AI as a Veterinarian? Why Artificial Intelligence Can Help but Should Not Treat Your Pet

Using AI for pet health may seem tempting, especially when your animal is unwell and clear answers are difficult to find. Over the past year, I have noticed more and more pet owners turning to AI in the hope of finding answers they have not received elsewhere. This is completely understandable, but it is also where one of the greatest risks lies. AI can be an excellent assistant, but it should not be responsible for treating your pet. The world is changing, and all of us are using AI in one way or another.

In this article, I want to explain how AI can help and why it should never be given responsibility for creating a pet’s treatment plan. A pet owner recently said something that stayed with me: “We have tried to help our dog using Google and AI.”

I understand this completely. When a pet is unwell, previous treatment has not produced the hoped-for result and clear answers seem impossible to find, people naturally begin searching for them themselves. This does not come from carelessness. On the contrary, it often comes from deep concern and a genuine wish to help. In such a situation, AI can be a useful assistant.

AI may help you:

• translate blood test results and veterinary explanations into more understandable language;
• organise your pet’s medical history and create a timeline of results;
• identify which test results have changed over time;
• prepare questions for your veterinarian;
• organise information and instructions after an appointment and identify unanswered questions;
• create a timeline of symptoms, dietary changes and wellbeing that can be shared with your veterinarian.

The problem begins when AI stops being an information assistant and starts creating the treatment plan.

Why AI for pet health cannot see the whole picture

AI only sees the information a person gives it. It cannot physically examine an animal or assess pain, dehydration, posture, the colour of the mucous membranes, breathing, muscle loss, digestive patterns or overall condition. It also may not ask enough detailed follow-up questions to reveal whether an important piece of information has been left out.

AI may not know every diagnosis, medication, previous reaction or change in test results. It cannot reliably determine which problem needs the most urgent attention right now. In some situations, for example, the most important first step may simply be helping the animal start eating again. Only after that can the next supportive steps be considered.

There is another important point. An AI response can sound extremely confident even when a significant part of the picture is missing. That is what makes it difficult. An answer may sound logical, calm and well reasoned, but that does not automatically make it the right next step for a particular animal.

The most dangerous recommendation may not be entirely wrong

A particular product, supplement or new food may be good in itself. That does not mean it is appropriate for this animal, at this stage of illness, together with this medication or before another problem has been stabilised.

The order of changes matters when supporting a pet’s health. If we change the food and medication while adding several supplements at the same time, we can no longer tell what helped, what did not suit the animal or what made the situation worse. Sometimes the most important first step is helping the animal begin eating again. Digestion, weight, nutrient intake, the microbiome and other supportive measures can then be addressed gradually.

More interventions do not always lead to a better outcome. Quite often, the opposite is true. The important thing is to do the right things, in the right order and for the right reason.

The way AI is used makes a major difference

People often talk about AI as though everyone were using the same tool at the same level. In reality, there can be significant differences:

• whether a free or paid version of an AI tool is being used;
• how thoroughly the situation has been described;
• how accurate the information provided is;
• whether the AI has received test results, medication details, a timeline of symptoms and dietary changes;
• whether the user can critically evaluate the answer;
• whether the AI has already been given background information, corrections and clarification during a longer conversation.

A paid version does not automatically make every answer correct, just as a free version does not mean that no useful information can be obtained. In more complex health and nutrition questions, however, the model’s capabilities, available context and previously provided information can make a major difference to the relevance of its answer.

The same AI can produce answers of very different quality depending on how it is used. AI is not simply a search engine. It needs well-formulated questions, sufficient background information and a user who is willing to assess its first answer critically.

Even then, AI should never be the only decision-maker. The question is not simply whether we should use AI. What matters is which tool is being used, what information it has received and whether the person reading the answer has enough knowledge to evaluate it. AI does not become a reliable veterinarian merely because its answer sounds convincing.

I never treat an AI answer as the final truth

I also use AI to analyse information, organise my thoughts and examine a situation from different angles. My recommendations, however, are never based on a single AI response.

In fact, I challenge AI quite often. When it suggests something, I ask follow-up questions, question its logic, correct its assumptions and direct it back to the circumstances of the individual animal. I frequently adjust the ideas it offers because they may be too general, too quick or may combine several changes that should not be introduced at the same time in real life.

A real-life example

I recently discussed feeding a dog raw egg with AI. My dog Faddy currently has sensitive digestion and is very selective about food, but he tolerated raw egg well. My question was not whether a dog could be given egg at all. I wanted to understand what a sensible limit might be, considering how often it would be offered, the amount and Faddy’s current condition.

AI initially responded with a typical general warning. It said that raw egg may contain bacteria and that feeding large amounts of raw egg white over a long period may affect biotin absorption. These statements were not exactly wrong, but the answer failed to consider the context.

I pointed out that we were discussing a whole egg rather than the continuous feeding of large quantities of raw egg white. I also provided the amount, frequency and details of the dog’s overall diet. AI then adjusted its answer. Avidin in egg white can bind biotin, but the yolk is also a good source of biotin. The risk of biotin deficiency is mainly associated with feeding large quantities of raw egg white over a long period, rather than occasionally offering a whole raw egg.

