AI is entering veterinary practice. Here's what vets want AI tools to do, what they refuse to let them do, and the standards the profession is now setting.
Artificial intelligence is arriving in veterinary practice, and the profession has some very clear ideas about how it should, and should not, be used.
Imagine you bring your dog in for an appointment. The vet listens carefully, examines her, considers the history. Then they turn to the computer, where an AI tool has drafted the consultation notes, flagged a differential diagnosis worth considering, and pre-populated the prescription form. The vet reviews all of it, makes their own judgment, and explains the plan to you.
That is a version of AI in veterinary practice that most vets would welcome. It saves time, reduces the administrative burden of a demanding job, and supports rather than replaces the clinical thinking that good animal care depends on.
But there is another version: one in which AI generates a diagnosis without clear explanation, clinical records are processed by an algorithm trained on unknown data, or a triage tool tells a worried owner to wait when they should not. That version worries the profession deeply.
The difference between these two scenarios is precisely what the veterinary sector is now working to define, and the answers are more specific, and more instructive, than you might expect.
Veterinary professionals are not, by and large, hostile to artificial intelligence. A major survey published in 2025 in the American Journal of Veterinary Research found that 43% of respondents were more optimistic than skeptical about AI in veterinary medicine, and that familiarity with the technology was the single strongest predictor of that optimism. In other words, the more vets knew about AI, the more positively they tended to view it (Pham et al., 2025).
And yet only around 39% of veterinary professionals were actually using AI tools in their practice at the time of the survey (Digitail/AAHA, 2024). Something is sitting in the gap between interest and action, and it is not technophobia. It is a set of very reasonable professional questions that, until recently, the AI industry has not been answering well enough.
That is beginning to change. The American Veterinary Medical Association established a Task Force on Emerging Technologies and Innovation to develop guidance for practitioners navigating these questions (AVMA, 2024). Internationally, the Royal College of Veterinary Surgeons, the UK’s primary veterinary regulator, convened over a hundred stakeholders for a dedicated roundtable on AI governance in May 2024 (RCVS, 2024), and the British Veterinary Association published a formal policy position on AI in December 2025 (BVA, 2025).
Together, these developments amount to the profession saying, clearly and collectively: we are interested in this technology, and here is what it needs to look like before we trust it.
The use cases where veterinary professionals are most enthusiastic are also, not coincidentally, the ones that carry the lowest clinical risk.
At the top of the list is administrative support. Vets spend a significant portion of their working day on documentation: clinical records, referral notes, prescription management, and follow-up communications. AI tools that can transcribe a consultation, draft a clinical note, or automate appointment reminders are already in use in some practices, and the professional response has generally been positive. The Digitail/AAHA survey found that improving administrative efficiency was the most widely cited benefit of AI, above clinical applications (Digitail/AAHA, 2024). For a profession grappling with serious workforce pressures and burnout, that matters.
Diagnostic imaging is the second major area of genuine clinical enthusiasm. AI systems trained to analyze radiographs, ultrasound images, and histopathology slides are becoming more sophisticated, and the peer-reviewed evidence base is building. Vets working with high volumes of imaging, or in settings where specialist review is not immediately available, see real potential in tools that can flag abnormalities, prioritize cases, and provide a structured second opinion. What they ask for in return is honesty about how the tool performs: its accuracy, its known limitations, and the species, breeds, and case types it has been validated on (PMC, 2024).
Triage support is a third area of growing interest. AI-assisted symptom checking, whether for telephone triage, online inquiries from pet owners, or in-practice prioritization, can help manage demand in busy practices. Here again, the appetite is real but conditional. Vets want tools that direct cases to a professional rather than away from one, and that are transparent about the limits of what a symptom-checking algorithm can and cannot determine.
If the “want” list is broadly about support, efficiency, and evidence-based assistance, the “do not want” list is built around a single principle that runs through virtually every piece of professional guidance published on this topic: AI must not replace clinical judgment.
Veterinarians bear responsibility for the care they provide. When an AI tool generates a diagnosis and that diagnosis is wrong, or when a plausible-sounding output is accepted without adequate scrutiny, the consequences fall on the animal, the client, and the licensed professional, not the software company. The AVMA and its counterparts in the UK and beyond have all been explicit on this point: AI is a tool to support veterinary decision-making, not a substitute for it (AVMA, 2024; RCVS, 2024; BVA, 2025).
Related to this is a deep concern about over-reliance. In fast-moving clinical settings, plausible outputs can be given undue weight. A 2026 article in Frontiers in Veterinary Science identified the risk that AI framing, specifically the way results are presented and the confidence with which they are expressed, can subtly encourage passive acceptance rather than active professional review (Frontiers in Veterinary Science, 2026a). Good tools are designed with this risk in mind. They present outputs as prompts for professional consideration rather than conclusions, include uncertainty flagging, and are built to be questioned.
