposal machinery to eliminate it.

In simplified terms: do not inhibit the protein—remove it.

Whether PROTACs themselves become important veterinary therapeutics remains uncertain. What matters is the broader direction they represent: more specific targets, more sophisticated mechanisms, and more individualized therapy.

Now combine that with precision diagnostics, better characterization of each patient, and algorithms that may help estimate which treatment has the greatest probability of success. That is very different from one-size-fits-most medicine.

Innovation and Evidence Are Not the Same Thing

Everything described so far sounds extraordinary, but veterinary medicine has always been attracted to new technology: a new implant, drug, diagnostic, supplement, device, surgical technique—and now, a new algorithm.

We must remember that innovation and evidence are not the same thing. Something can be technologically impressive and clinically useless. An AI system can produce a confident answer without producing a correct one.

Some veterinary AI studies generate remarkable headlines, but a closer look may reveal a small dataset, internal validation, no independent external testing, highly selected patients, specialized equipment, or endpoints that do not reflect real clinical use.

That does not mean the technology is bad. It means we need to ask better questions:

    ●    Was the algorithm independently validated?

    ●    Was it tested on animals resembling those seen in everyday practice?

    ●    What happens when images are imperfectly positioned?

    ●    What happens when a patient has multiple diseases?

    ●    How often is the system wrong—and how is it wrong?

False positives and false negatives do not have identical consequences. A missed diagnosis can cause harm, but large numbers of false positives can also lead to more testing, biopsies, expense, anxiety, and unnecessary treatment. “Accuracy” alone does not tell us everything we need to know.

There is also the problem of automation bias. When a computer gives us an answer, we tend to grant it credibility—particularly when it is expressed confidently. If an algorithm reports a 93% probability of pulmonary metastasis, how much does that influence the veterinarian’s independent interpretation?

Machines will make mistakes, just as veterinarians do. The greater danger is that we stop questioning the machine.

Will AI Replace Veterinarians?

I do not think that is the most useful question.

Parts of veterinary work probably will become automated. Computers may become better than humans at recognizing certain patterns in medical images, laboratory results, or gait data. That is not necessarily a threat, because diagnosis is not the same thing as medicine.

Suppose an algorithm reports an 82% probability that a dog has a particular disease. What happens next?

Should we perform another test? Which one? Should we treat now or wait? What are the consequences if the algorithm is wrong? How do age, concurrent disease, medication use, cost, and the family’s goals change the recommendation? How certain should we be before recommending surgery? What if the algorithm conflicts with the clinical examination?

Those are not merely pattern-recognition problems. They are judgment problems.

Judgment may become one of the veterinarian’s most valuable skills.

The Veterinarian May Become the Interpreter of Technology

Veterinarians once practiced with relatively little diagnostic information. Today, we have an enormous amount. By 2035, we may have too much.

A patient record could contain continuous activity and sleep data, heart- and respiratory-rate trends, weight changes, gait analysis, laboratory trajectories, imaging findings, AI-generated differential diagnoses, risk scores, genomic information, drug-response predictions, and automated disease alerts.

The problem will no longer be, “How do I get more information?” It will be, “Which of this information actually matters?”

Veterinarians will become less valuable as repositories of facts. No human can compete with a computer at storing information. Our value will lie elsewhere: understanding context, recognizing uncertainty, evaluating evidence, identifying when the data do not make sense, communicating risk, and balancing benefit, harm, cost, and quality of life.

We must understand the patient and the person caring for that patient. We must make decisions when there is no single correct answer.

Ironically, the more technology enters veterinary medicine, the more valuable those human skills may become.

Who Owns the Data?

If a dog’s collar collects health data 24 hours a day, who owns that information? Is it the pet owner, veterinary hospital, or company that manufactured the device?

Can the company use those data to train an algorithm? What happens if the company disappears? Can another veterinary platform access the information? Can 10 years of patient data move into a different practice-management system?

And when an algorithm makes a recommendation, do we know why?

Veterinarians do not need to become computer scientists, but they will need to understand the clinical limitations of these systems. What population trained the model? Which animals were excluded? What was the model designed to predict? Does that outcome matter to the patient in front of us?

The veterinarian of 2035 may need to understand data quality almost as well as diagnostic-test quality.

The Veterinary Appointment Could Become More Personal

Much of today’s appointment is devoted to reconstructing what happened. When did the problem begin? Is the animal eating, drinking, vomiting, limping, or sleeping normally?

Imagine if much of that information were already summarized before the patient entered the hospital:

> Activity decreased 11% beginning six weeks ago. Body weight declined 4%. Nighttime respiratory rate increased. Medication compliance is 92%. Video analysis suggests progressive right pelvic-limb asymmetry.

Instead of spending half the appointment trying to establish a timeline, we could spend more of it deciding what to do.

Paradoxically, technology may make veterinary appointments more personal, not less. Technology handles some of the measurement; the veterinarian spends more time interpreting, explaining, and helping the family make decisions.

The Biggest Innovation May Be Convergence

If you ask me to name the biggest veterinary innovation of the next five to 10 years, I would not choose AI, wearables, genomics, or a new class of drugs in isolation.

I would choose the convergence of all of them.

Imagine a patient whose health is continuously monitored at home. Subtle changes are detected before disease becomes clinically obvious. AI integrates those signals with years of medical records, laboratory results, imaging, and genetic information. The veterinarian receives an alert and investigates earlier. Treatment is selected not only according to the name of the disease, but according to the biology of that individual animal. Once treatment begins, we continuously measure whether it worked.

That is a fundamentally different model of veterinary medicine:

    ●    From episodic medicine to continuous medicine

    ●    From reactive medicine to potentially predictive medicine

    ●    From treating populations toward treating individuals

But one enormous condition applies to everything I have described: we have to prove that it works.

Not in a technology company’s demonstration. Not in a carefully selected dataset. Not simply because an algorithm generated an impressive accuracy statistic.

We need prospective studies, independent validation, real-world clinical populations, and comparisons with existing standards of care. Ultimately, we must show that these technologies do more than generate information. They must produce better outcomes for animals.

Innovation should never be measured by how impressive the technology appears. It should be measured by whether our patients are better because we used it.

Technology will become extraordinarily good at giving us answers. Our job will be to make sure we are still asking the right questions.

What Do You Think?

Which technology will change veterinary medicine the most over the next decade: artificial intelligence, wearable devices, precision medicine, gene therapy, new cancer treatments, continuous monitoring—or something completely different?

I plan to take several of these technologies and examine them individually—not simply what companies claim they can do, but what the science actually shows. What is real? What is hype? What are the limitations? And what might genuinely change the way we practice veterinary medicine?

Veterinary medicine in 2035 will look different. The important question is whether we will use new technologies simply because we can—or demand evidence that they actually make our patients better.