The bacterial risk was also initially presented out of proportion, as if one fresh raw egg automatically created an exceptional danger for a dog. People also eat raw egg in certain dishes, such as steak tartare served with raw egg yolk.

This does not mean that there is no bacterial risk. It means that the word “raw” alone does not automatically make a food unsuitable. The risk depends on factors such as the quality and freshness of the egg, storage, hygiene, quantity and frequency of feeding. A dog’s digestive system is also adapted differently to process animal-based food. This does not make dogs immune to foodborne pathogens, but it does mean that precisely the same risk assessment cannot always be applied to humans and dogs.

This is a good example of how AI may repeat a general warning without immediately considering the quantity, frequency, overall diet and health of the individual animal. Its most dangerous mistake may not be an incorrect fact. The problem may arise when a correct piece of information is presented without context or applied to the wrong animal, amount or situation.

The problem is often not the information but how it is applied

AI may provide entirely accurate pieces of information. Problems arise when they are applied in the wrong amount, in the wrong circumstances or to the wrong animal.

I often consult my own study materials and information shared by integrative veterinarians, and I may also provide that material to AI so that it can take those perspectives into account. Conventional and integrative veterinary medicine may approach certain questions differently.

I therefore try to consider different sources and bring them together in the context of the individual animal. A great deal of information available online is partly correct but presented without considering dosage, the stage of illness or the animal’s complete situation. I rarely rely on the recommendation of only one source or one specialist. The more complex the case, the more important it is to examine different perspectives and build a coherent picture.

AI can help us think, but a person must still be able to judge whether its answer is practically useful. That requires knowledge, experience and the ability to see the wider picture.

For the past seven years, I have continuously studied animal nutrition, digestion and ways of supporting the body. During that time, I have worked through thousands of pages of educational material. This knowledge gives me a framework for assessing:

• whether the information is relevant to a particular animal;
• which steps are currently the highest priority;
• which recommendations can wait;
• which changes should not be made at the same time;
• when a veterinarian must be involved.

Without this background, AI recommendations can easily become a long list of foods, supplements and possible solutions that are all introduced at once in the hope that one of them will work.

When it feels as though the options are running out

People often contact FendaF only after the conventional approach has not produced the hoped-for result or when they feel that they are running out of options.

In these situations, we look at the existing information as part of a broader picture. We draw on experience, nutritional knowledge and the thinking used by integrative veterinarians, where the animal is considered as a whole rather than focusing only on an isolated symptom.

A veterinary appointment is often only one moment in a longer journey. The owner may receive a great deal of information during the visit, but at home they must put those recommendations into practice, monitor the animal’s reactions and decide which questions should be discussed at the next appointment. At this stage, an owner may need someone who can help them calmly work through the information, remain consistent with the agreed approach and avoid introducing too many changes at once.

We can help owners consider:

• where to begin;
• which steps can wait;
• what should be changed one step at a time;
• how to monitor the animal’s response;
• which observations should be shared with the veterinarian;
• which questions should be discussed at the next appointment;
• when the veterinarian should be contacted again.

Sometimes an owner needs more than information. They may need someone who listens, helps organise their thoughts and keeps the process on a sensible path. The aim is not to replace the veterinarian, but to bring together the veterinarian’s recommendations, the animal’s daily wellbeing and the owner’s observations.

At FendaF, we do not diagnose illnesses or change medication prescribed by a veterinarian. We help owners understand the available information, make better-informed choices and support their pets thoughtfully, one step at a time. When necessary, we can also help formulate questions to discuss with the veterinarian and consider which changes have been observed since the appointment.

We do not replace veterinarians. We help owners apply the knowledge received from their veterinarian in everyday life.

Use AI wisely

AI for pet health cannot replace a physical examination, medical responsibility or the professional ability to assess the whole picture. It can, however, be a valuable assistant for organising information, preparing questions and understanding an agreed plan.

The most important goal is not to collect as many recommendations as possible. It is to understand which step is appropriate now, which can wait and when help is needed from someone who can assess the whole situation.

Supporting health does not always mean doing more. It means doing the right things in the correct order and for a clear reason. An animal’s condition will not necessarily improve because someone finally discovers the “perfect” food, supplement or treatment protocol. More lasting change is often supported by moving away from constant quick fixes and following a thoughtful, gradual process.

First consider which area needs the most attention now. Choose one area, not five at the same time. Make one considered change and give the animal’s body time to respond. These changes do not have to be large or dramatic. They are often improvements to the basic foundation, and a stronger foundation can support more lasting results.

When we stop jumping from one solution to another and begin seeing the connections, the entire process becomes calmer. Every setback no longer feels like a new emergency, every decision does not need to be questioned endlessly and the animal receives something its body often needs most: consistency.

Recovery does not come from constantly changing direction. It develops through a considered plan, correctly ordered steps and enough time for the body to respond. AI may help us understand those steps, but choosing them still requires knowledge, experience and the ability to see the whole picture.