Opacity is equally unwelcome. Many AI systems operate as “black boxes”, producing outputs without explaining how they arrived at them. In a consumer context, this may be tolerable. In a clinical context, it is not. Vets need to understand what a tool was trained on, how it was validated, and, most critically, where it is likely to fail. A landmark 2026 audit of commercial veterinary AI products found a significant transparency gap: a wide disparity between the clinical capabilities claimed in marketing materials and the publicly available documentation needed to actually evaluate those claims (Frontiers in Veterinary Science, 2026b). The profession has noticed, and it is not impressed.
Data governance is the third major area of concern. The Digitail/AAHA survey found that 53.9% of respondents cited data security and privacy as a significant worry about AI in practice (Digitail/AAHA, 2024). Pet owners have a legitimate interest in this too. When a vet uses an AI tool during a consultation, the animal’s health data, clinical history, and owner details may be processed, stored, or in some cases used to train or refine the AI model itself. Unlike human healthcare, veterinary practice currently operates without a sector-specific federal data protection framework in the United States, which means that the responsibility for asking the right questions falls squarely on practices and the professionals running them. Vets want clear answers from providers on where data goes, who can access it, how long it is kept, and whether client data may be used for purposes beyond the immediate clinical task. The profession is beginning to require those answers as a condition of adoption.
Finally, vets want consistency. An AI tool that performs well in testing but updates silently, or behaves differently between versions without explanation, cannot be safely embedded in clinical practice. Version control, transparent update processes, and revalidation commitments are not technical niceties; they are professional necessities.
What makes the current moment distinctive is that the profession is not simply expressing preferences. It is building the tools to hold AI providers accountable.
This is the direction of travel. The AVMA’s task force is developing guidance that could, in time, shape professional standards, accreditation requirements, and the competencies expected of new graduates entering the profession (AVMA, 2024). State licensing boards are also beginning to consider where AI governance intersects with existing standards of professional conduct. Internationally, the UK’s veterinary regulator has indicated that its own Codes of Professional Conduct may be updated to reflect AI use, a signal of where the broader profession is heading (RCVS, 2024). Across the board, the message is consistent: AI is welcome, but it needs to earn its place.
For pet owners, this is reassuring news. The veterinary profession is not sleepwalking into AI adoption. It is asking hard questions, setting standards, and making clear that clinical responsibility remains exactly where it has always been: with the qualified professional in the room.
The AI tools that will become genuinely embedded in good veterinary practice will be the ones that do the unglamorous work, freeing up time, supporting decisions, and making data easier to manage. They will be honest about what they cannot do. They will be transparent about how they handle sensitive information. And they will always leave the clinical judgment, the part that requires knowledge, experience, and care, to the vet.
That is what the profession is asking for. It is a reasonable ask, and the technology industry would do well to meet it.
PerkyPet is on a mission to do just that. It was built from the outset with input from veterinarians around the world; designed to reduce burnout and surface previously invisible data to support clinical decision-making, not to replace the professionals at the heart of animal care. It is held to the highest standards of transparency and clinical responsibility.
Pham, T. et al. (2025). Familiarity with artificial intelligence drives optimism and adoption among veterinary professionals: 2024 survey. American Journal of Veterinary Research, 86(S1). https://avmajournals.avma.org/view/journals/ajvr/86/S1/ajvr.24.10.0293.xml
Digitail / AAHA (2024). AI in Vet Med: Insights from Digitail & AAHA’s Survey. https://digitail.com/blog/39-2-of-veterinary-professionals-use-ai-tools-in-their-practice-digitail-and-aaha-survey/
Read, J. (2024). What might the growth of artificial intelligence mean for veterinary healthcare? Veterinary Record. https://bvajournals.onlinelibrary.wiley.com/doi/10.1002/vetr.4669
Royal College of Veterinary Surgeons (2024). AI Roundtable report published. https://www.rcvs.org.uk/about-us/news-and-views/news/ai-roundtable-report-published
British Veterinary Association (2025). BVA Policy Position on Artificial Intelligence in the Veterinary Profession. https://www.bva.co.uk/media/6777/bva-position-on-artificial-intelligence.pdf
American Veterinary Medical Association (2024). Building a framework for responsible AI in veterinary medicine. https://www.avma.org/news/building-framework-responsible-ai-veterinary-medicine
Frontiers in Veterinary Science (2026a). Ethical considerations of artificial intelligence in veterinary medicine decision-making. https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2026.1780868/full
Frontiers in Veterinary Science (2026b). A systematic audit of transparency and validation disclosure in commercial veterinary artificial intelligence. https://www.frontiersin.org/journals/veterinary-science/articles/10.3389/fvets.2026.1761038/full
PMC (2024). The potential application of artificial intelligence in veterinary clinical practice and biomedical research. https://pmc.ncbi.nlm.nih.gov/articles/PMC10864457/
